QOS Bkgpaper

INTERNATIONAL TELECOMMUNICATION UNION
TELECOMMUNICATION DEVELOPMENT BUREAU

Document: 2

Global Seminar on Quality of Service and Consumer Protection
Geneva, Switzerland, 31 August-1 September 2006

ICT Quality of Service Regulation:
Practices and Proposals

BACKGROUND PAPER
FINAL VERSION

September 2006

International Telecommunication Union (ITU)

This background paper was prepared by Robert Milne ([email protected]), Antelope Consulting,
United Kingdom. The views expressed in this paper are those of the author and do not necessarily
reflect the views of ITU or its members.
The author wishes to express his sincere appreciation to Luis Cardozo, Bobby Hannan, Mberra
Gngr, Claire Milne, Sebastian Moeller, Enrique Prez, Nicolas Silva, Nancy Sundberg, Volker Sypli
and Eric Yam.
This is one of two background papers to be presented for comments at the ITU Global Seminar on
Quality of Service and Consumer Protection, Geneva, Switzerland, 31 August-1 September 2006.
Further information can be found on the web site at http://www.itu.int/ITU-D/treg.

2006 ITU

International Telecommunication Union
Place des Nations
CH-1211 Geneva, Switzerland

3
Contents
Summary..................................................................................................................................................6
1 The motivation for quality of service monitoring..........................................................................7
1.1 Aims.....................................................................................................................................7
1.1.1 Helping customers to make informed choices.........................................................7
1.1.2 Checking claims by operators.................................................................................8
1.1.3 Understanding the state of the market.....................................................................8
1.1.4 Maintaining or improving quality in the presence of competition..........................8
1.1.5 Maintaining or improving quality in the absence of competition............................8
1.1.6 Helping operators to achieve fair competition........................................................9
1.1.7 Making interconnected networks work well together .............................................9
1.2 Ways of achieving the aims.................................................................................................9
1.3 Activities............................................................................................................................11
1.4 Technicalities.....................................................................................................................12
2 The operational arrangements for quality of service monitoring.................................................15
2.1 Defining measurements.....................................................................................................15
2.2 Setting targets....................................................................................................................15
2.3 Making measurements.......................................................................................................18
2.4 Auditing measurements.....................................................................................................19
2.5 Publishing measurements..................................................................................................20
2.6 Ways of sharing responsibilities........................................................................................23
2.7 Ensuring compliance..........................................................................................................26
2.8 Relating quality to profitability..........................................................................................30
2.9 Resolving quality problems directly..................................................................................31
3 The selection of areas of focus for quality of service monitoring................................................32
3.1 Measurements to be defined..............................................................................................32
3.1.1 Importance to customers........................................................................................32
3.1.2 Practicality for operators.......................................................................................34
3.1.3 Comparability between operators..........................................................................34
3.2 Targets to be set.................................................................................................................36
3.2.1 Usefulness to customers........................................................................................36
3.2.2 Realism for operators............................................................................................37
3.3 Measurements to be published...........................................................................................38
3.3.1 Accessibility to customers.....................................................................................38
3.3.2 Helpfulness to customers.......................................................................................39
3.3.3 Fairness to operators..............................................................................................39

4
3.4 Measurements to be made..................................................................................................40
3.4.1 Objective and subjective methods.........................................................................40
3.4.2 Passive and active methods...................................................................................41
3.4.3 Network-by-network and end-to-end methods......................................................42
3.5 Services to be monitored....................................................................................................43
3.5.1 Layers in services..................................................................................................43
3.5.2 Network segments used by services......................................................................45
3.5.3 Aspects of services................................................................................................46
Conclusion..............................................................................................................................................48
Abbreviations.........................................................................................................................................49
Endnotes.................................................................................................................................................50

Box 1 The meaning of quality of service.............................................................................................. 12
Box 2 The nature of measurements and targets..................................................................................... 13
Box 3 Conventions for measurement names......................................................................................... 14
Box 4 Monitoring by the regulator in India........................................................................................... 17
Box 5 Monitoring by the regulator in Chile.......................................................................................... 18
Box 6 Monitoring by the regulator in Ireland....................................................................................... 20
Box 7 Monitoring by the regulator in Brazil ......................................................................................... 22
Box 8 Monitoring by the regulator in Australia.................................................................................... 28
Box 9 Monitoring by the regulator in the United Kingdom.................................................................. 29
Box 10 Monitoring by the regulator in Colombia................................................................................. 30
Box 11 Relations between quality of service and regulation of profitability in the United States........ 31
Box 12 The evolution of quality of service monitoring in the EU and OECD countries...................... 33
Box 13 Some sources of measurements and targets from ETSI and ITU standards............................. 38
Box 14 Measurements for voice, video and text................................................................................... 41
Box 15 Measurements related to communication................................................................................. 47

Figure 1 Activities during quality of service monitoring...................................................................... 11
Figure 2 Examples of responsibilities for defining measurements and setting targets.......................... 16
Figure 3 Examples of publishable measurements arranged in a bar chart............................................ 21
Figure 4 Techniques for ensuring fulfilment of quality of service obligations..................................... 26
Figure 5 Analyses of complaints to regulators in two countries in Africa............................................ 34
Figure 6 Examples of measurements across three networks................................................................. 43
Figure 7 The transformation from current networks to next generation networks................................ 44
Figure 8 Access and core networks....................................................................................................... 45

5
Table 1 Direct contributions of quality of service monitoring to achieving the aims............................. 9
Table 2 Direct benefits to stakeholders of quality of service monitoring............................................. 10
Table 3 Examples of publishable measurements arranged in a table.................................................... 21
Table 4 Disadvantages of allocations of responsibility in quality of service monitoring..................... 24
Table 5 Examples of allocations of responsibility in quality of service monitoring............................. 25
Table 6 Examples of apparent impairments possibly not generating fault reports................................ 35

6
Summary
For many years some telecommunications regulators have had schemes for quality of service
monitoring that recognised its importance in their markets. More recently other regulators have been
drafting regulations and, in some cases, publishing results. Yet there have been various apparent
setbacks: various regulators have stopped their work on quality of service after publishing consultation
papers, regulations or initial quality of service measurements. The problem overall may be that the
resources of regulators are insufficient: regulators have many urgent and important tasks and require
time to appreciate fully the aims of, and tasks in, quality of service monitoring.
Several regulators now have plans to devise new schemes or revise existing schemes for quality of
service monitoring. This paper is intended to help those and other regulators, by starting to study how
experience so far can guide quality of service monitoring in the future, particularly in countries where
it is not yet well established.
Section 1 looks at the question of when quality of service monitoring by regulators is desirable.
Broadly speaking, the section concludes that quality of service measurements can be beneficial in
several ways, at least if they are published for all operators, but that quality of service targets can be
detrimental unless they are applied only to dominant operators. Though there are exceptions to this
conclusion it suggests that fewer targets may be needed than are currently imposed.
Section 2 deals with operational arrangements for quality of service monitoring by regulators. There
are various ways in which operators can be involved in the arrangements for defining measurements,
setting targets, making measurements, auditing measurements and publishing measurements. There
are also ways in which customers can be involved, even in defining measurements and making
measurements. The cases studied illustrate wide variations in how quality of service monitoring is
done. The more and less successful aspects of the cases are not always what might be expected.
Section 3 turns to the question of what quality of service monitoring by regulators is desirable. Many
measurements have been defined and many targets have been set. Regulators and operators need to
choose those that match their main aims. The section describes criteria for choosing the measurements
to be defined, the targets to be set and the measurements to be published. It also discusses various
other choices that regulators and operators must make, such as how measurements are to be made
(particularly for voice and related means of communication) and which services are to be monitored in
current network and Next Generation Network (NGN) environments. These choices are particularly
important as communication services become layered on Internetwork Protocol (IP) networks, when
operators will use wholesale services, and bundle retail services, both within and between layers.
Quality of service monitoring inevitably differs in different countries, with their own industry
structures and customer preferences. It is unlikely ever to be perfect, as regulators are unlikely ever to
have full enough information about the markets. This paper does not pretend to say what regulators
always ought to do. However, it does suggest the following overall guidelines for discussion:
Using operator skills and customer opinions through widespread consultations, working groups and
open meetings helps to make monitoring effective.
The measurements made should be important to customers, practical for operators and comparable
between operators. They should concentrate on few aspects of services.
The measurements published should be accessible to customers, helpful to customers and fair to
operators. They should be made available in ways appropriate to the culture of the intended users.
Any targets set should be useful to customers and realistic for operators. They are most likely to be
desirable for wholesale services (and often retail services) of dominant operators.
Monitoring should entail regularly examining, and understanding the basis for, the measurements.
Some of these guidelines may seem too obvious to discuss, but they are not followed everywhere.

7
1 The motivation for quality of service monitoring
This section examines the possible aims of quality of service monitoring by regulators and makes
inferences about the conditions in which quality of service measurements and quality of service targets
are beneficial. It includes discussions of the activities and concepts in quality of service monitoring.
1.1 Aims
The ITU World Telecommunication Regulatory Database indicates that in 2005 regulators had
responsibilities for quality of service monitoring in about 84% of countries and for quality of service
standards in about 76% of countries
1
. Thus quality of service monitoring could have important effects
on the development of telecommunications worldwide.
Quality of service monitoring requires that measurements be made and, sometimes, that targets be
attained. As operators often make measurements for themselves already, action by regulators could be
unnecessary. However, regulators might monitor quality of service for several reasons. Some are more
important when competition is strong, and others are more important when competition is weak.
Quality of service monitoring by regulators can have the following aims:
Helping customers to make informed choices.
Checking claims by operators.
Understanding the state of the market.
Maintaining or improving quality in the presence of competition.
Maintaining or improving quality in the absence of competition.
Helping operators to achieve fair competition.
Making interconnected networks work well together.
These aims are discussed individually below.
1.1.1 Helping customers to make informed choices
The price is an important factor in choosing a service. However, it is not the only one: customers may
care about price more than quality, but once they have settled on the price they want the best quality
available at that price. In fact quality can be more important than price. Thus in Colombia a survey
showed that the most important factor in choosing a mobile telephony operator was coverage for 28%
of customers, quality of service for 24% of customers and promotions for 21% of customers
2
.
Business customers often care more about quality than price, because problems with quality are more
likely to be costly. For example, in the United Kingdom (UK) 59% of small and medium businesses
rated the offer of the best quality and reliability, not the cheapest deal possible, as the most
important factor when choosing an operator
3
. Even 55% of residential customers of the main
incumbent operator valued quality and reliability above price
4
.
Different quality levels can coexist at different prices: in India each long distance operator can have
two call-by-call selection codes representing different quality levels, in Australia each fixed telephony
operator can offer services without the customer service guarantee as well as ones with it, and in many
countries different internet cafes or different calling cards have different quality levels. In such cases
customers need to know about quality levels but not to have them constrained.
For services that are bundled together, with one price covering several services, the quality of specific
aspects of the services can influence choices greatly: customers are likely to choose bundles that have
the best versions of those services that are most important to them.
Measurements that are published can help customers to choose well. Targets that are the same for all
operators, however, may limit choices of quality and price.

8
1.1.2 Checking claims by operators
Operators may make claims in advertisements about their services or the services of their competitors.
An example occurred in the UK, when two mobile telephony operators complained to the advertising
regulator that a third mobile telephony operator advertised the best call success rate ever. The
advertisements were based on measurements that were made using a method adopted jointly by the
three operators. Earlier measurements that were made using this method had been published. Most of
the complaints were upheld, because the third operator had ignored unpublished but audited
measurements for the other two operators, had implied that the telecommunications regulator had
published the measurements, and had identified the call success rate with network performance
5
.
(They also included a discussion of whether call success rates of 99.2% and 99.3% were different for
customers in practice.)
Measurements that are published can be used in checking claims by operators about services.
1.1.3 Understanding the state of the market
Figures about rollout may not be enough to show how well policies are succeeding. (For example,
numbers of base stations might reach required levels but the base stations might not be maintained
well enough to operate properly.) They are therefore of interest to policy makers. They are also of
interest to operators, who wish to understand strengths and weaknesses of competitors.
Measurements can show gaps in services that could be filled by market entry or that need new policies
for particular groups of people, geographic areas or operating conditions (such as emergencies).
1.1.4 Maintaining or improving quality in the presence of competition
Competition can improve quality. However, cutting prices is often the preferred way of competing,
partly because it can be done much faster than improving quality. When cutting prices involves cutting
costs, competition can even reduce quality. Quality reductions due to cuts in costs can be difficult to
reverse, as new staff may need to be trained or deferred investments may need to be brought forward.
Even in fully competitive markets quality may be poor, because of ferocious competition, rapid
expansion or overstretched management. Thus in various countries both customer support and network
equipment can become overloaded in aggressive campaigns for internet access customers, for
example. Similarly, during rapid expansion of the mobile networks in India quality fell well short of
the targets
6
.
Measurements that are published, perhaps with targets, can oppose forces that can tend to lower
quality levels in some competitive environments.
1.1.5 Maintaining or improving quality in the absence of competition
A dominant operator may be subject to price controls, unwilling to boost supply or insensitive to
customer wishes. In such circumstances the operator might try to maintain margins by reducing the
quality of a retail service. (For example, public payphone faults and directory enquiry call failures
might rise because of cuts in staff.) The operator might even introduce a service at a higher quality
level and lower the quality level of the standard service to encourage customers to move to the
better service
7
.
In addition, a dominant operator might suggest that to protect customers any quality of service
obligations imposed on it should be imposed on other operators also. Doing this might lead to uniform
quality levels but would limit the speed and flexibility with which other operators could offer services.
Measurements that are published, perhaps with targets, for the retail services of dominant operators
can counter the adverse effects of other forces on dominant operators. Targets that are the same for all
operators, however, may act as barriers to entry.

9
1.1.6 Helping operators to achieve fair competition
The business of a competitive operator might depend crucially on the facilities of a dominant operator,
by interconnecting with, or just reselling, them. The dominant operator is then providing a wholesale
service to operators (not a retail service to end users). For retail competition to be fair, a competitive
operator buying a wholesale service from a dominant operator should get the same quality level from
that service as the retail arm of the dominant operator does
8
.
Targets that represent required minimum quality levels for the wholesale services of dominant
operators can achieve this (perhaps in conjunction with other actions, such as separating the
managements of the retail and wholesale arms).
1.1.7 Making interconnected networks work well together
An end-to-end path may pass through several networks, each controlled by a different operator. No
individual operator has responsibility for the end-to-end path, but a dominant operator may have a
serious effect on end-to-end paths set up by other operators. (The effect of any other operator on end-
to-end paths is on average much less serious.)
Targets that determine required minimum quality levels nationally for end-to-end paths can ensure that
one operator (and, in particular, a dominant operator) does not degrade excessively the quality of end-
to-end paths set up by others. Of course, targets for measurements of customer support that are not
dependent on network equipment are not relevant to the quality of end-to-end paths.
1.2 Ways of achieving the aims
The aims in Section 1.1 mention various activities occurring in quality of service monitoring. Those
that contribute directly to achieving the aims of quality of service monitoring are making
measurements, publishing measurements and setting targets. Table 1 indicates approximately which
activities contribute to which aims.

Table 1 Direct contributions of quality of service monitoring to achieving the aims
aim
activity monitored operators
helping
customers to
make
informed
choices
checking
claims by
operators
under-
standing the
state of the
market
maintaining
or
improving
quality in
the presence
of
competition
maintaining
or
improving
quality in
the absence
of
competition
helping
operators to
achieve fair
competition
making
inter-
connected
networks
work well
together
dominant operators only
making
measurements
all operators +
dominant operators only +
publishing
measurements
all operators + + + +
dominant operators only + + +
setting
targets
all operators +
Note: A + occurring in an entry in Table 1 indicates that the activity contributes to achieving the
corresponding aim in Section 1.1. Of course, making measurements is assumed to be required by publishing
measurements and setting targets, but making measurements is not awarded + in Table 1 just because
publishing measurements or setting targets is awarded +.
Source: Robert Milne (Antelope Consulting).

10
The aims in Section 1.1 do not always point to exactly the same conclusions. For instance:
Different aims can suggest different emphases. In particular, helping customers to make informed
choices is important mainly when choices are inconvenient or costly to change. This can happen if
customer equipment is tied to a particular operator, contracts last for years or telephone numbers
are not portable
9
. It therefore occurs with lifetime subscriptions as introduced in India and the
United States (US)
10
. Sometimes, however, customer choices are not inconvenient or costly to
change and do not by themselves justify quality of service monitoring. This is so for call-by-call
selections of alternative operators without long term commitments.
Different aims can suggest different measurements. Thus using a single indicator of customer
satisfaction for each individual operator, instead of multiple measurements, might help potential
customers but not other operators or policy makers
11
.
Different aims can suggest different services for monitoring: For example, helping operators to
achieve fair competition might involve imposing targets on wholesale services but not on retail
services.
Different aims can suggest different treatments of operators: resellers might not be obliged to make
measurements for the aspects of services that they passed to customers unchanged but might be
obliged to publish the measurements for those aspects made by the original operators, just to
inform their own customers.
Quality of service monitoring can become complex. It is likely to be made cheaper, easier and more
effective by concentrating on particular aims and simple schemes. Doing this involves deciding on the
monitoring to be done (making measurements, publishing measurements or setting targets) and on the
operators to be monitored (dominant operators only or all operators). Table 2 offers guidance on
making these decisions, by summarising roughly how different choices benefit various stakeholders.

Table 2 Direct benefits to stakeholders of quality of service monitoring
stakeholders
monitoring monitored operators
policy
makers
customers
other
operators
making
measurements
all operators +
dominant operators only +
publishing
measurements
all operators +++ ++
dominant operators only + + ++
setting
targets
all operators +
Note: A + occurring in an entry in Table 2 indicates that the activity benefits the corresponding stakeholders,
with the number of occurrences of + being loosely the number of aims in Section 1.1 in which the stakeholder
benefits from the activity. Of course, making measurements is assumed to be required by publishing
measurements and setting targets, but making measurements is not awarded + in Table 2 just because
publishing measurements or setting targets is awarded +.

11
Table 2 can itself be reduced to even simpler guidelines. In brief:
Making measurements for all operators could benefit policy makers.
Publishing measurements for all operators could benefit customers.
Setting targets for dominant operators could benefit other operators.
These guidelines do not mention universal service obligations, which often include minimum quality
levels. In reality universal service providers (the operators having universal service obligations) are
dominant operators, either through being major forces in the national market or through serving areas
where there are few alternatives. In this paper universal service providers are regarded as dominant
operators.
1.3 Activities
Figure 1 illustrates a complete process for quality of service monitoring by regulators, in which the
activities mentioned in Section 1.2 can recur several times. (The flow chart shows how activities
succeed each other and are repeated periodically and gives alongside each activity an example of the
questions or statements that form the subject of the activity.)

Figure 1 Activities during quality of service monitoring
Note: The repeat loops in Figure 1 are simplifications, as auditing, for example, may reveal defects in
measurement definitions that can be corrected without formal review and consultation processes. The numbers
used as examples in Figure 1 are not intended to represent best practice.
reviewing
achievements
consulting
stakeholders
repeat
every
review
period
repeat
every
reporting
period
making
measurements
publishing
measurements
auditing
measurements
defining
measurements
setting
targets
Should the fault repair time continue to be measured?
Should the fault repair time be measured?
The fault repair time has been measured to be 1.357 days.
Has the fault repair time been measured accurately enough?
The fault repair time is the average time taken to repair reported faults.
The fault repair time should be no more than 1.5 days.
reviewing
achievements
consulting
stakeholders
repeat
every
review
period
repeat
every
reporting
period
making
measurements
publishing
measurements
auditing
measurements
defining
measurements
setting
targets
Should the fault repair time continue to be measured?
Should the fault repair time be measured?
Has the fault repair time been measured accurately enough?
The fault repair time is the average time taken to repair reported faults.
The fault repair time should be no more than 1.5 days.

12

1.4 Technicalities
There is not one meaning of quality of service that is universally adopted. However, the one chosen
for this paper is widely adopted. Box 1 says what it is and identifies various related terms that are not
used in this paper.

Box 1 The meaning of quality of service
In ITU-T recommendation E.800, quality of service is the collective effect of service performances, which
determine the degree of satisfaction of a user of the service
12
. This definition is adopted in this paper. Quality
of service therefore concerns aspects of services that users experience directly. It can be contrasted with
network performance, which, again according to E.800, is the ability of a network portion to provide the
functions related to communication between users"
12
.
Different terminology is used in some documents from ITU and other organisations. In particular:
The term quality of service may refer to just quality determined by network performance or even just to a
way of getting different levels of IP network performance by classifying and controlling traffic. The term
quality of experience is then sometimes be used for what is regarded as quality of service in this paper.
The term type of service was formerly used in IP networks but has now been superseded by other terms
related to classifying and controlling traffic.
The term class of service may describe a level of network performance appropriate to particular services
in IP and other data networks.
The term grade of service has been used for many years. In E.600, grade of service is a number of
traffic engineering variables used to provide a measure of adequacy of a group of resources under
specified conditions
13
. It often refers to a value characterising the quality of one aspect of a service (such
as the proportion of successful calls).
There is no firm boundary around what constitutes quality of service: jitter and wander for leased lines could
be measured in quality of service monitoring if they were important to users. In practice, however, quality of
service monitoring is applied to rather few services.

For the purposed of this paper, when quality of service is monitored there must at least be
measurements that are made according to defined measurement methods. These measurements allow
discussions of quality of service to be quantitative and to recognise that different aspects of a service
may reach different quality levels (as, for example, faults may be repaired fast but the proportion of
successful calls may be low).
As indicated in Section 1.2, measurements are not just: some may be published and some may have
targets. Box 2 describes how measurements and targets are related.

13
Box 2 The nature of measurements and targets
Quality of service monitoring requires measurements that characterise the quality of aspects of services. The
definitions of the measurements to be made are called parameters, indicators, metrics, measures or
determinants in different documents. The term parameter is generally used in standards but is avoided in
this paper for simplicity. The term indicator is often used for measures like the number of main telephone
lines per 100 inhabitants that are outside the scope of this paper.
A target is a potential value (or range of values) for a measurement that must be attained if quality is to be
regarded as satisfactory. Targets are called objectives, benchmarks, thresholds, standards or reference
values in different documents. In addition, all these terms have other meanings. The term target is
sometimes confined to desired or planned quality levels, typically in successive years. Then the term
threshold might refer to quality levels that are required immediately, but when used in that way it has the
disadvantage that for many measurements high values are bad and low values are good (so lintel would be
more appropriate than threshold).
Defining measurements involves providing measurement descriptions and measurement methods. A
measurement description says what quality is characterised by the measurements. Often a shortened form of
the measurement description is used as the measurement name, but the full measurement description is needed
for showing when measurements are comparable. Thus the measurement descriptions the proportion of faults
repaired within 2 days and the proportion of faults repaired within a time identified by the operator
evidently refer to measurements that are not usually comparable with one another, but the measurement name
the fault repair timeliness might apply to either of them.
Making measurements involves applying measurement methods. Some measurements may be singled out as
publishable measurements, which characterise the quality in a way that is helpful enough to publish. Other
measurements may be useful for other purposes, such as analysing trends.
Measurements may have targets. For instance:
The fault repair time might have the average time taken to repair reported faults as a publishable
measurement description, with 1.5 days as a target for it.
The call setup success ratio might have the percentage of calls successfully set up after receiving dial
tone as a publishable measurement description, with 97% as a target for it.
The packet transmission time might have the average time taken to transmit 56-byte packets between
representative sources and destinations as a publishable measurement description, with 40 milliseconds
as a target for it.
The packet loss ratio might have the percentage of 56-byte packets lost during transmission between
representative sources and destinations as a publishable measurement description, with 1.0% as a target
for it

The measurements made vary widely between regulators. However, they tend to reflect similar
concerns and therefore to have related, but not identical, measurement methods. Box 3 illustrates this,
by giving names to most of the measurements that are mentioned in this paper. Measurement names
are not sufficient for defining measurements in regulations: the call completion ratio, for example, can
refer to measurements defined in several different ways (Some of which, incidentally, have little to do
with quality of service from operators and much to do with how frequently people answer phones.)
However, measurement names are useful in identifying the general intentions of measurements.

14
Box 3 Conventions for measurement names
Measurement names should be regarded as short explanations of the measurements, not as full measurement
descriptions. In particular, the measurement names in this paper may be short for different measurement
descriptions in different examples and may differ from the names used by regulators. (For example, what
exactly constitutes a fault report varies between regulators.)
Many measurements fit one of the following patterns and, in this paper, have corresponding uniform names:
The measurements can be times taken to do actions. In this paper they have names of the form the x
time, which signifies the average time taken to do the action x, from initiating it (by making a request
or pressing a button, for example) to completing it. Thus the service supply time signifies the average
time to supply services, the fault repair time signifies the average time to repair faults, the complaint
resolution time signifies the average time to resolve complaints, the call setup time signifies the
average time to setup calls, and the packet transmission time signifies the average time to transmit
packets. (Times other than averages are also sometimes useful, as in the maximum time taken to repair
95% of faults , for example.)
The measurements can be proportions of occurrences of actions done in time. In this paper they have
names of the form the x timeliness, which signifies the result of dividing the number of occurrences of
the action x that are done on time by the total number of occurrences of the action x. Thus the service
supply timeliness signifies the proportion of occasions in which services are supplied on time, the fault
repair timeliness signifies the proportion of occasions in which faults are repaired on time, and the
complaint resolution timeliness signifies the proportion of occasions in which complaints are resolved on
time. (The term timeliness is specific to this paper, as there is no term in widespread use.)
The measurements can be frequencies of occurrences of events per customer. In this paper they have
names of the form the x rate, which signifies the result of dividing the number of occurrences of the
event x by the total number of customers. Thus the fault report rate signifies the number of fault
reports per customer and the complaint rate signifies the number of complaints per customer.
The measurements can be proportions of occurrences of events having a specific property. In this paper
they have names of the form the x y ratio, which signifies the result of dividing the number of
occurrences of the event x that have the property y by the total number of occurrences of the event
x. Thus the bill error ratio signifies the proportion of bills that are erroneous, the call setup success
ratio signifies the proportion of call setups that are successful, the call drop ratio signifies the
proportion of calls that are dropped, the packet loss ratio signifies the proportion of packets that are lost
and the bit error ratio signifies the proportion of bits that are erroneous. (The term ratio is often
replaced by the term rate, bur a rate often implies dependence on time, money or some other quantity,
such as the number of customers: a bit error rate and a bit error ratio, for example, could be different.)

15
2 The operational arrangements for quality of service monitoring
This section considers how quality of service monitoring by regulators operates, particularly with the
involvement of operators and customers. Operators can be involved in defining measurements, setting
targets, making measurements, auditing measurements and publishing measurements. Customers can
be involved, too, in defining measurements and making measurements. The section also deals with
techniques for encouraging and enforcing compliance with quality of service obligations.
2.1 Defining measurements
Choices of measurements depend on what customers in the country are most concerned about. Choices
of targets depend on what operators in the country can hope to achieve. (How measurements and
targets can be chosen is discussed further in Section 3.1 and Section 3.2.) The choices can have
unforeseen side effects, as they can focus management attention on some aspects of services to the
possible detriment of others. Getting the balance right needs an understanding of internal company
dynamics and potential industry performance that many regulators do not expect to achieve. The
knowledge and experience of operators can help with this.
Regulators are often unwilling to work closely with dominant operators. However, they may be more
willing to work broadly with all operators to define measurements. Also, they may welcome activities
by other monitors using magazines and websites that define and make measurements or conduct
customer satisfaction surveys. (These other monitors form a miscellaneous class, including
consumer associations, pressure groups, internet service reviewers and monitoring system vendors.) In
general:
Letting dominant operators continue to use measurement methods that they have used for some
time may be expedient if no other operators have to make measurements
14
.
Letting all operators define their own measurements usually leads to measurements that are not are
comparable between different operators.
Encouraging other monitors to define and make measurements assists with monitoring the aspects
of services that concern customers. Of course, other monitors might not exist or might not monitor
aspects of services mainly of concern to operators and policy makers. They might also make
measurements in ways that have low costs but do not represent customer experience adequately.
Involving an industry group of operators in the definition of measurements uses the skills of
operators
15
. The group may take a long time to agree and implement the definitions, partly because
individual operators wish to minimise their own disruption and expense and partly because it may
need to examine measurement procedures in detail. However, the results may be acceptable to
everyone, especially if the group includes customer representatives and external experts. The
regulator may still need to check independently that the measurements defined are comparable
between operators inside the group. (They are usually not comparable with measurements defined
by operators outside the group.)
Examples of regulators making extensive use of industry groups are described in Section 2.5.
2.2 Setting targets
When quality of service is monitored, targets may be set. In principle target setting could be the
responsibility of various organisations, just as with measurement definition in Section 2.1. However,
in practice, target setting is done by regulators if it is done at all. Yet there are potential applications of
target setting by other organisations. In particular, if operators set their own targets they could offer
varied choices of quality and price but be obliged to meet commitments: in consultation with
regulators they operators would devise, publicise and commit to targets and publish indications of
whether these targets were attained.

16
In brief:
Letting dominant operators set their own targets is unlikely to be effective if no other operators
have to make measurements, as there will be no ways of comparing quality
Letting all operators set their own targets could broaden choices of quality and price while
requiring operators to fulfil their commitments. However, this does not seem to happen, perhaps
because regulators that are more likely to let operators set targets are also less likely to require
targets.
Encouraging other monitors (such as pressure groups and monitoring systems vendors) to set
targets is also unusual, possibly because other monitors usually deal only with internet services, for
which expectations are less well developed than for telephony services.
Involving an industry group of operators in the setting of targets uses the skills of operators. It may
limit the choices available to customers, just as would setting targets without the involvement of
the group. Setting targets needs less detail than defining measurements, so an industry group is less
useful for it. In its case the knowledge and experience of operators can be taken into account by
due consultation and prior monitoring to determine realistic quality levels.
Figure 2 shows who defines measurements and sets targets in various countries. Different owners of
the definitions and settings may be suited to different weightings of the aims in Section 1.1: regulators
may need to own more if they monitor just dominant operators than if they monitor all operators.

Figure 2 Examples of responsibilities for defining measurements and setting targets
Note: The hatched area in Figure 2 indicates possible responsibilities, as no targets can be set if there are no
measurements defined.
targets
set by
operator
targets
set by
nobody
targets
set by
group of
operators
without regulator
targets
set by
group of
operators
with regulator
targets
set by
regulator
measurements
defined by
other monitor
measurements
defined by
nobody
measurements
defined by
group of
operators
without regulator
measurements
defined by
group of
operators
with regulator
measurements
defined by
regulator
targets
set by
other monitor
measurements
defined by
operator
Australia: fixed telephony
Canada: fixed telephony
Cte d'Ivoire: mobile telephony
Brazil: all telephony
Egypt: mobile telephony and all internet
India: all telephony and broadband internet
Malaysia: all telephony and all internet
Turkey: all telephony
Venezuela: all telephony
Australia: mobile telephony and broadband internet
Colombia: all telephony
Pakistan: mobile telephony and all internet
Singapore: all telephony and broadband internet
UK: fixed telephony
Australia: narrowband internet
Canada: mobile telephony and all internet
Colombia: all internet
Singapore: narrowband internet
UK: mobile telephony
UK: all internet
targets
set by
operator
targets
set by
nobody
targets
set by
group of
operators
without regulator
targets
set by
group of
operators
with regulator
targets
set by
regulator
measurements
defined by
other monitor
measurements
defined by
nobody
measurements
defined by
group of
operators
without regulator
measurements
defined by
group of
operators
with regulator
measurements
defined by
regulator
targets
set by
other monitor
measurements
defined by
operator
Australia: fixed telephony
Canada: fixed telephony
Cte d'Ivoire: mobile telephony
Brazil: all telephony
Egypt: mobile telephony and all internet
India: all telephony and broadband internet
Malaysia: all telephony and all internet
Turkey: all telephony
Venezuela: all telephony
Australia: mobile telephony and broadband internet
Colombia: all telephony
Pakistan: mobile telephony and all internet
Singapore: all telephony and broadband internet
UK: fixed telephony
Australia: narrowband internet
Canada: mobile telephony and all internet
Colombia: all internet
Singapore: narrowband internet
UK: mobile telephony
UK: all internet

17
Measurements often have targets associated with them, in the expectation that targets will help to
improve quality. However, targets (and the associated measures for ensuring compliance) are not
always effective in improving quality. Box 4 describes one such case involving the regulator in India,
the Telecom Regulatory Authority of India (TRAI). Targets were set after networks had been
monitored for 2 years. They were intended to lead to improvements in successive years. However, the
fixed networks could not be modernised fast enough to achieve the targets for fault report rates, which
had therefore to be deferred for 3 years. Moreover, the mobile networks achieved their targets in just
75% of situations. (Box 4 also illustrates how measurements may need to change as the market
develops, as discussed in Section 3.1.1, and how measurements must be comparable between
operators, as discussed in Section 3.1.3.)

Box 4 Monitoring by the regulator in India
Fixed telephony
TRAI published for every 3 months from April 2003 to J une 2005 17 measurements with targets for up to 75
fixed telephony operators in 27 geographic areas (treating different operating areas as different operators).
Participation in the scheme was compulsory. The targets, set in J uly 2001, were different for successive years.
TRAI revised the scheme after a review, for the following reasons
16
:
Quality had remained unsatisfactory (with measurements for fixed wireline networks lagging some years
behind the targets), though the targets had been set after 2 years of measurement and consultation.
Some measurements (such as the delay before getting dial tone) were still required but no longer important
and others (such as the delay before refunding deposits) were important but not yet required.
Fixed wireless networks were more like mobile networks than fixed wireline networks, which probably
had more access network faults and fewer dropped calls.
TRAI has published for every 3 months since J uly 2005 14 measurements with targets for 71 fixed telephony
operators in 27 geographic areas
17
. Participation in the scheme is compulsory
18
. The targets are generally the
same for successive years. The measurements include the service supply timeliness (separately for initial and
various changed services), the fault repair timeliness (separately for various periods of time), the fault repair
time, the fault report rate, the bill error ratio, the complaint resolution timeliness, the proportion of rapidly
answered attendant calls and the call completion ratio. Not all of the measurements are published: for fixed
wireless networks, the measurements published are like those published for fixed wireline networks but the
measurements required are like those required for mobile networks.
Mobile telephony
TRAI published for every 3 months from April 2003 to J une 2005 10 measurements with targets for up to 122
mobile telephony operators in 23 geographic areas (treating different operating areas as different operators).
Participation in the scheme was compulsory. The targets, set in J uly 2001, were different for successive years.
TRAI revised the scheme after a review, for the following reasons
16
.
Quality had remained unsatisfactory (with measurements failing to reach the targets in 25% of cases),
though the targets had been set after 2 years of measurement and consultation.
Some measurements (such as the fault repair time) were still required but no longer important and others
(such as proportions of calls blocked by wireless channel congestion) were important but not yet required.
Some measurements (such as the proportion of successful calls) were not comparable between operators,
as they depended on the operators in ways that were not described precisely enough.
TRAI has published for every 3 months since J uly 2005 14 measurements with targets for 128 mobile
telephony operators in 23 geographic areas
17
18
. The targets are
generally the same for successive years. The measurements include the bill correction timeliness, the bill error
ratio, the complaint resolution timeliness, the proportion of rapidly answered attendant calls (separately for
various periods of time), service coverage (separately for buildings, cars and streets), service availability, the
call setup success ratio, the call drop ratio and the conversational voice quality.

18
2.3 Making measurements
Regulators impose quality of service obligations for telephony services much more often than for
internet services. For internet services many rely on customer opinion surveys and on other monitors
using magazines and websites. These other monitors may even involve users in making measurements,
by distributing test tools and collecting test results. Occasionally these arrangements are made more
formal. For example, in Australia the regulator publishes an annual report about tests made by internet
users and in Chile the regulator required internet access operators to provide measurement tools to
their customers. The future may see much more widespread use of such arrangements: user terminals
besides computers are becoming ever more advanced and in principle could report back to operators
ample end-to-end quality of service measurements.
Box 5 describes the scheme of the regulator in Chile, the Subsecretara de Telecomunicaciones
(Subtel). In this scheme internet operators make some measurements but also distribute tools so that
their customers can make measurements. This scheme is unusual in various other ways, as it does not
apply to telephony services but it applies to both access and transit internet services and it obliges
operators to publish measurements on their own websites.

Box 5 Monitoring by the regulator in Chile
Internet access
Operators published from February 2002 to December 2004 2 measurements of narrowband (dialup) internet
access services
19
. The measurements were expressed as weekly, monthly, quarterly and annual trend graphs
on the websites of the individual operators. These websites also provided software for download so that
narrowband and broadband internet access users could make other measurements. Participation in the scheme
was compulsory
20
. The measurement methods, contents and formats and the software for download were
agreed by an industry group working with Subtel. The measurements made by the operators determined the
internet session login success ratio and the model pool utilisation. The measurements made by the users
determined the internet session login time and the data transfer rates for services including file transfer, web
browsing and email.
Subtel suspended the compulsory scheme pending a review, in order to update the telecommunications
legislation and introduce quality of service monitoring for new services, such as those using the internet. The
internet access operators are no longer publishing measurements.
Internet transit
Operators have published since J anuary 2001 3 measurements of internet transit services separately for each
connected internet access operator and each other internet transit operator
19
. The measurements are presented
as weekly, monthly, quarterly and annual trend graphs on the websites of the individual operators.
Participation in the scheme was compulsory
21
. The measurement methods, contents and formats were agreed
by an industry group working with Subtel. The measurements include the packet transmission time (delay),
the packet loss and the occupancy of the link with each connected internet access operator and each other
internet transit operator.
Subtel suspended the compulsory scheme pending a review, in order to update the telecommunications
legislation and introduce quality of service monitoring for new services, such as those using the internet. The
internet transit operators are nonetheless still publishing measurements.

Some measurements (such as fault repair times) are best made by operators. Others are best made by
external measurement agencies, because doing so makes them more comparable between different
operators and, sometimes, less costly. In particular, drive-around tests are more comparable and less
costly if they are done at similar times, between similar places and in similar circumstances for all
mobile telephony operators. Co-operation between the operators in defining and making such
measurements can be valuable in identifying occasional anomalies due to faulty measurement
equipment or incorrect applications of measurement procedures.

19
Operators and regulators often favour the use of customer opinion surveys. Surveys intended for use in
quality of service monitoring by regulators could be the responsibility of operators, other monitors or
regulators. Savings might be made by obliging operators to include questions due to regulators in their
own surveys. However, the surveys might then need to include many questions, because of differences
in the information wanted by the operators and the regulators. For example, the operators might want
to assess interest in possible new services and the regulators might want to ask about unsolicited sales
calls, public payphone faults, directory enquiry responses, and emergency call responses. To
regulators customer opinion surveys may be most valuable when customer support from operators is
inadequate (so customers do not bother to complain or report faults) and therefore when surveys by
operators are likely to be unsatisfactory.
In large countries having several geographic areas, many measurements are likely to be made. Thus in
India there can be at least 774 measurements about fixed operators and 1792 measurements about
mobile operators every 3 months. Of course the operators may have correspondingly many customers,
so the costs of the measurements per customer may be low. In fact the costs of measurements may
bear most heavily on small operators, because the number of test calls, for example, needed to make
precise enough measurements in a given geographic area is independent of the size of the operator.
Calculating and allocating the costs of measurements can be contentious
22
. Measurements can raise the
costs of all operators, including those that already make similar measurements. In principle the
measurements made could characterise the quality of many aspects of services for different market
segments in different geographic areas. However, the costs are related to the number and variety of the
measurements made and can be reduced by concentrating on few particularly useful measurements.
Reviewing the choice of measurements from time to time helps the measurements to remain those
important to customers and practical for operators, as discussed in Section 3.1.1 and Section 3.1.2.
Where measurements are made by an operator or are audited for an operator the costs are most
obviously carried by the individual operator. In other cases the costs may be recovered from licence
fees by the regulators or from all the operators in an industry group sharing an external measurement
agency, perhaps in proportion to subscriber numbers, coverage or revenue.
2.4 Auditing measurements
Measurements made by external agencies can be submitted to the regulator or even published before
the operators see them (as happens for drive-around tests in the UK). Measurements that are seen by
the operators before they are submitted to the regulator should be signed off by senior employees in a
self-certification process. They are therefore probably audited for accuracy by or for the operators.
Nonetheless regulators may want independent auditing of measurements. (For example, in Malaysia
the regulator audits all the dominant operators, and other selected operators, every year, even though
the law permits jail sentences for failures to fulfil quality of service obligations.) However,
independent auditing of all measurements can be inconvenient and costly and can delay publication
greatly. To increase confidence in the measurements but limit costs regulators might decide to audit
just some measurements (or data used in calculating some measurements), selected at random or
through being suspect. Regulators can also check measurements in other ways, such as comparing
them with the complaints received by the regulators or performing mystery shopping to test
complaint handling. All these ways need careful design, as audit procedures can easily become too lax
or too burdensome. The possibility of independent auditing can itself be quite effective in ensuring the
accuracy of measurements: quality of service measurements may not produce major newspaper
articles, but legal proceedings for falsifying them will do so.
The definitions of measurements themselves, as well as the subsequent measurements, require auditing
both before the definitions are implemented and thereafter: though the definitions may be intended to
be precise enough to achieve comparability, they may remain open to slightly different interpretations.
If the auditors for different operators and regulators meet together, they can find and resolve minor
problems of comparability, agree the main parts of each measurement procedure needing
investigations in audits, and validate the auditing techniques of each other.

20
Box 6 shows that the regulator in Ireland, the Commission for communications Regulation (ComReg),
and its predecessor, the Office of the Director of Telecommunications Regulation (ODTR), decided
that fairly simple quality of service monitoring was not helpful enough to continue. The consequences
of the audits (rejected measurements, delayed publications and extra costs) were factors in reaching
this decision. The decision contrasts with that in the UK described in Section 2.5, perhaps because the
influence of quality on customer choices was thought to differ in the two countries.

Box 6 Monitoring by the regulator in Ireland
Fixed telephony
ODTR and subsequently ComReg published for every 6 months from J anuary 2001 to December 2003 3
measurements for up to 9 fixed telephony operators of residential or business services (treated separately)
23
.
Participation in the scheme was voluntary (except for the universal service provider). Operators did not
always need to make all of the measurements. Mergers and withdrawals from participation left only 6
operators late in the life of the scheme. The measurements included the service supply timeliness, the fault
repair timeliness and the complaint resolution timeliness.
ComReg suspended the voluntary scheme after a review, for the following reasons:
Lack of comprehensive coverage reduced the usefulness of the scheme to customers.
Measurements were often rejected after audits for not being comparable with others.
Costs were disproportionate to the margins of small operators.
Choices of operators by residential customers (but not necessarily business customers) were based mainly
on pricing.
Comreg is developing new quality of service obligations that will apply to the universal service provider.
Mobile telephony
ODTR and subsequently ComReg have not imposed quality of service obligations on mobile telephony
operators.
2.5 Publishing measurements
Publishing measurements is central to achieving some of the aims described in Section 1.1. In
particular, it is essential if customers are to base choices of services on general conclusions, not just
personal anecdotes.
Publishing measurements usually involves distributing written words through websites. However,
written words may not be the most suitable for the general public, at least in rural and remote districts
or in oral societies where spoken words predominate. Publishing some measurements through radio
broadcasts or freephone calls could reach more people. Of course journalists and other intermediaries
might still prefer written words that they could summarise for the general public. Where written words
are suitable there are other ways of providing them besides using websites or even newspapers. In
particular, quality of service deficiencies could be publicised in SMS messages and bills.
In the countries discussed in this paper, publishing measurements depends on written words, often in
tables of numbers. However, tables of numbers are not always used exclusively: bar charts are used in
Singapore, trend graphs are used in Chile, and bar charts and coverage maps are used in the UK, for
example.
Table 3 gives a simple example of a table of numbers for a particular measurement period and
geographical area, displaying three measurements (the fault repair time, the packet transmission time
and the packet loss) for four internet services from three operators (X, Y and Z). The table is designed
according to some of the criteria listed in Section 3.3 so that it can be accessible to customers, helpful
to customers and fair to operators: the measurements are presented for different operators in the same
table, published without extra irrelevant numbers, rounded to at most two figures, expressed

21
consistently (in this case with higher numbers always meaning worse quality), separated for different
services from the same operator, and accompanied by explanations of unusual quality levels.

Table 3 Examples of publishable measurements arranged in a table
service name fault repair time
(target: 1.5 days)
packet
transmission time
(target: 40
milliseconds)
packet loss
(target: 1.0%)
explanatory remark
X Supernet 3.5 days 34 milliseconds 3.5% The quality was affected by
slow fault repair by the
backhaul operator.
Z Business Net 0.3 days 26 milliseconds 0.5% The measurements were made
when the service had very few
customers.
Z Residential Net 1.1 days 38 milliseconds 1.2% The measurements were made
for this area jointly with others,
not for this area separately.
Y Packetspeed 1.4 days 34 milliseconds 0.8%
Note: The coloured numbers in Table 3 exceed the targets. The numbers and the uses of targets for internet
services in Table 3 are not intended to represent best practice.

Figure 3 represents much of the same numerical information as a bar chart (with different scales for
the three different measurements). The explanations of unusual quality levels (and indeed the numbers
themselves) could be added to the bar charts as comments.

Figure 3 Examples of publishable measurements arranged in a bar chart
Note: The horizontal lines in Figure 3 represent the targets. The numbers and the uses of targets for internet
services in Figure 3 are not intended to represent best practice.

1 2 3
X Supernet
Z Business Net
Z Residential Net
Y Packetspeed
packet transmission time
(target: 40 milliseconds)
packet loss
(target: 1.0%)
comparative
measurements
fault repair time
(target: 1.5 days)
better
worse
1 2 3
X Supernet
Z Business Net
Z Residential Net
Y Packetspeed
packet loss
(target: 1.0%)
comparative
measurements
fault repair time
(target: 1.5 days)
better
worse
1 2 3
X Supernet
Z Business Net
Z Residential Net
Y Packetspeed
packet loss
(target: 1.0%)
comparative
measurements
fault repair time
(target: 1.5 days)
better
worse

22
Table 3 represents how measurements might be tabulated when they are intended to be compared by
customers (as opposed to journalists, for example). In many countries the tables have other
arrangements: in Brazil and Colombia, for example, the measurements for fixed telephony operators
are presented for different operators in different tables and do not provide explanatory remarks.
In some countries many measurements from many operators are published at many times. Most of
these measurements are unlikely to interest many potential customers. However, they may help to
demonstrate that the regulator is fair and open. Box 7 outlines the detailed monthly publishing done by
the regulator in Brazil, the Agncia nacional de telecomunicaes (Anatel). This can result in the
publication of at least 1190 measurements about fixed operators and 1034 measurements about mobile
operators every month. (In this example, as in the other examples in this paper, the counts of
measurements do not include numbers that are published but that are essentially irrelevant, in that they
are simply used in deriving the measurements.) In both India and Brazil the numbers of measurements
are extremely high partly because measurements are published for many separate geographic areas.

Box 7 Monitoring by the regulator in Brazil
Fixed telephony
Anatel published for every month from J uly 1998 to December 1999 14 measurements for 35 fixed telephony
operators in 34 geographic areas (treating different operating areas as different operators). Anatel then
replaced the 14 measurements by 34 measurements, as further detailed measurements were wanted.
Anatel has published for every month since J anuary 2000 34 measurements with targets for 35 fixed
telephony operators in 34 geographic areas
24
. The measurements are presented as tables separately for each
operator. Participation in the scheme is compulsory
25
. The measurements include the service supply timeliness
for address changes, the fault repair timeliness (separately for various periods of time), the fault report rate,
the fault repair timeliness for public payphones, the fault report rate for public payphones, the bill error ratio
(separately for local, national and international bills), the complaint resolution timeliness, the proportion of
rapidly answered attendant calls (separately for day and night), the call setup success ratio (separately for day
and night and for local, national and international calls) and the call completion ratio (separately for day and
night and for local, national and international calls).
Anatel has also published for every month since March 2005 the number of complaints per 1000 subscribers
for each of these operators.
Mobile telephony
Anatel published for every month from J anuary 2000 to February 2004 9 measurements with targets for 35
mobile telephony operators (treating different operating areas as different operators). Anatel then replaced the
9 measurements by 22 measurements, as the mobile telephony operators changed to different technologies and
operating areas.
Anatel has published for every month since September 2003 22 measurements with targets for 47 mobile
telephony operators
26
27
. The measurements include the fault repair
timeliness, the bill error ratio, the complaint rate (separately for general and coverage or congestion
complaints), the complaint resolution timeliness, the proportion of rapidly answered attendant calls (separately
for morning, evening and night), the call setup success ratio (separately for morning, evening and night), the
call drop ratio (separately for morning, evening and night) and the call completion ratio (separately for
morning, evening and night).
Anatel has also published for every month since March 2005 the number of complaints per 1000 subscribers
for each of these operators.

If only one, dominant, operator has quality of service obligations it may publish its own
measurements. However, if several operators are to be compared easily, their measurements are best
presented near each other

in the same publication. Measurements of multiple operators are published
on single websites for every 3 months in Canada and Singapore and for every 12 months in France, for
example, as well as in almost all the countries discussed in detail in this paper. (However, only some

23
of these websites can present measurements of multiple operators together on a single page, and most
do not cover all classes of operators.)
Collecting measurements from multiple operators introduces delays and administrative overheads.
Slightly less co-ordination is needed if measurements are published by individual operators separately,
as is done in Belgium and Spain for telephony services and in Chile for internet services. Doing this
may be satisfactory if comparisons are expected to be made only by enthusiasts and experts such as
specialist journalists, not by typical members of the general public. However, it still leaves regulators
with responsibilities for specifying uniform contents and formats for the measurements of different
operators, vetting the measurements and telling customers about the existence of the measurements.
Publishing many measurements from many operators at many times requires skills that several
regulators and operators do not have
28
. However, the task can be simplified in various ways. For
instance:
The numbers of measurements can be reduced. Customers are unlikely to be interested in many
measurements, especially when the measured differences between operators are small.
The numbers of operators can be reduced. Operators that have too few customers or too few
revenues (or that are not dominant, of course) could be exempted from quality of service
obligations.
The numbers of times of publication can be reduced. In a market that is developing quickly doing
this may not be helpful to customers or fair to operators, but in a market that is fairly mature many
actions to improve quality will not have noticeable effects for several months.
The numbers of geographic areas with separate measurements can be reduced. However, in many
countries geographic differences (including those between urban and rural districts) can be
important.
2.6 Ways of sharing responsibilities
Responsibilities for several of the activities outlined in Section 1.3 can be shared between the
regulator, the operators, other monitors and the customers. The factors involved in allocating
responsibilities to different stakeholders are discussed in Section 2.1, Section 2.2, Section 2.3, Section
2.4 and Section 2.5. Table 4 summarises the discussions by expressing the advantages as
disadvantages of alternative allocations of responsibilities. As with any such summary the statements
are simplifications and not substitutes for proper examinations of the factors.

24
Table 4 Disadvantages of allocations of responsibility in quality of service monitoring
responsible organisation
regulator
group of
operators
with
regulator
group of
operators
without
regulator
other
monitor
operator customer nobody
may be self-
serving.
does not
achieve any
quality aims.
defining
measurements
all operators
does not use
skills of
operators.
may be slow
to conclude.
may not look
at all
aspects.
may be slow
to conclude.
may not look
at all
aspects.
does not use
skills of
operators.
does not
provide fair
comparisons.
is not
feasible.
may lead to
poor quality.
may lead to
poor quality.
may lead to
poor quality.
setting
targets
all operators
limits
service
offerings.
may be slow
to conclude.
limits
service
offerings.
may be slow
to conclude.
limits
service
offerings.
may not look
at all
aspects.
does not get
commitment
by operators.

is not
feasible.

does not
achieve any
quality aims.
making
measurements
all operators
does not use
findings by
operators.
is not
realistic
always.
is not
realistic for
current
telephony.
needs audits.
is not
realistic for
fixed
telephony.
needs audits.
is not
realistic for
current
telephony.
may not test
enough.
does not get
economies
of scale.
needs audits.
is not
realistic for
current
telephony.
is not
represent-
ative.
may lead to
poor quality.
auditing
measurements
all operators
needs spot
checks by
regulator.
is not
feasible.
needs spot
checks by
regulator.
is not
feasible.
is not
prudent.
publishing
measurements
all operators
does not get
endorsement
by regulator.
does not get
endorsement
by regulator.
does not
provide easy
comparisons.
is not
feasible.
may lead to
poor quality.

As illustrated in Section 2.2 responsibilities are allocated differently in different countries. Table 5
provides further examples of allocations.

25
Table 5 Examples of allocations of responsibility in quality of service monitoring
responsible organisation
regulator
group of
operators
with
regulator
group of
operators
without
regulator
other
monitor
operator customer nobody
Canada:
fixed
telephony

defining
measurements
all operators
India: all
telephony
Spain: fixed
telephony
Chile
(suspended):
all internet
France:
mobile
telephony
UK: fixed
telephony
UK: mobile
telephony
UK: all
internet

Canada:
mobile
telephony
Canada:
fixed
telephony
Spain: fixed
telephony

setting
targets
all operators
India: all
telephony
France:
mobile
telephony

UK: all
telephony
Canada:
mobile
telephony
Canada:
fixed
telephony

making
measurements
all operators
France:
mobile
telephony
India:
mobile
telephony
UK: fixed
telephony
UK: mobile
telephony
UK: all
internet
India: all
telephony
Chile
(suspended):
all internet
Chile
(suspended):
all internet
Canada:
mobile
telephony
auditing
measurements
all operators
India: all
telephony
UK: fixed
telephony
UK: mobile
telephony

Canada:
mobile
telephony
Canada:
fixed
telephony

publishing
measurements
all operators
India: all
telephony
France:
mobile
telephony
UK: fixed
telephony
UK: mobile
telephony
UK: all
internet
Chile
(suspended):
all internet
Spain: fixed
telephony

Canada:
mobile
telephony

26
2.7 Ensuring compliance
Quality of service monitoring by regulators imposes on operators obligations to make measurements
and, sometimes, to attain targets. The obligations may be formalised in licences for separate operators
(as in Algeria, Morocco and Oman, for example) or in regulations for multiple operators (as in Cte
dIvoire, Malaysia and the United Arab Emirates, for example).
Quality of service obligations are only likely to be fulfilled if legitimate concerns are addressed. A
requirement for successful quality of service monitoring by regulators, therefore, is to take into
account the opinions of operators and consumers fully and openly. This can be achieved through a
consultation process, for obligations in licences as well as for obligations in regulations.
Typically at the start and end of a consultation process the regulator circulates documents that discuss
the policy options and proposals. The process then both brings extra knowledge and experience to bear
on complex problems and shares ownership of the solutions: in J ordan recently the consultation
process resulted in a greatly reduced burden for the regulator as well as the operators, by halving the
number of measurements and the frequency of auditing
29
.
Imposing obligations without due consultation about something as complex as quality of service is
likely to diminish the authority of regulators, not enhance it. Yet even after due consultation regulators
need techniques for ensuring that operators fulfil their obligations. These techniques may rely on
encouragement (and, in particular, on the power of publicity) or on enforcement (and, in particular, on
immediate effects on profits). In general:
The power of publicity is greatest when there is competition and the published measurements
distinguish between different operators: condemning operators collectively may achieve little.
Major differences in quality levels can be singled out for special attention while minor ones are left
for investigation by experts. Regulators in countries with many years of competition sometimes
tend towards using encouragement, arguing that different quality levels could emerge following
market forces and that detailed regulation needs too many people.
Immediate effects on profits attract attention from senior management in operators but may distort
measurement reporting within operators and to regulators. Usually enforcement entails comparing
measurements with targets (not just comparing measurements between operators), though the
targets might be set to match the prevailing quality levels. Regulators in countries with few years of
competition sometimes emphasise enforcement more than encouragement when they introduce
quality of service obligations, but they may find that it needs more effort than they can spare.
In some countries, encouragement and enforcement are applied more forcefully to dominant operators
than to other operators, especially where targets are set only for dominant operators.
Figure 4 highlights techniques for ensuring that operators fulfil quality of service obligations. Serious
and persistent failures to fulfil many obligations might be handled by using even more drastic
enforcement techniques that are not related directly to quality of service, such as withdrawing licences
or transferring franchises.

Figure 4 Techniques for ensuring fulfilment of quality of service obligations

encouragement enforcement
publishing all
measurements
and unattained
targets
changing prices publicising
deficiencies to
customers
monitoring the
implementation
of remedial
plans
demanding extra
measurements
and detailed
targets
imposing fines requiring
compensation
for customers
excluding access
to government
contracts
encouragement enforcement
publishing all
measurements
and unattained
targets
changing prices publicising
deficiencies to
customers
monitoring the
implementation
of remedial
plans
demanding extra
measurements
and detailed
targets
imposing fines requiring
compensation
for customers
excluding access
to government
contracts

27
These techniques are not always practicable (or available under the law, of course). In particular:
Publishing all measurements and unattained targets can be laborious. However, it can help to show
that the regulator is fair and open. Publishing at least some measurements is central to helping
customers make informed choices: as discussed in Section 2.5, it is often the main technique for
encouraging compliance with quality of service obligations.
Publicising deficiencies to customers (by putting remarks in bills, messages or advertisements)
needs firm comparisons with other operators or against targets.
Demanding extra measurements and detailed targets could lead to an emphasis on measurement
procedures instead of problem solutions. Nonetheless, it can be appropriate when the actions
needed to improve quality can be effective fast. Thus when quality fell well short of the targets at
points of interconnection in India, all operators were directed to provide interconnections within 90
days of being paid
30
. They were also required to make measurements of congestion at points of
interconnection every month, instead of every three months, in the expectation that reductions in
the times taken to provide interconnections would have observable effects rapidly
31
.
Monitoring the implementation of remedial plans may require external agencies skilled in network
design and operation to assist the regulators. Accompanying it by direct intervention in the
activities of the operators may cause confusion about responsibilities and duties.
Requiring compensation to customers may not be feasible when customers do not need accounts
(so credits and rebates cannot be given, as with certain calling cards) or bills are often wrong. It is
most useful when customers have better information than regulators about the quality that they
receive, can request compensation directly without recourse to regulators or arbitrators, and notice
different quality levels very easily (so the compensation can vary with the severity of the
deficiency). Compensation is therefore often paid mainly for long fault repair times in fixed
wireline networks, as in Chile and India (and also in the UK and some states of the US, for
example)
32
. (However, in Mexico it is paid more generally.) For similar reasons compensation is
suited to wholesale services, so in Canada it is paid to both wholesale and retail customers of
incumbent operators
33
.
Imposing fines can involve extensive legal processes and may take a long time (as may various
other techniques for enforcing compliance with quality of service obligations). For example, the
regulator in Brazil imposed fines on several operators in 2005, as well as in earlier years, for
quality problems that arose in 1999, 2000 and 2001 (though if fewer cases need to be investigated
the delay in imposing a fine can be just 1 year.),
34
.
Changing prices (by introducing quality factors into price controls, effectively with rewards for
good quality as well as penalties for bad quality) needs careful design if it is to act as an incentive
to improving quality
35
. As indicated in Section 2.8 the relation between quality of service and price
controls is not always clear.
Excluding access to government contracts can be difficult to make proportionate to failures by
operators and may not be applicable to several operators at once.
The techniques require effort and expertise, of various sorts, with processes that operate efficiently and
regularly whenever operators report measurements. Nonetheless they deserve consideration if the aims
in Section 1.1 are to be achieved. In many countries several techniques are available, to provide
graduated penalties. Thus in Turkey the penalties may involve monitoring the implementation of
remedial plans, requiring compensation to customers, or imposing fines.
Using several techniques instead of one can lead to less legalistic procedures and more proportionate
remedies. Box 8 outlines a scheme used by the regulator in Australia, the Australian Communications
and Media Authority (ACMA), and its predecessor, the Australian Communications Authority (ACA).
This scheme itself depends on several techniques (publishing measurements, monitoring the
implementation of remedial plans and requiring compensation to customers).

28
Box 8 Monitoring by the regulator in Australia
Fixed telephony
ACA and subsequently ACMA have published for every 3 months since J anuary 1998 4 measurements for up
to 8 fixed telephony operators in 7 geographic areas (subdivided further into urban, major rural, minor
rural and remote districts)
36
. The measurements are presented as tables separately for each operator.
Participation in the scheme is compulsory for fixed telephony operators having large enough customer
numbers
37
. The measurements include the service supply timeliness (separately for initial and various changed
services), the fault repair timeliness and the proportion of fulfilled appointments. Reviews have changed some
scheme details, most recently by removing the obligation on resellers to make and report measurements
38
.
The scheme was developed to get operators to provide and maintain telephony services for residential and
small business customers (having at most 5 connections) speedily. It specifies times in which operators should
supply services, repair faults, fulfil appointments and pay compensation. It thereby defines a customer service
guarantee from participating operators that is enforced using the following rules
37
:
An operator who exceeds the time on any occasion must pay compensation of AUD12 (USD9.5) to
residential customers and AUD20 (USD16) to business customers (roughly half the monthly rental in both
cases), for each of the first 5 days of lateness, and AUD40 (USD32) for each subsequent day of lateness.
An operator who exceeds the times on at least 10% of occasions in at least 2 successive quarters or who
shows significant performance decline is asked about possible causes and remedial plans.
An operator can be exempted if it exceeds the times for reasons beyond its control.
An operator must offer services with the customer service guarantee but may also offer services with
better or worse guarantees.
In 2005, the dominant operator exceeded 4% of service supply times, 9% of fault repair times and 6% of
appointment times, with average compensation payments of AUD58 (USD46), AUD25 (USD20) and AUD14
(USD11) respectively
39
.
Mobile telephony
ACA and subsequently ACMA have other quality of service monitoring schemes. Among them are ones
assessing the performance of operators offering mobile telephony, the reliability of the universal service
provider, and the responsiveness of operators giving priority to customers with life-threatening medical
conditions. Some of these give force to work by an industry group of operators, the Australian
Communications Industry Forum (ACIF), which has drawn up several codes of practice for telephony service
customer protection and network operation
40
. (ACIF is now part of a larger group, the Communications
Alliance.) Another industry group, the Internet Industry Association (IIA), has drawn up codes of practice for
internet service customer protection
41
.

Some operators will not fulfil quality of service obligations voluntarily, so encouragement needs to
have the force of law behind it. This problem of compliance occurs with other obligations (such as
adherence to codes of practice) and makes many operators prefer regulation, which shares burdens
fairly, to persuasion, which imposes burdens only on compliant operators.
Box 9 describes one case, that of the regulator in the UK, the Office of communications (Ofcom), and
its predecessor, the Office of telecommunications (Oftel), where encouragement to fulfil quality of
service obligations was insufficient. Schemes involving industry groups were used for both fixed and
mobile telephony. Originally operators were not obliged to participate. However, the voluntary
scheme for fixed telephony operators was found to be insufficient and was replaced by a compulsory
one, though the corresponding scheme for mobile telephony operators was retained with the proviso
that it was developed further.

29
Box 9 Monitoring by the regulator in the United Kingdom
Fixed telephony
An industry group published for every 3 months from J uly 1995 to December 2004 5 measurements and 4
customer satisfaction survey results for up to 13 fixed telephony operators of residential or business services
(treated separately). Participation in the scheme was voluntary (except for the universal service providers).
Operators did not always need to make all of the measurements. Mergers and withdrawals from participation
left only 5 operators late in the life of the scheme. Nonetheless in 2003 the publication website received an
average of 77000 visits per month, with a maximum of 140000 visits per month.
Ofcom replaced the voluntary scheme with a compulsory one after a review, for the following reasons
3
:
Lack of comprehensive coverage reduced the usefulness of the scheme to customers.
Falling participation in the group made measurement and audit procedures difficult to improve.
Costs could be decreased by fuller participation in the group and improvements in procedures (except
perhaps for customer satisfaction surveys).
Choices of operators were influenced heavily by quality.
Existing ways of comparing between operators were not regarded as satisfactory by residential customers.
A new industry group is publishing for every 6 months from J anuary 2006 5 measurements for 20 fixed
telephony operators of residential or business services (treated separately)
42
. The measurements are presented
as bar charts comparing operators. Participation in the scheme is compulsory for fixed telephony operators
having large enough revenues, call volumes and life times, even if they do not own networks
43
. The
measurements were defined by the group in a framework laid down by Ofcom. The measurements include the
service supply timeliness, the fault repair timeliness, the fault report rate, the complaint resolution timeliness
and the bill error ratio.
Oftel also published for every 6 months from October 1999 to September 2003 the number of complaints per
1000 subscribers for each of these operators.
Mobile telephony
Oftel published for every 6 months from April 1999 to May 2003 3 measurements for 4 mobile telephony
operators in 9 geographic areas (covering 70 towns as well as main roads). Participation in the scheme was
voluntary. Operators sometimes blocked publication and took a long time to develop improved measurement
and audit procedures. Very few people knew about the publication website, which in 2003 received an
average of 4000 visits per month.
Ofcom replaced the voluntary scheme with another voluntary one (after a review, for the following reasons
3
:
The group was expecting to consider improvements in measurement and audit procedures.
Comparability would be increased and costs would be decreased by co-ordinating tests already done
separately by operators in the group.
Choices of operators were influenced heavily by quality.
Another new industry group is publishing from September 2006 3 measurements for 4 mobile telephony
operators. The measurements er updated twice monthly and are intended to cover the country in 6 months
44
.
The measurements are presented as tables and coverage maps comparing operators. Participation in the
scheme is voluntary, but there may be a further review if the information published does not help customers or
does not cover almost all service provision
43
. The measurements were defined by the group. The
measurements include the call setup success ratio, the call drop ratio and the conversational voice quality.
Oftel also published for every 6 months from October 1999 to September 2003 the number of complaints per
1000 subscribers for each of these operators.

30
2.8 Relating quality to profitability
As indicated in Section 2.7 regulators may try to enforce quality of service obligations by changing the
prices charged by operators. In particular they may introduce quality factors into price caps, as has
been done in some states of the US, for example. Box 10 describes what is done by the regulator in
Colombia, the Comisin de Regulacin de Telecomunicaciones (CRT).

Box 10 Monitoring by the regulator in Colombia
Fixed telephony
A government agency has published for every 3 months since J anuary 2003 9 measurements with targets for
up to 42 fixed telephony operators in up to 31 geographic areas (treating different operating areas as different
operators)
45
. The measurements are presented as tables separately for each operator. Participation in the
scheme is compulsory
46
. The measurements include the service supply time (separately for initial and changed
services), the fault repair time (separately for internal and external plant) and the fault report rate (separately
for internal and external plant). The regulations prescribe details of a customer satisfaction survey as well as
of these measurements
47
.
The price cap rules allow variations in local tariffs between different operators and geographic areas and are
not imposed on some operators in some geographic areas. The variability extends to the treatment of quality
through the quality factor, which is different for each operator that is subject to the price cap rules. Each year,
for each such operator the quality-dependent price cap is the product of the quality factor and the quality-
independent price cap (which is formed by subtracting a productivity factor from the retail price index).
For the start of the scheme in 2000 the quality factor was set to 100% or 84% (depending on when the
operators reported their measurements to CRT). Since then it has been a weighted sum of figures derived from
measurements of the customer satisfaction level, the service supply time, the fault repair time and the fault
report rate. The figures are derived by raising to a certain power the results of mapping the measurements on
to truncated linear scales between 100% (for more satisfactory quality) and 50% (for less satisfactory quality).
The certain power changed in 2005 from 0.25 to 0.20 (so the minimum possible quality factor changed from
84% to 87%)
46
. In addition the weightings, more satisfactory values and less satisfactory values change
from year to year, to reflect changes in what is realistic for the networks. For instance
48
:
The customer satisfaction level had a weighting of 25% in 2001 and 40% in 2006, a more satisfactory
value of 68.1 in 2001 and 80.0 in 2006, and a less satisfactory value of 53.7 in 2001 and 76.0 in 2006.
The service supply time had a weighting of 25% in 2001 and 10% in 2006, a more satisfactory value of
100.8 days in 2001 and 10 days in 2006 and a less satisfactory value of 410.6 days in 2001 and 15 days
in 2006.
The fault repair time had a weighting of 25% in 2001 and 20% in 2006, a more satisfactory value of 2.5
days in 2001 and 1 day in 2006 and a less satisfactory value of 5.6 days in 2001 and 2 days in 2006.
The fault report rate had a weighting of 25% in 2001 and 30% in 2006, a more satisfactory value of
39.2% in 2001 and 23% in 2006 and a less satisfactory value of 83.6% in 2001 and 33% in 2006.
The calculated quality factor is often more than 90% but is sometimes the minimum possible (84% or 87%).
Mobile telephony
Another government agency has published for every 3 months since J anuary 2003 3 measurements for 6
mobile telephony operators
49
. The measurements include the proportion of successful customer appeals
against decisions of operators.

Even when regulators do not introduce quality factors into price caps, they wish to ensure that
regulating the profitability of operators makes quality levels more appropriate (or at least does not
make them less appropriate). Box 11 summarises the evidence from the US about the relation between
quality levels and ways of regulating the profitability of dominant fixed telephony operators. At this
stage it offers no final clear conclusions.

31
Box 11 Relations between quality of service and regulation of profitability in the United States
The states of the US have been regulating incumbent fixed telephony operators for many years. They do so in
a framework determined by the national regulator, the Federal Communications Commission (FCC), that
allows many variations between states but that largely determines the measurements made for quality of
service. FCC published the measurements for every 3 months from J anuary 1991 to December 1995. FCC has
also published the measurements for every 12 months since J anuary 1996
50
. The measurements include the
service supply time, the fault repair time, the fault report rate, the complaint rate and service availability.
Each operator serves several states. In principle the measurements can therefore be analysed to check whether
differences in quality levels are correlated with differences in how states regulate operators.
The states have typically shifted from rate of return regulation to various forms of incentive regulation.
Rate of return regulation limits the profits kept by operators by assessing them in such a way that operators
recover their costs. Incentive regulation encourages operators to reduce costs by letting them keep some of the
profits due to cost reductions. The most widespread kind of incentive regulation involves price caps to limit
the prices charged to customers, with the price cap being defined by subtracting a productivity factor from a
retail or wholesale price index. Other kinds of incentive regulation involve sharing profits with customers and
deferring reductions in prices to rate of return levels.
Rate of return regulation does not encourage cost reductions greatly. Indeed it might be expected to improve
quality above optimum levels, because operators would suffer no penalties for capital investment that would
make life easier for their engineers. Incentive regulation, by contrast, might be expected to lower quality
levels because operators would be constrained in how they maintained margins. Some states in the US have
introduced a quality factor alongside the productivity factor in price caps to overcome this potential
problem in incentive regulation: the profits kept by operators would be increased or decreased according to
whether quality levels had risen above or fallen below targets.
Several studies of retail services in the US have used the FCC publications to consider whether differences in
quality levels are correlated with choices between rate of return regulation and incentive regulation
51
.
However, at this stage the evidence is inconclusive and might even seem contradictory on initial inspection:
different studies use different ways of defining incentive regulation, counting quality deficiencies and
examining other influences on quality levels, so their results are not comparable
52
.
Fewer studies have examined wholesale services. Their quality levels do seem to have fallen while there have
been price caps
53
. However the fall does not seem to be correlated with the use of incentive regulation instead
of rate of return regulation
54
.
2.9 Resolving quality problems directly
Quality of service obligations make the operators take the responsibility for solving problems. This
may not be appropriate to all problems in all circumstances: in countries where incumbent operators
have few funds and alternative operators cover limited geographic areas, investments in quality may
need to be centralised. An approach that aims to eradicate the root causes of the problems may achieve
more than encouragement or enforcement.
In particular, if the problems can be traced to deficiencies in staff skills, there may be requirements for
national training programmes available to the staff of all operators. Poor customer support is
sometimes a symptom of such deficiencies.
High fault report rates and low proportions of successful calls generally indicate a need for improved
network equipment. In fact network modernisation can improve quality of service greatly
55
. As
network modernisation is motivated largely by wishing to reduce running costs, regulators can
stimulate it by relating the profits kept by operators to prices instead of costs
56
.
Quality of service improvements that depend on staff development may fall away without sustained
effort, but those that result from capital investment are likely to last until the next major change in
technology. Suitable staff recruitment and process simplification can sometimes produce
improvements quickly, but many actions to improve quality are slow to have their full effects.

32
3 The selection of areas of focus for quality of service monitoring
This section examines the likely constraints on quality of service monitoring by regulators. It describes
criteria for choosing the measurements to be defined, the targets to be set, the measurements to be
published and the measurements to be made. The section then discusses which services are to be
monitored in current network and NGN environments.
3.1 Measurements to be defined
The numbers of measurements required by regulators vary widely between countries: the examples in
Section 2.4 and Section 2.5 show variations by factors of 10. There are many measurement definitions
to choose between: ITU has currently at least 150 recommendations related to quality of service, and
several other organisations, such as ETSI and IETF, have also developed related standards
57
.
Regulators can narrow their choices to measurements that are relevant to one or more of the aims in
Section 1.1. They can then restrict attention further to measurements that as far as possible are:
Important to customers.
Practical for operators.
Comparable between operators.
These points are discussed in some detail below.
3.1.1 Importance to customers
The measurements defined for quality of service monitoring by the regulator need to be important to
customers (though operators will often choose to have extra measurements for their own purpose).
This paper suggests, in particular, that:
The measurements should be the ones needed by those current or potential customers who would
be most seriously affected by unsatisfactory quality. (In particular, measurements that are published
should be helpful to customers in the ways listed in Section 3.3.1.) In some countries the customers
most seriously affected might be people with disabilities. In other countries they might be the users
of the most popular service. (For example, in the Philippines SMS messages, often from poor users,
contribute as much revenue as mobile voice calls, so at some stage SMS transmissions might
require quality of service monitoring even if voice calls did not.)
The measurements should be made over short enough periods of the year and for small enough
geographic areas to point to problems
58
. Average measurements are not always adequate: for small
operators quality may fluctuate greatly over some months and for large operators quality may vary
between districts. (For example, unsatisfactory pair gain systems in some districts of Australia
prevent broadband internet delivery, have poor voice quality, and get congested, but there are too
few of them have much effect on aggregate quality of service measurements.)
The measurements should be reviewed, to see whether they need to be changed, as the market
changes and different aspects of services become most important. In a developing country, the
number of years taken to supply fixed network connections may be important. As a network
expands, the proportion of successful calls may need attention. Once a network matures, customer
support attracts most complaints. The details of individual measurement definitions may also
evolve as quality improves: service supply times for initial network connections may come to be
measured in days rather then in years and eventually in terms of adherence to times specifically
requested by customers. (In fact developments in technology doe not always lead to improvements
in quality: conversions to and from voice over IP, together with mobile access, could make end-to-
end delay unsatisfactory when it was not unsatisfactory before.)
The measurements should be consistent between services, particularly if the services are seen as
substitutes for one another. Several measurements are likely to deal with aspects such as customer
support that are common to many services, as discussed in Section 3.5.3.

33
Box 12 outlines the experience of the European Union (EU) and the Organisation for Economic Co-
operation and Development (OECD), in which various measurements have gradually lost their
importance.

Box 12 The evolution of quality of service monitoring in the EU and OECD countries
The European Commission set up a study in 1989 of possible quality of service measurements
59
. Some
measurements (such as the delay before getting dial tone) were dismissed as being unnecessary in digital
networks, but others seemed relevant according to the surveys reported in the study. The European
Commission chose 9 measurements for incorporation in law in 1998
60
. These measurements underlie quality
of service monitoring for universal service providers (and sometimes other fixed telephony operators) in
countries that have joined or are joining the EU. In the new regulatory framework of 2002 the European
Commission made two of them (the proportion of unsuccessful calls and the call setup time) optional as their
values were good enough
61
.
The OECD chose 6 measurements like those of the European Commission in 1990
62
. These measurements are
discussed regularly in the OECD biannual reports on telecommunications. They have evolved with the
development of the market. In particular, for many OECD countries now the service supply time for initial
network connections is too small to be reported and the response times for operator services are less important
than the prices of the services
63
.

Ideally the aspects that concern customers are found, and periodically reviewed, through customer
opinion surveys. Such surveys can be expensive: one regulator rejected a pilot survey covering only
two geographic areas because of the expense. However, they may reduce the costs of operators by
showing where operators are paying too much attention to aspects of services that customers are not
concerned about
64
.
In the absence of surveys, records of complaints can help to guide the choice of measurements.
However, often these records are open to interpretation in various ways, as described in Section 3.4.1,
and do not represent the concerns of potential customers fully: people give up trying to complain.
(Supply times for connections are particularly hard to complain about when there are no connections.)
Often many complaints are about prepaid or postpaid billing. Figure 5 illustrates this by summarising
analyses of complaints made to two regulators in Africa in 2002-2003 and in 2004-2005
65
. (The
analyses resemble each other in several respects but not in all respects: the proportions of complaints
allocated to the classes billing, pricing, administration and contract breach are rather similar
for the two countries, but the proportions of complaints allocated to the classes disconnection,
service supply and misrepresentation are very different for the two countries.)

34
21%
20%
16%
11%
7%
1%
14%
2%
4%
4%
billing
disconnection
pricing
fault
service supply
misrepresentation
administration
contract breach
call failure
other
24%
15%
3%
12%
1%
10%
1%
9%
20%
5%

Figure 5 Analyses of complaints to regulators in two countries in Africa
3.1.2 Practicality for operators
The measurements defined for quality of service monitoring by the regulator need to be implementable
by operators for reasonable costs in reasonable times using consistent measurement and audit
procedures. This paper suggests, in particular, that:
The measurements should not require more tests and calculations than are likely to be needed to
characterise differences in quality that are perceptible to customers, now or in the future. Most
services may need few measurements each and several of those measurements may be common to
many services (as in Malaysia).
The measurements should be the same as or similar to ones that operators already make (or will
benefit from making) for their own purposes, if possible.
The measurements should deal with matters that operators can control: operators make choices
(about numbers of base stations, for example) that weigh up the costs and benefits of reducing
problems and that may affect levels of quality and pricing.
The measurements that need to be made for separate periods of the year and for separate
geographic areas should be aligned with those already used by the operators for customer support
and network operation, if possible.
Operators can benefit from sharing in quality of service monitoring. They may find that it helps them
to work out where to direct their investment: high fault repair times may point to improving customer
support, whilst high fault report rates may point to improving network equipment
66
.
The probable causes of particular deficiencies might be identified and quantified as far as possible in
treatments of complaints and fault reports. They could be useful to operators and regulators but not to
customers. They would be omitted from documents provided to customers or other operators, for
reasons of confidentiality and comprehensibility. (For example, to operators disconnections caused by
non-payment of bills, equipment failure, and equipment theft are all different but to customers, one
disconnection looks much like another.)
3.1.3 Comparability between operators
For comparability to be achieved, measurements must allow comparisons between operators offering
similar services. (Similar services can substitute for each other to some extent, so services are certainly
similar if they differ only in implementation features that customers do not experience directly.) This
paper adopts this view in proposing that:
The measurement descriptions (explained in Box 2) should expose any significant dependencies on
choices by individual operators. This has particular importance for measurements that are counts of

35
events occurring in certain intervals of time: the intervals might be defined once for all operators or
separately by individual operators. (For example, the proportion of faults repaired within 2 days
might be described as characterising the ability of the operator to repair faults, and the proportion
of faults repaired within a time identified by the operator might be described as characterising the
ability of the operator to fulfil fault repair commitments but should not be described as
characterising the ability of the operator to repair faults.) Also, if there are fewer significant
dependencies on individual operators the measurements will generally be more helpful to
customers.
The measurement methods (explained in Box 2) should be precise enough that differences in
interpretation and implementation should not lead to systematic (non-random) differences in
measurements. (As measurement methods are statistical their results are subject to some
randomness.) This may be hardest to achieve for measurements of complaint handling, but it can
even be difficult for measurements of traffic handling. (For example, measurements of call set up
time may need to take account of whether calls end on user terminals or on network devices such as
central mail boxes, when calls are forwarded, and how number portability and number translation
are implemented, as well as the terminal types and application implementations for user terminals.)
The details of measurement methods may need to be discussed in groups of operators before they can
be settled. Table 6 provides an example of what might need to be decided even for counting fault
reports, by listing some apparent impairments that when reported by a user might or might not
generate new fault reports. Some of these, such as the repeated reports of events, could affect the fault
report count greatly. Others might be given different treatments by operators without affecting the
count significantly
67
.
Table 6 Examples of apparent impairments possibly not generating fault reports
reported by that user already
reported by a different user already
on more than one connection using the same route
on more than one connection not using the same route
in customer premises equipment maintained by the operator
in customer premises equipment not maintained by the operator
in additional application software supplied by the operator
in additional application software not supplied by the operator
due to normal operation of the service
cured by the user before testing
not detected during testing
not present after testing
due to maintenance at times notified to customers in advance
due to maintenance at times not notified to customers in advance
due to problems of another operator notified to customers in advance
due to problems of another operator not notified to customers in advance
due to disasters caused by human activity
due to disasters not caused by human activity

36

After the measurement descriptions and measurement methods have been refined, comparability may
be achieved. However, it may be comparability between operators in one group in one country only.
Often even the measurements for fixed telephony and for mobile telephony are not comparable
(though this may change as fixed-mobile convergence occurs). Not only the number of measurements,
but also the precision of measurement definitions varies within and between countries
68
. Where
measurements are the same in outline they may be different in detail, so comparisons can be very
approximate.
3.2 Targets to be set
As discussed below, if targets are set they should as far as possible be:
Useful to customers.
Realistic for operators.
3.2.1 Usefulness to customers
Targets can be put to various uses: as indicated in Section 1.1, they can be useful in maintaining
quality or in improving quality. Different kinds of target are needed for quality maintenance and for
quality improvement. More fully:
Some targets represent required minimum quality levels
69
. (For example, the proportion of
successful calls might be required to be at least 95%.) They are intended for maintaining quality
above levels that have already been achieved in the country. To set them regulators have to know
what quality levels have been reached.
Other targets represent desired or planned quality levels, typically set for successive years. (For
example, the proportion of successful calls might be desired or planned to be at least 96% after 1
year, 97% after 2 years and 98% after 3 years.) They are intended for improving quality from levels
that have already been achieved in the country. To set them regulators have to know what quality
levels could be reached, in what times and at what costs.
The choice between these kinds of target depends on the main aims of quality of service monitoring.
Required minimum service quality levels are more widely chosen, and more generally applicable, than
desired or planned quality levels.
Targets of either kind may be more appropriate to wholesale services than to retail services and to
dominant operators than to other operators, for the following reasons:
Policy makers are likely to be concerned with end-to-end quality levels determined by access and
core networks. The incumbent operator is likely to be dominant in the markets for making and
receiving calls using fixed access and core facilities
70
.
Customers are sometimes more interested in differences between quality levels than in absolute
quality levels, at least for retail services
71
. In such cases comparisons between operators may be
more useful than comparisons with targets. Of course customers may still welcome constraints on
the quality levels of a dominant operator.
Operators are likely to be interested in the quality levels that they can expect to be available from
those operators whose facilities they use for interconnection or resale. Often the main facilities
available are those of a dominant operator.

37
3.2.2 Realism for operators
Though targets can be useful they must be realistic if they are to be effective. Making them realistic
can be difficult, for the following reasons:
The targets should relate to the quality levels that customers want. However, these levels might
change as quality improves and might vary for different market segments in different geographic
areas.
The targets should be set after observing what is possible for the operators, so the regulator
appreciates what quality levels have been reached and what quality levels could be reached.
However, making the observations delays the introduction of the targets
The targets might be set by adapting values determined through earlier measurements by the
operators, used in other countries or proposed in international standards. However, these values
may be for measurements that are not comparable with the measurements chosen. In particular,
targets often differ greatly between countries. For example, in India and the Philippines the target is
to have no more than 1 error per 1000 bills and in Malaysia and Romania the target is to have no
more than 20
72
.
The targets should avoid limiting the available choices of quality and price, provided that operators
and regulators publish enough information for customers about the choices. For internet services,
for example, differences in quality due to differences in actual bandwidth may justify different
prices for connections having the same nominal bandwidth.
There are many international standards relating to quality of service measurements and targets: several
organisations have defined measurements and some have set targets. Box 13 mentions briefly those
that are most relevant to this paper. They are ETSI, which often looks at topics from a broader
perspective than that of the EU, and ITU, which carries out standardisation work through ITU-T and
ITU-R study groups. The study groups in ITU covering quality of service are ITU-T Study Group 2
(Operational aspects of service provision, networks and performance), ITU-T Study Group 12
(Performance and quality of service) and ITU-T Study Group 19 (Mobile telecommunications
networks)
73
.

38
Box 13 Some sources of measurements and targets from ETSI and ITU standards
ETSI has been responsible for standardising the quality of service measurements required by the European
Commission. It has therefore had definitions of measurements for fixed and mobile telephony services for
many years and definitions of measurements for internet services for rather fewer years
74
. The proposals for
measurements of internet services follow user surveys to ascertain the aspects of services that customers were
most concerned about
75
. These definitions leave open many choices. Though many of them are suitable as
inputs to national definitions, few of them can provide comparability without further work.
ETSI has also been devising quality of service measurements for mobile services with 3GPP. Besides voice
telephony and messaging the services include video telephony, video streaming, file transfer, web browsing
and email
76
. Measurements are defined both including and excluding application interfaces (thereby taking
into account more or less of the user experience). The measurement methods are accompanied by information
on test procedures, profiles and calculations leading towards comparability. Targets are not considered,
though some experimental investigations of targets have been done
77
.
Many ITU-T recommendations consider quality of service. They are useful in several areas (though they do
not provide consistent sets of measurement methods covering all aspects of services). Those most
immediately relevant to fixed and mobile telephony are E.721 and E.771, which include targets based on
typical reference connections for various calls (local, national, international, fixed and mobile)
78
.
Some targets in ITU-T recommendations relate to network performance, rather than quality of service, as they
are calculated bottom up, from network equipment, not top down, from user experience. However, there
are exceptions, such as the upper bounds on end-to-end delays that prevent conversations from deteriorating
and that are recommended in G.114
79
. Also targets for voice, video and data applications suggested by some
empirical work are summarised in G.1010
80
.
Some such voice, video and data applications are put in the context of IP network performance by Y.1541
81
.
However, several targets for IP networks are still rather tentative, as some applications have not been assessed
widely by users and other applications may never actually be implemented on IP networks. (In particular
some of the packet loss targets in Y.1541, but not those in G.1010, are unnecessarily demanding for most
applications that are expected to use IP networks: they are intended to deal with the transport of tightly
synchronized bit streams over IP and do not take account of the packet loss concealment that is implemented
with many voice encodings or the reliable delivery that is used in file transfers and similar applications.)

3.3 Measurements to be published
As discussed below, if measurements are published, they should as far as possible be:
Accessible to customers.
Helpful to customers.
Fair to operators.
3.3.1 Accessibility to customers
Measurements intended for publication should be published at locations or times that let their intended
users access them. This paper suggests, in particular, that:
The measurements should be published in a form directed either to customers or to others (such as
journalists) active for customers. These are the intended users.
The measurements should be presented appropriately to reach customers. In some countries
publishing through radio broadcasts or freephone calls would reach many more people than
publishing through newspapers or websites. In other countries readers would expect written words
and prefer tables of numbers to graphs or bar charts (contrary to widespread belief).
The measurements should have their locations or times of publication widely publicised and easily
found.

39
The measurements should be published at the same locations or times for all operators that are
intended to be compared.
3.3.2 Helpfulness to customers
Measurements intended for publication should be expressed in ways that their intended users find
helpful. (This may seem obvious, but it is overlooked surprisingly often.) This paper suggests, in
particular, that:
The measurements should deal with the aspects of services that customers are most concerned
about, as discussed in Section 3.1.1. In particular, as explained in Section 3.4.3, measurements of
traffic handling should be end-to-end, not network-by-network, if possible.
The measurements should relate to the aspects of services that customers can experience directly.
(For example, customers experience dropped calls directly but do not experience wireless network
traffic channel congestion directly, though it is one cause of dropped calls.)
The measurements should be published without being accompanied by extra, irrelevant, numbers.
(For example, the call setup success ratio may be of interest to users, but the number of test calls is
not of interest to users: auditors or others should have checked whether the number of test calls is
large enough to make the measurements precise enough.)
The measurements should be described in terms that are comprehensible to customers. This should
be possible if customers can experience the implications of the measurements directly, though
keeping the descriptions precise enough not to mislead may be difficult.
The measurements should assist with comparisons between operators. For this purpose they are
better presented in tables with entries for all the operators in one measurement period, not, for
example, in tables with entries for one operator over all the measurement periods.
The measurements should be expressed no more and no less precisely than is needed to
characterise differences in quality that are perceptible to customers and to suppress differences in
measurements that are likely to be random. Rounding the measurements to at most 2 significant
figures usually helps to achieve this
82
. It also corresponds very approximately with some
experimental results about human perception (though differences in perception vary with
circumstances). An alternative replaces measurements by a scale having perhaps five points (as in
France), where, to minimise subjective ranking, each point must represent a range of possible
numbers
83
.
The measurements should be presented consistently with each other. (For example, whether calls
can be made and whether messages can be sent should be presented in terms of either successful
calls and successful messages or unsuccessful calls and unsuccessful messages, not in terms of
successful calls and unsuccessful messages.)
Measurements can be weighted and summed to give single figures of merit. As indicated in Section
2.8 this happens in Colombia. Weightings need to result from conducting customer opinion surveys if
important information is not to be lost by using them
84
.
3.3.3 Fairness to operators
Measurements should be expressed in ways that are fair to operators. For instance:
The measurements for two services from the same operator should be published separately when
the services are intended to have different levels of quality (for business and residential customers,
say).
The measurements in two geographic areas might be published separately for all operators if
services in the two geographic areas are likely to have different levels of quality. However,
separate publication in this case might not always be realistic given the networks and organisations
of the operators.

40
The measurements might be annotated with explanations of deficiencies by an operator when the
operator needs to use the facilities of another operator in a way affecting quality. (For example, an
alternative operator may have faults on lines that are leased from a dominant operator that does not
offer fast protection switching.) Customers can then know both how satisfactory the service is and
why the service may have problems.
The measurements for a service might be published with annotations by the regulator when they
may not reflect a true or lasting state of affairs. This might happen if the service is too new to have
enough customers to provide statistically significant (precise enough) measurements.
3.4 Measurements to be made
When measurements are being defined, the measurement methods must be specified carefully if
comparability is to be achieved. Many fine details need to be clarified, as illustrated in Section 3.1.3.
Also, choices need to be made in each of the following broad classes of method:
Objective and subjective methods.
Passive and active methods.
Network-by-network and end-to-end methods.
3.4.1 Objective and subjective methods
Measurement methods may be objective (when they involve counting or timing events) or subjective
(when they involve asking for the opinions of customers). Measurement methods that involve counting
records of complaints are somewhere in between: the counting is objective, but there may be
subjectivity in decisions to complain or in interpreting complaints. In particular, measurements of
complaint handling are related to how operators perform customer support. For instance:
Whether calls from customers ever reach call centre agents, and are ever counted, can depend on
how many agents operators employ.
Whether calls from customers are classified as complaints or as queries can depend on how
articulate and assertive customers are.
Whether complaints about faults can be checked immediately can depend on which parts of
networks operators control.
Though quality of service is concerned with user satisfaction, objective measurement methods are
often preferable to subjective ones, for the following reasons:
Subjective measurements may place too much weight on the opinions of people who express
themselves strongly.
Subjective measurements result in distributions of opinions that can be difficult to summarise
helpfully and fairly. (For example, a mean opinion score has little value if users are split equally
between those who are very satisfied and those who are very dissatisfied.)
Subjective measurements make comparability within cultures difficult to achieve. (For example,
surveys by different operators would need to use the same questions, be made at the same times, be
focussed on the same populations and use the same mail or other techniques for getting responses.)
Subjective measurements make comparability between cultures difficult to achieve. The notion of
satisfaction may be quite different for different ethnic and religious groups.
Subjective measurements are complicated and costly.
As far as possible objective measurements methods are correlated with, or at least intuitively related
to, user perceptions.

41
Over time there may be some movement from objective measurements of mechanical service (such as
the proportion of calls that are successful) to subjective measurements of human service (such as the
proportion of customers that are satisfied with customer support). Measurements made by objective
measurement methods in different countries are rarely comparable in detail even if they are in outline,
but they might be made comparable by refining the measurement descriptions and measurement
methods. Measurements made by subjective measurement methods, might never be comparable
between countries, because of cultural differences.
Services have satisfactory quality only if users can communicate accurately enough. What accurately
enough means may vary between users and between uses. It is an obvious candidate for subjective
measurements but objective measurements are often favoured. Box 14 discusses this briefly, especially
for voice.

Box 14 Measurements for voice, video and text
Voice quality can be assessed by various methods described in ITU-T recommendations. In particular:
Subjective assessments use surveys of customer opinions about particular calls to calculate a Mean
Opinion Score (MOS), described in P.800
85
. In practice the opinions must be obtained by asking about test
calls, as otherwise they are difficult to distinguish from general feelings about the services. Even when test
calls are used the results can be affected strongly by the attitudes and environments of the survey
participants and may not be reproduced in other similar surveys.
Objective assessments for listening use estimators such as Perceptual Evaluation of Speech Quality
(PESQ), described in P.862
86
. The results can predict the quality of single messages (as in voice mail) but
not the quality of interactive conversations: they estimate distortion but not some other factors that
influence voice quality for conversation (such as absolute delay, listening level, loudness loss, sidetone
and echo) and that can be particularly important for voice over IP. The results can be converted into
numbers on the MOS scale in ways that predict the MOS for listening to some extent.
Objective assessments for conversation use the E-Model, described in G.107
87
. The results take into
account many factors that influence voice quality for conversation but are not yet established
experimentally in some cases (such as when low bit rate voice encodings are used together or with room
noise). For network planning the results are calculated using reference connections devised by operators,
but for quality of service monitoring they are calculated using typical connections experienced by
customers, preferably with many of the E-model factors derived from In-service Non-intrusive Monitoring
Device (INMD) measurements as in P.562
88
. The results can be converted into numbers on the MOS scale
in ways that predict the MOS for conversation to some extent.
Measurements of perceived or predicted quality made by different methods are unlikely to be comparable
with each other, so to achieve comparability all operators should use the same method.
Quality assessments for video are less well developed than those for voice, though work has been done on
subjective and objective assessments for viewing digital television, for example, as in J .144
89
. Even the audio
requirements can range from those for telephony to those for surround sound in interactive gaming. J ust as
with voice there is a distinction between single messages (as in video streaming) and interactive
conversations. At this stage video telephony services, for example, are likely to be treated in quality of service
monitoring as if they were voice telephony services (so that aspects like lip synchronisation are ignored) but
some targets can become inappropriate: the call setup time, for example, may be greater for video telephony
than for voice telephony.
Quality assessments for text can usually have simple tests of matching between the sent and received texts.
3.4.2 Passive and active methods
Measurements of traffic handling (such as those for the call setup success ratio) can use real traffic or
test traffic. More fully:
Measurements using real traffic require network equipment to count the traffic. They are passive
or non-intrusive, in that they do not add extra traffic to the network. They may not be

42
immediately comparable even within individual networks, because equipment from different
suppliers can have different counting mechanisms. (Also, they do not typically match the
experiences of users because the traffic is counted only if it gets into the networks along the
wireline or wireless links from their edges.)
Measurements using test traffic require customer equipment from which test calls are made at the
edge of the network, so the experiences of users can be represented. They are active or
intrusive, in that they add extra traffic to the network and could lead to extra disruption and
expense, especially for a poorly performing network. They are comparable between different
networks only if different operators make them at similar times, between similar places and in
similar circumstances. In particular, drive-around tests for mobile networks that are co-ordinated
between the operators in the main geographic areas lead to measurements that are more comparable
and less costly
90
.
If measurements of traffic handling are to be comparable, they have to use real traffic for all operators
or test traffic for all operators. However, operators may well wish to supplement measurements of test
traffic with their own measurements of real traffic (such as off-line analyses of signalling records), to
understand network performance better, especially at critical points.
Measurements using test traffic need enough tests to give the required confidence in the
measurements
91
. The traffic would pass from traffic-weighted locations to traffic-weighted locations
during a part of the day chosen to balance the wish to find how networks perform during the busy hour
(when the traffic is most intense) with the wish to economise by spreading tests through the day.
3.4.3 Network-by-network and end-to-end methods
The quality of service provided for a network path depends on all the networks that the path goes
through. (For example, as illustrated in Figure 6, the delay between an end user on network X and an
end user on network Z that passes through a network Y is the sum of three delays, which are from the
end user in X to a point of interconnection of X with Y, from the point of interconnection of Y with X
to a point of interconnection of Y with Z, and from the point of interconnection of Z with Y to the end
user in Z.)
Measurements of traffic handling (such as those for delay and packet loss) can be network-by-network
or end-to-end. More fully:
In network-by-network measurements, the individual networks are considered in isolation, so an
end-to-end delay along a path is calculated from the delays across the individual networks used by
the path. The measurements for a network therefore do not represent what customers experience, at
least for paths ending in other networks, but they do represent what the operator controls.
In end-to-end measurements, the individual networks are considered in combination, so an end-to-
end delay along a path is measured directly (and the delay averaged over all paths through a
network takes account of paths ending outside the network). The measurements for a network
therefore represent what customers experience on average, but they do not represent what the
operator controls.
End-to-end measurements are more helpful to customers than are network-by-network ones, as they
concentrate on aspects of services that customers experience directly. They can be fair to operators if
published measurements are annotated to explain service deficiencies due to other operators. (For
resale, too, measurements should concentrate on aspects of services that customers experience
directly, by ensuring that, for example, fault repair times include both the times taken by resellers to
process fault reports and the time taken by the original operators of the resold networks to repair
faults.)
Figure 6 shows the distinction between network-by-network and end-to-end measurements when there
are three networks (X, Y and Z), one of which is providing transit services to the other two. (It is
applicable to other measurements, such as packet loss, as well as delay.)

43
Figure 6 Examples of measurements across three networks
3.5 Services to be monitored
Services arise from complex combinations of customer wishes and network implementations. For
quality of service monitoring, particularly in an NGN environment, they can be analysed in terms of
the following notions:
Layers in services.
Network segments used by services.
Aspects of services.
3.5.1 Layers in services
Telephony originally combined transmission, switching (over access and core networks) and voice
communication. One operator would supply all these in combination, perhaps along with some content
services such as directory enquiries. Quality of service monitoring covered the combined service.
Customers never noticed that the implementation had transmission and switching layers.
Then other services were introduced, such as narrowband (dialup) internet access. These other services
could be regarded as either as applications on a par with directory enquiries or as different forms of
communication. Often they had poor quality levels during their early years but were nonetheless
exempt from quality of service monitoring to avoid inhibiting their growth.
Now, however, there are more new services. Some of these are parts of former combined services that
are now sold separately. (For example, carrier preselection separates access and core networks and
voice over IP splits out voice communication.) Splitting services into separately sold parts may be
even more common in an NGN that lets many services use one network. Figure 7 outlines roughly
how the current network environment, in which different services are supported by different networks,
is giving way to the NGN environment, in which different services are supported by shared networks.
(Of course, there are more services and networks than are shown in Figure 7: large operators may have
50 separate networks.)

network X network Y network Z
end user
network-by-network measurements:
end-to-end measurements:
p
a
t
h
measured by
operator X measured by
operator Y measured by
operator Z measured by
operator Z
measured by
operator Z
point of
interconnection
point of
interconnection
end user
network X network Y network Z
end user
network-by-network measurements:
end-to-end measurements:
p
a
t
h
measured by
operator Z
measured by
operator Z
point of
interconnection
point of
interconnection
end user

44
Figure 7 The transformation from current networks to next generation networks

The layers in the implementation are changing to separate more sharply content, communication,
network and transmission. The changes in the layers come with changes in businesses. For instance:
Operators that operate in single layers need both wholesale services to interconnect networks (the
horizontal arrows in Figure 7) and wholesale services to use adjacent layers (the vertical arrows in
Figure 7). For the regulators, layer-by-layer regulation becomes important alongside, or sometimes
instead of, service-by-service regulation. Quality of service measurements of traffic handling for
interconnecting networks (such as those for delay and packet loss) can be used both in assessing the
end-to-end quality of that layer and in designing the layer above. Quality of service measurements
for customer support may be needed both within and between layers.
Operators that operate in several layers create new bundles of services straddling those layers,
especially by getting exclusive rights to content: they introduce walled gardens, in which their
customers access particular content with particular quality of service, instead of other content over
the public internet. For their customers, making informed choices becomes harder, because many
services need to be considered at once and tariff packages become more complicated. For their
competitors, offering equivalent services becomes harder if the new bundles exploit exclusive
rights. Quality of service measurements for the individual services are important for comparing
disparate bundles from different operators.
Operators that control access networks and supply voice communication services want to charge
for the use of suitable quality IP, which can support real time services such as voice
communication better than can current best effort IP, at least when the traffic is fairly heavy
92
.
Their attitudes are discussed briefly in Section 3.5.2. Quality of service measurements for best
effort IP may be needed to confirm that charging for suitable quality IP is not a strategem for
lowering the quality level of best effort IP to encourage customers to buy suitable quality IP.
Operators that formerly bundled user terminals in their services need to support many differently
perfoming terminal types and application implementations. Quality of service measurements of
traffic handling must either identify or exclude the terminal types and application implementations.
Operators may offer new real time services. The quality requirements and expectations of users
(including people with disabilities) need to be understood. Quality of service measurements need to
be defined, even if they are not used in quality of service monitoring by regulators.
Operators may exploit new terminals. Quality of service measurements of traffic handling could be
made end-to-end by providing suitable terminal capabilities.
multiple media (text, voice, video) communication
mobile
switched
network
transmission
fixed
switched
network
transmission
suitable quality IP routed and switched network
content
voice
com-
muni-
cation
text
com-
muni-
cation
voice
com-
muni-
cation
content content content
text
com-
muni-
cation
voice
com-
muni-
cation
best effort IP
routed
network
content
multiple media (text, voice, video) communication
mobile
switched
network
transmission
fixed
switched
network
transmission
suitable quality IP routed and switched network
content
voice
com-
muni-
cation
text
com-
muni-
cation
voice
com-
muni-
cation
content content content
text
com-
muni-
cation
voice
com-
muni-
cation
best effort IP
routed
network
content

45
Though these changes in the layers are significant, for the regulator they may mainly require the
application of existing techniques in slightly new situations.
The scope of quality of service monitoring differs between different layers. In particular:
Services in the content layer have little quality of service monitoring. The main exceptions are
operator assistance calls, directory assistance calls and emergency calls, which may be monitored
for the frequency and speed with which they are answered. (Other calls to operator call centres may
also be conveniently monitored in this way, even though their subjects are not specific to the
telecommunications industry.) For other services in the content layer, the content itself may be
monitored for other reasons, such as ensuring legality and conformance with rules about the
contents and lengths of calls with particular prices. However, this monitoring is not usually
regarded as a form of quality of service monitoring and may be performed by organisations quite
different from telecommunications regulators.
Services in the communication layer, such as telephony and messaging, are currently the main
focus of quality of service monitoring for end users. The monitoring is discussed in Section 3.5.3.
Services in the network layer providing IP routing and switching have increasing importance for
quality of service monitoring when they can be bought separately from communication services.
Currently internet services, and managed business IP network services such as IP virtual private
networks, can be bought in this way. In the future IP network services with different quality levels
are likely to become available more widely. At least when they are provided by dominant
operators, monitoring their quality can help both end users and alternative operators in the same
layer or the layer above.
Services in the transmission layer are bought by end users (corporate customers constructing their
own private networks) and by other operators. The quality of service characteristics of the
transmission, such as the bit error ratio of a digital leased line, are likely to be monitored by
corporate customers that are strong enough to need only occasional regulatory intervention in retail
services, but other operators buying wholesale services may need regulatory support.
Effective regulation of wholesale services can stimulate the growth of new operators, in the same layer
or the layer above. Ultimately there might be enough operators that customers have adequate choice.
There could still be a need for quality of service measurements for retail services to inform customers,
but there would be no need for quality of service targets for retail services, just as there would
ultimately be no need for retail price control
93
.
3.5.2 Network segments used by services
Regulatory requirements such as carrier preselection and local loop unbundling draw attention to the
distinction between access and core networks. Figure 8 illustrates this distinction in a simplified form.
(It ignores, for example, the possibility that aggregation or distribution networks may lie between
access and core networks.)

Figure 8 Access and core networks
end user
p
a
t
h
end user
access network
(typically star-
connected lower
bandwidth links)
core network
(typically mesh-
connected higher
bandwidth links
access network
(typically star-
connected lower
bandwidth links)
end user
p
a
t
h
end user
access network
(typically star-
connected lower
bandwidth links)
core network
(typically mesh-
connected higher
bandwidth links
access network
(typically star-
connected lower
bandwidth links)

46

Even in the current network environment, quality problems are more usual in access networks than in
core networks, because the network equipment has more vulnerable locations and is less easily
duplicated for protection. However, any shortages of bandwidth for voice communication are due to
core networks and mobile networks, not to fixed networks: in fixed networks the availability of
bandwidth for a call is usually guaranteed if the call setup is successful.
In an IP network the availability of bandwidth is not guaranteed in this way. (Part of the work on
creating standards for the NGN environment concerns letting IP networks have properties like this.)
Typically the bandwidth is more severely limited in IP access networks than in IP core networks. IP
access networks are likely to differentiate between packets associated with different services, so that
they can offer shorter delays and fewer losses to some services. IP core networks can sometimes be
implemented without doing this (though they need to maintain the markings that differentiate between
packets traversing them, to ensure similar behaviours in the originating and terminating IP access
networks). Measurements of delay and packet loss for IP access networks, and the corresponding
charges for interconnection, might therefore need to relate to specific services, even if those for IP
core networks do not do so.
Thus in IP access networks there might be a need to provide extra mechanisms so that real time
services such as voice communication can attain high quality levels. (These mechanisms might not be
needed to accommodate just one voice call per IP access line, but they could be needed for multiple
video calls, for example.) In particular, operators controlling access networks wish to be compensated
by other operators for providing these mechanisms and for losing revenue from their own voice
communication services. They might also wish to increase the appeal of their own services, ranging
from voice communication to video content, by bundling them together with the IP access network
service.
This is why some operators in the US oppose regulation that supports network neutrality. (The term
loosely describes avoiding some forms of discrimination between services and between operators in
access and core networks.) Opponents of regulation to support network neutrality argue that operators
should be able to charge other operators extra for supporting services that use extra network resources.
Proponents of regulation to support network neutrality argue that without it operators might (and in
some cases do already) discriminate in favour of their own services, even to the extent of blocking
voice over IP. Both opponents and proponents of regulation can claim to be favouring investment
(particularly in new networks, in the case of opponents, and in new applications, in the case of
proponents). Whether any such regulation is required can vary between countries: sharp separations
between wholesale and retail services of dominant operators, and competition through local loop
unbundling and other forms of resale could be enough.
3.5.3 Aspects of services
Usually services have many aspects, each of which may need its quality to be characterised. There can
therefore be many measurements to be defined. Several models have been devised to help with
defining measurements systematically
94
. This paper mainly draws a distinction between measurements
for customer support and measurements dealing with other aspects of services.
In many respects customer support is independent of services: all services require contact with
customers, for supplying them, resolving complaints, and repairing faults. Measurements of quality
related to customer support are therefore essentially independent of services: they concern the
numbers of times that current or potential customers make requests to operators (as measured by the
proportion of bills queried, for example) and the responsiveness of operators to requests (as measured
by the time taken to satisfy customers querying bills, for example). The most widely used of them
relate to supply (the service supply time or timeliness), faults (the fault repair time or timeliness and
the fault report rate), complaints (the complaint resolution time or timeliness and the complaint rate)
and billing (the bill error ratio).

47
To customers, measurements of quality related to other aspects of services also seem largely
independent of services. Even measurements that might be needed for future services are fairly
predictable (though the importance of those measurements to customers may change over time and the
targets may be very different for different services). Box 15 provides a model that describes these
measurements. (It is expressed in terms of services in the communication layer but is more generally
applicable, especially in conjunction with the naming conventions of Box 3.)
Though customer support is likely to receive attention for many services in many layers, in an NGN
environment quality of service monitoring by regulators is otherwise likely to be confined to those
services in the communication and network layers that most concern customers.

Box 15 Measurements related to communication
The user requirements for services often follow a pattern. In particular:
Users want to start uses of services successfully (dependably and speedily). For example, they want to set
up calls, transmit SMS messages, log into internet sessions or send emails. The quality is characterised by
the probability of starting uses successfully and the time taken by successful starts. Defining
measurements can be systematic and straightforward (for calls, for example, leading to the call setup
success ratio and the call setup time). However, the measurements are not necessarily important to users:
the probability of SMS messages transmitted successfully may be too low, but the time taken by successful
SMS transmissions may be satisfactory enough not to be measured. For services that users do regard as
having starts (such as GPRS and broadband internet sessions) measurements are usually unnecessary. Here
successful needs to be interpreted relative to user understanding of the service: for store-and-forward
services like SMS, MMS and email successful can mean getting to or from the central point of storage
and sending and receiving messages can be treated as separate aspects of the service that need separate
measurements.
Users want to continue uses of services for long enough to complete their communications. For example,
they want to use the connections made by successful call setups and internet session logins without
perceiving any breaks in them until breaking them deliberately. The quality is characterised by the
probability of continuing uses for long enough. Defining measurements can be systematic and
straightforward (for calls, for example, leading to the call retention ratio). For services that users do not
perceive as having connections (such as SMS) measurements are usually unnecessary. When
measurements are made using test traffic (not real traffic) long enough needs to be related to the
distribution of lengths wanted by customers, such as the durations of real calls.
Users want to use services for communication accurately. For example, they want the connections made
by successful call setups to allow voice conversations. The quality is characterised by user perceptions.
These may depend on the intended application: network delays, for example, are much more relevant to
interactive conversations than to single messages. Defining measurements is straightforward for services
that just provide only text messages but can be problematic for other services, especially if they provide
voice or video conversations.
These aspects are essentially generalisations of those referred to as accessibility, retainability and
integrity in ITU-T recommendation E.800 (though the terms accessibility, continuity and fidelity might
be more appropriate)
12
. Accessibility itself essentially comprises attributes referred to as reliability and
speed in I.350
95
. The ETSI and 3GPP work on mobile services follows approximately similar lines
96
. Other
models are discussed in G.1000
94
.

48
Conclusion
This paper starts to assemble and analyse information about current quality of service monitoring by
regulators that can be used to guide future schemes. It describes the motivation (in Section 1),
operational arrangements (in Section 2) and selection of areas of focus (in Section 3) for quality of
service monitoring. On the basis of experience so far it puts forward detailed guidelines for discussion
for each of these topics, as well as overall guidelines.
Clearly there is still much to be done to extend and examine the information. However, the paper does
at least point to possible ways forward.

49
Abbreviations
3GPP Third Generation Partnership Project
ACA Australian Communications Authority
ACIF Australian Communications Industry Forum
ACMA Australian Communications and Media Authority
Anatel Agncia nacional de telecomunicaes
Cintel Centro de investigacin de telecommunicaciones en Colombia
ComReg Commission for communications Regulation
CRT Comisin de Regulacin de Telecomunicaciones
CRTC Canadian Radio-television and Telecommunications Commission
ECTEL Eastern Caribbean TELecommunications authority
ETSI European Telecommunications Standards Institute
EU European Union
FCC Federal Communications Commission
GPRS General Packet Radio Service
GSM Global System for Mobile communications
ICT Information and Communication Technology
IETF Internet Engineering Task Force
IIA Internet Industry Association
INMD In-service Non-intrusive Measurement Device
IP Internetwork Protocol
ITU International Telecommunication Union
ITU-D ITU Telecommunication Development Sector
ITU-R ITU Radiocommunication Sector
ITU-T ITU Telecommunication Standardisation Sector
MMS Multimedia Messaging Service
MOS Mean Opinion Score
NGN Next Generation Network
ODTR Office of the Director of Telecommunications Regulation
OECD Organisation for Economic Co-operation and Development
Ofcom Office of communications
Oftel Office of telecommunications
PESQ Perceptual Evaluation of Speech Quality
SMS Short Messaging Service
Subtel Subsecretara de telecomunicaciones
TCP Transmission Control Protocol
TRAI Telecom Regulatory Authority of India
TRC Telecommunications Regulatory Commission
UDP User Datagram Protocol
UK United Kingdom
US United States

50
Endnotes

1
In this paper the term regulator covers government ministries, not just separate regulators, and the term
operator covers all service providers, not just network operators.
2
See Cintel, Brief Estudio de Mercado 2005, February 2006, at http://www.cintel.org.co/noticia.php3?nt=4716.
3
See Oftel, Reporting Quality of Service Information to Consumers: A consultation on the regulatory options for
acquiring and publishing quality of service measures for telecommunication services, November 2003, at
http://www.ofcom.org.uk/consult/condocs/qualitystate/qosoriginal/qos.pdf.
4
See Oftel, Consumers use of fixed telephony: Q14 August 2003, October 2003, at http://www.ofcom.org.uk/
static/archive/oftel/publications/research/2003/q14fixres1003.pdf.
5
For the adjudication by the industry group that regulates advertising see Advertising Standards Authority,
Non-broadcast Adjudication, November 2005, at http://www.asa.org.uk/asa/adjudications/non_broadcast/
Adjudication+Details.htm?Adjudication_id=40467.
6
See TRAI, Report on QoS Parameters related to Congestion on Point of Interconnections (POIs) of CMSPs,
J anuary 2006, at http://www.trai.gov.in/trai/upload/Reports/2/report17jan06.pdf.
7
For a discussion in economic terms of such problems with dominant operators see Sappington, D. E. M.,
Regulating Service Quality: A Survey, Journal of Regulatory Economics, November 2005, Volume 27, Issue
2, Pages 123154.
8
For interconnection, wholesale service targets for measurements of traffic handling are more demanding than
the corresponding retail service targets (because they consider partial paths, between a point of interconnection
and an end user, not complete paths, between one end user and another). For resale, wholesale service targets for
measurements of traffic handling are the same as the corresponding retail service targets, but wholesale service
targets such as that for the fault repair time may need to be more demanding than the corresponding retail service
targets (to allow time for customer care procedures, for example).
9
Residential broadband internet service exemplifies this in several countries. The operators may well require
contracts lasting one or two years, to cover the costs of equipment in the customer premises and in the access
network. They may well also provide email addresses, such as customer_name@isp_name.com, that are not
portable from their networks, and storage that is not accessible only from outside their networks. Nonetheless
there may be no co-ordinated published quality of service monitoring.
10
Reducing churn might be a reason for introducing lifetime subscriptions in India, where monthly churn rates
vary between 3.5% and 6.0% approximately.
11
Also, customer satisfaction surveys can be misleading: customers that have changed their operators may not
want to admit that doing so was unwise, so satisfaction with the alternative operators to whom they have
changed may appear higher than satisfaction with the dominant operators from whom they have changed. This
effect can be reduced if the survey sample size is made large enough to distinguish each individual operator, but
doing this can be expensive. Thus customer satisfaction surveys that produce results about individual operators
are usually insufficient on their own: in Australia and Colombia, for example, such surveys are performed every
12 months and every 6 months respectively as accompaniments to the quality of service measurements.
12
See ITU-T Recommendation E.800, Terms and definitions related to quality of service and network
performance including dependability, August 1994, at http://www.itu.int/rec/T-REC-E.800.
13
See ITU-T Recommendation E.600, Terms and definitions of traffic engineering, March 1993, at
http://www.itu.int/rec/T-REC-E.600.
14
When it was state-owned the incumbent operator in the UK published quality of service measurements. It
stopped doing so for commercial reasons in preparation for privatisation. It resumed doing so in October 1987,
under pressure from the regulator.
15
An industry group could be either self-regulatory or co-regulatory. The distinction between these is often
rather blurred. One way of making the distinction sharp is to say that the regulator does not participate in a self-
regulatory group (except to require that the group fulfil its mandate and perhaps to provide the secretariat) but
that the regulator does participate in a co-regulatory group. The role of such groups has been formalised in

51

Australia, for example, where industry groups can submit codes of practice to the regulator, who can then
register them and make them compulsory.
16
See TRAI, Consultation Paper on Review of Quality of Service (QOS) Parameters of Basic and Cellular
Mobile Telephone services, February 2005, at http://www.trai.gov.in/trai/upload/ConsultationPapers/13/
conpaper23feb05.pdf.
17
For links to the reports see http://www.trai.gov.in/Reports_list_year.asp.
18
See TRAI, Regulation on Quality of Service of Basic and Cellular Mobile Telephone Services, 2005, J uly
2005, at http://www.trai.gov.in/trai/upload/Regulations/3/regu1jul05.pdf.
19
For links to the reports see http://www.subtel.cl/servlet/page?_pageid=58&_dad=portal30&_schema=
PORTAL30.
20
See Subtel, Fija indicadores de calidad del servicio de acceso a internet y sistema de publicidad de los
mismos, J une 2001, at http://www.subtel.cl/pls/portal30/docs/folder/wsubtel_contenidos_sitio/subtel/
marcoregulatorio/norm_tec_internet/norma_tecnica_internet/res_669_calidad_de_serv_internet.pdf.
21
See Subtel, Fija indicadores de calidad de los enlaces de conexin para cursar el trfico nacional de Internet
y sistema de publicidad de los mismos, J une 2000, at http://www.subtel.cl/pls/portal30/docs/folder/
wsubtel_contenidos_sitio/subtel/marcoregulatorio/norm_tec_internet/norma_tecnica_internet/res_698_trafico_in
ternet.pdf.
22
Figures provided by 4 operators in the UK showed variations by factors of 74 in costs per customer and 34 in
costs per minute (though variations by factors of 7 and 5 were perhaps more usual, with the high costs being
outliers). Organisations making measurements generally regarded the costs as very low. See:
Oftel, Reporting Quality of Service Information to Consumers: A consultation on the regulatory options for
acquiring and publishing quality of service measures for telecommunication services, November 2003, at
http://www.ofcom.org.uk/consult/condocs/qualitystate/qosoriginal/qos.pdf.
Ofcom, A Statement on providing quality of service information to consumers: A Consultation on quality
parameters including a Notification and Draft Direction, September 2004, at http://www.ofcom.org.uk/
consult/condocs/qualitystate/qos/qos.pdf.
23
For links to the reports see http://www.askcomreg.ie/about_us/Consumer%5FInformation.53.LE.asp#306.
24
For the database see http://sistemas.anatel.gov.br/sgiq/Default.asp?SISQSmodulo=6347&SISQSsistema=540.
25
See Anatel, Regulamento de Indicadores de Qualidade do Servio Telefnico Fixo Comutado RIQ, October
2005, at http://www.anatel.gov.br/biblioteca/Resolucao/2005/Anexo_res_417_2005.pdf.
26
For the database see http://sistemas.anatel.gov.br/sgiq/Default.asp?SISQSmodulo=8463&SISQSsistema=164.
27
See Anatel, Regulamento de Indicadores de Qualidade do Servio Mvel Pessoal SMP, April 2003, at
http://www.anatel.gov.br/biblioteca/Resolucao/2003/anexo_res_335_2003.pdf.
28
Doubts about whether the benefits of publication outweigh the costs may be why in 2003 measurements were
published in only 6 of the 15 member countries of the EU, despite apparent obligations to publish under the
directives. At the same time 10 member countries imposed targets for fixed telephony universal operators and 3
member countries imposed targets also on other fixed telephony operators. See European Commission, Ninth
report on the implementation of the Electronic Communications Regulatory Package, Annex 2, COM(2003) 715,
December 2003, at http://europa.eu.int/eur-lex/lex/LexUriServ/site/en/com/2003/com2003_0715en01.pdf.
29
See TRC, Explanatory Memorandum on the Quality of Service Framework, February 2006, at
http://www.trc.gov.jo/Static_English/doc/Technical/Explanatory Memorandum on the Quality of Service
Framework.pdf.
30
See TRAI, TRAIs Direction to all Service Providers on Provision of Interconnection, J une 2005, at
http://www.trai.gov.in/trai/upload/PressReleases/79/pr7jun05.pdf.
31
See TRAI, TRAI issues show cause notices to mobile operators, March 2006, at http://www.trai.gov.in/trai/
upload/PressReleases/5/pr7mar06.pdf.

52

32
In Chile customers are entitled to reductions in monthly rentals if faults persist for at least 12 hours and to
rebates if faults persist for at least 3 consecutive days. In India customers are entitled to rebates if faults persist
for at least 3 consecutive days (with rebates amounting to 7 days for faults repaired in at most 7 days, 15 days for
faults repaired in at most 15 days, and 1 month otherwise). See:
Subtel, Acerca del contrato de servicio, August 2006, at http://www.subtel.cl/servlet/page?_pageid=58&
_dad=portal30&_schema=PORTAL30.
TRAI, Regulation on Quality of Service of Basic and Cellular Mobile Telephone Services, 2005, J uly 2005, at
http://www.trai.gov.in/trai/upload/Regulations/3/regu1jul05.pdf.
33
There are 13 measurements for retail services and 24 measurements for wholesale services. Of the 24
measurements for wholesale services, 21 are made competitor-by-competitor, so that compensation can be
directed to the correct competitors. See:
CRTC, Telecom Decision CRTC 2005-20 Finalization of quality of service rate rebate plan for competitors,
March 2005, at http://www.crtc.gc.ca/archive/ENG/Decisions/2005/dt2005-20.htm.
CRTC, Telecom Decision CRTC 2005-17 Retail quality of service rate adjustment plan and related issues,
March 2005, at http://www.crtc.gc.ca/archive/ENG/Decisions/2005/dt2005-17.htm.
34
The fines amounted to more than BRL4300000 (USD1600000). The problems largely, but not solely,
concerned the service supply time, the fault repair time, the complaint resolution time and the call setup success
ratio. See:
Anatel, Anatel multa operadoras de telefonia fixa em R$ 2,6 milhes por descumprirem metas de qualidade,
May 2005, at http://www.anatel.gov.br/index.asp?link=/biblioteca/Releases/2005/release_03_05_2005(5).pdf.
Anatel, Anatel multa Telemar em R$ 107,5 mil por descumprir metas de qualidade, April 2005, at
http://www.anatel.gov.br/Tools/frame.asp?link=/biblioteca/releases/2005/release_07_04_2005.pdf.
Anatel, Descumprimento de Metas de Qualidade na telefonia fixa gera mais R$ 1,3 milho em multas, March
2005, at http://www.anatel.gov.br/Tools/frame.asp?link=/biblioteca/releases/2005/release_22_03_2005.pdf.
Anatel, Anatel aplica mais de R$ 300 mil em multas contra a Telemar e a Brasil Telecom, March 2005, at
http://www.anatel.gov.br/Tools/frame.asp?link=/biblioteca/releases/2005/release_17_03_2005(2).pdf.
35
For example, there is a theoretical justification for the view that penalties proportional to profits may increase
investment in quality but that penalties proportional to revenues may increase or decrease investment in quality
according to circumstances, provided that (in both cases) they are less than the maximum values permitted by the
law. See Weisman, D. L., Price regulation and quality, Information Economics and Policy, March 2005,
Volume 17, Issue 2, Pages 165-174.
36
For links to the reports seehttp://www.acma.gov.au/ACMAINTER:STANDARD::pc=PC_1378.
37
See ACMA, Telecommunications (Customer Service Guarantee) Standard 2000 (No. 2), February 2004, at
http://www.acma.gov.au/acmainterwr/consumer_info/csg/csg_consolidatedstandard04.pdf.
38
See ACMA, Telecommunications Industry Monitoring and Reporting Arrangements, August 2004, at
http://www.acma.gov.au/acmainterwr/aca_home/issues_for_comment/discussion/archive/telecommunications_in
dustry_monitoring_and_reporting.pdf.
39
See ACMA, Telecommunications Performance Report 2004-05, November 2005, at http://www.acma.gov.au/
acmainterwr/_assets/main/lib100449/tpr 2005.pdf.
40
For links to the codes see http://www.acif.org.au/ACIF_documents/codes.
41
For links to the codes see http://www.iia.net.au/index.php?option=com_content&task=section&id=3&
Itemid=33.
42
For the bar charts see http://www.topcomm.co.uk/.
43
See Ofcom, A Statement on providing quality of service information to consumers: A Consultation on quality
parameters including a Notification and Draft Direction, September 2004, at http://www.ofcom.org.uk/
consult/condocs/qualitystate/qos/qos.pdf.
44
For the coverage maps see http://www.topnetuk.co.uk/.

53

45
For the database see http://www.sui.gov.co/reportesSUI/SUI_ReporteTele.htm.
46
See CRT, Resolucin 1250 de 2005, J une 2005, http://www.crt.gov.co/Documentos/Normatividad/
ResolucionesCRT/00001250.pdf.
47
See CRT, Resolucin 575 de 2002, December 2002, http://www.crt.gov.co/Documentos/Normatividad/
48
See CRT, Resolucin 338 de 2000, December 2000, http://www.crt.gov.co/Documentos/Normatividad/
49
For links to the reports see http://www.crt.gov.co/crt_2001-2004/paginas/internas/biblioteca/regulatorio_b.htm.
50
For the database see http://svartifoss2.fcc.gov/eafs/adhoc/table_year_tab.cfm?reportType=4305.
51
See:
Ai, C., and Sappington, D. E. M., The Impact of State Incentive Regulation on the U.S. Telecommunications
Industry, Journal of Regulatory Economics, September 2002, Volume 22, Issue 2, Pages 133-159.
Banerjee, A., Does Incentive Regulation Cause Degradation of Retail Telephone Service Quality?,
Information Economics and Policy, J une 2003, Volume 15, Issue 2, Pages 243-269.
Clements, M. E., Local Telephone Quality of Service: A Framework and Empirical Evidence,
Telecommunications Policy, J une-July 2003, Volume 28, Numbers 5-6, Pages 413-426.
Roycroft, T. R., and Garcia-Murrilo, M., Trouble Reports as an Indicator of Service Quality: The Influence
of Competition, Technology, and Regulation, Telecommunications Policy, November-December 2000,
Volume 24, Numbers 10-11, Pages 947-967.
52
See Sappington, D. E. M., The Effects of Incentive Regulation on Retail Telephone Service Quality in the
United States, Journal of Network Economics, December 2003, Volume 3, Issue 4, Pages 355-375, at
http://www.rnejournal.com/articles/sappington-RNE_5_dec_03final.pdf.
53
See Uri, N. D., Service Quality Effects of Incentive Regulation on Access Service in Telecommunications in
the United States, European Journal of Law and Economics, November 2003, Volume 16, Issue 3, Pages 369-
390.
54
See Zimmerman, P. R., Regional Bell Operating Company Entry Into Long-Distance and Non-Price
Discrimination Against Rival Interexchange Carriers: Empirical Evidence from Panel Data, Applied Stochastic
Models in Business and Industry, November 2003, Volume 19, Issue 4, Pages 269290.
55
For example, in the UK by September 1989 the main incumbent operator could repair 86% of faults in 1
working day, compared with 65% in September 1988. The improvement was attributed to network
modernisation spurred by competition. At about the same time the operator, under pressure from the regulator,
started to pay compensation for excessive times for service supply and fault repair. If the committed time for
service supply or fault repair was missed by 2 working days, customers were paid GBP5 (USD8.4) per working
day, with up to GBP1000 (USD1700) for residential customers and GBP5000 (USD8400) for business
customers if they suffered financially.
56
For suitable correlations, showing that in the US more network modernisation has occurred with incentive
regulation than with rate of return regulation, see Ai, C., and Sappington, D. E. M., The Impact of State
Incentive Regulation on the U.S. Telecommunications Industry, Journal of Regulatory Economics, September
2002, Volume 22, Issue 2, Pages 133-159.
57
For a summary of the work of ITU and other organisations on quality of service see ITU, Quality of Service
and Network Performance, March 2004, at http://www.itu.int/pub/T-HDB-QOS.02-2004.
58
Other statistics besides averages, such as variances, can also help with showing variability but not immediately
in identifying the cause of variability.
59
See Milne, C. B., and Mitchell, J ., Quality of Service indicators for the telephone service to residential
consumers, Final Report to the Commission of the European Communities, Polytechnic of Central London,
October 1989.

54

60
See European Commission, Directive 98/10/EC of the European Parliament and of the Council of 26
February 1998 on the application of open network provision (ONP) to voice telephony and on universal service
for telecommunications in a competitive environment, February 1998, at http://portal.etsi.org/public-interest/
Documents/Directives/Standardization/Directive_1998_10.pdf.
61
See European Commission, Directive 2002/22/EC of the European Parliament and of the Council of 7 March
2002 on universal service and users' rights relating to electronic communications networks and services
(Universal Service Directive), March 2002, at http://portal.etsi.org/public-interest/Documents/Directives/
Standardization/Universal_Service_Directive.pdf.
62
See OECD, Quality of service indicators, Chapter 5 of Performance indicators for public
telecommunications operators, November 1990, at http://www.oecd.org/dataoecd/34/59/2366923.pdf.
63
See OECD, Quality of service, Chapter 7 of Communications Outlook 2003, May 2003, at
http://www1.oecd.org/publications/e-book/9303021E.pdf.
64
An example is sometimes provided by the call setup success ratio. For 35 operators in 20 developing
countries, a customer uses between 200 and 1,600 minutes per year of mobile telephony. If the average call
duration is 2 minutes, the customer using 200 minutes per year may not distinguish between call setup success
ratios of 99% and 99.9%, and the customer using 1,600 minutes per year may not distinguish between call setup
success ratios of 99.9% and 99.99%. When the customer is actually selling the use of the telephone informally,
the call setup success ratio could well be unimportant.
65
Similarly 33% of complaints to the regulator in Germany in 1999 were about incomprehensible bills, 41% of
complaints to the regulator in the US in 2000 were about inaccurate information, and 28% of complaints to the
ombudsman in Australia in 2005 concerned prepaid or postpaid billing, almost independently of the services
(fixed telephony, mobile telephony and internet). See:
OECD, Quality of service, Chapter 8 of Communications Outlook 2001, May 2001, at
http://www1.oecd.org/publications/e-book/9301021E.pdf.
ACMA, Telecommunications Performance Report 2004-05, November 2005, at http://www.acma.gov.au/
acmainterwr/_assets/main/lib100449/tpr 2005.pdf.
66
For an analysis suggesting that this is so, by relating investment types to the quality of particular aspects of
services in the US, see Clements, M. E., Local Telephone Quality of Service: A Framework and Empirical
Evidence, Telecommunications Policy, J une-J uly 2003, Volume 28, Numbers 5-6, Pages 413-426.
67
For definitions of 5 measurements designed by an industry group (in a process lasting over a year) and
intended to achieve comparability between fixed telephony operators see Topcomm, Quality of Service
Definitions (Applicable to Quarter 1, 2006), J uly 2006, at http://www.topcomm.co.uk/topcomm_QOS.pdf.
68
A co-ordinated effort to define measurements for regulators and operators spanning several countries might
provide comparability between operators in different countries. It could also reduce costs, for regulators and
operators alike. It could be a response to the increasing tendency of mobile telephony operators to become
multinational. Perhaps ECTEL, with the same dominant operator in all its countries, can aspire to such co-
ordination. (Measurement definitions at the level of the ETSI ones for the EU leave too many options open to
ensure comparability.)
69
Targets were originally things to aim for. Now in widespread usage (as in this paper) they are things for which
there are penalties if they are missed. Arguably targets that represent minimum required quality levels should be
called something other than targets, but no other term is much more satisfactory.
70
Also, in some countries of the EU, where calling parties pay for calls, the mobile telephony operators are seen
as dominant in the market for receiving calls using mobile access facilities.
71
Subjective voice quality assessments distinguish differential tests from absolute tests accordingly. See ITU-T
Recommendation P.800, Methods for subjective determination of transmission quality, August 1996, at
http://www.itu.int/rec/T-REC-P.800.
72
In fact the requirement in the Philippines is to have no more 1 error per 1000 bills per year, so it is particularly
stringent if bills are sent every 3 months.
73
See:

55

ITU, ITU-T Study Group 2 (Study Period 2005 - 2008), August 2006, at http://www.itu.int/ITU-
T/studygroups/com02/index.asp. ITU-T Study Group 2 is responsible for studies relating to: principles of
service provision, definition and operational requirements of service emulation; numbering, naming,
addressing requirements and resource assignment including criteria and procedures for reservation and
assignment; routing and interworking requirements; human factors; operational aspects of networks and
associated performance requirements including network traffic management, quality of service (traffic
engineering, operational performance and service measurements); operational aspects of interworking
between traditional telecommunication networks and evolving networks; and evaluation of feedback from
operators, manufacturing companies and users on different aspects of network operation.
T/studygroups/com12/index.asp. ITU-T Study Group 12 is responsible for Recommendations on the end-to-
end transmission performance of terminals and networks, in relation to the perceived quality and acceptance
by users of text, data, speech, and multi-media applications. It is the lead study group on performance and
quality of service. (Though this work includes the transmission implications of all networks and all
telecommunications terminals, and also includes work on performance and resource management, it has a
special focus on IP quality of service, interoperability and implications for NGN.)
T/studygroups/com19/index.asp. Study Group 19 is responsible for studies relating to: network aspects of
mobile telecommunications networks, including International Mobile Telecommunications 2000 (IMT-2000)
and beyond, wireless Internet, convergence of mobile and fixed networks, mobility management, mobile
multimedia functions, internetworking, interoperability and enhancements to existing ITU-T
Recommendations on IMT-2000.
74
See:
ETSI EG 202 057-1 V1.2.1, Speech Processing, Transmission and Quality Aspects (STQ); User related
quality of service parameter definitions and measurements; Part 1: General User related quality of service,
August 2005, at http://pda.etsi.org/pda/queryform.asp.
quality of service parameter definitions and measurements; Part 2: Voice telephony, Group 3 fax and modem
data services, September 2002, at http://pda.etsi.org/pda/queryform.asp.
quality of service parameter definitions and measurements; Part 3: quality of service parameters specific to
Public Land Mobile Networks (PLMN), April 2005, at http://pda.etsi.org/pda/queryform.asp.
quality of service parameter definitions and measurements; Part 4: Internet access, August 2005, at
http://pda.etsi.org/pda/queryform.asp.
75
See ETSI TR 102 276 V1.1.1, User Group; Users' Quality of Service Criteria for Internet Access in
Europe, October 2003, at http://pda.etsi.org/pda/queryform.asp.
76
See:
ETSI TS 102 250-1 V1.1.1, Speech Processing, Transmission and Quality Aspects (STQ); quality of service
aspects for popular services in GSM and 3G networks; Part 1: Identification of Quality of Service aspects,
October 2003, at http://pda.etsi.org/pda/queryform.asp.
aspects for popular services in GSM and 3G networks; Part 2: Definition of Quality of Service parameters
and their computation, March 2006, at http://pda.etsi.org/pda/queryform.asp.
aspects for popular services in GSM and 3G networks; Part 3: Typical procedures for Quality of Service
measurement equipment, September 2005, at http://pda.etsi.org/pda/queryform.asp.
aspects for popular services in GSM and 3G networks; Part 4: Requirements for Quality of Service
measurement equipment, October 2003, at http://pda.etsi.org/pda/queryform.asp.

56

aspects for popular services in GSM and 3G networks; Part 5: Definition of typical measurement profiles,
November 2005, at http://pda.etsi.org/pda/queryform.asp.
aspects for popular services in GSM and 3G networks; Part 6: Post processing and statistical methods,
October 2004, at http://pda.etsi.org/pda/queryform.asp.
77
See ETSI TR 102 274 V1.1.2, Human Factors (HF): Guidelines for real-time person-to-person
communication services, J anuary 2004, at http://pda.etsi.org/pda/queryform.asp.
78
See:
ITU-T Recommendation E.721, Network grade of service parameters and target values for circuit-switched
services in the evolving ISDN, May 1999, at http://www.itu.int/rec/T-REC-E.721.
ITU-T Recommendation E.771, Network grade of service parameters and target values for circuit-switched
public land mobile services, October 1996, at http://www.itu.int/rec/T-REC-E.771.
79
See ITU-T Recommendation G.114, One-way transmission time, May 2003, at http://www.itu.int/rec/T-REC-
G.114.
80
See ITU-T Recommendation G.1010, End-user multimedia quality of service categories, November 2001, at
http://www.itu.int/rec/T-REC-G.1010.
81
See ITU-T Recommendation Y.1541, Network performance objectives for IP-based services, February 2006,
at http://www.itu.int/rec/T-REC-Y.1541.
82
For example, with rounding to 2 significant figures, fault repair times of 1.1 and 1.9 days would be treated as
different and written as 1.1 days and 1.9 days respectively, but fault repair times of 15.1 days and 15.9 days
would be treated as the same and written as 16 days.
83
A scale having few points can exaggerate differences between quality levels unfairly: a regulator that endorsed
operators as quality approved for particular aspects of services would be using a scale having two points, and
if 2.0 days was the upper bound on approved times 1.9 days would be good and 2.1 days would be bad
despite their closeness.
84
Weighted sums of measurements are at least better than average rankings of operators, obtained by averaging
over the rankings of the operators given by the measurements. Such averages treat minor differences between
operators for less important measurements as being as significant as major differences between operators for
more important measurements. They have nonetheless occasionally been devised by other monitors and quoted
by broadband internet operators.
85
See ITU-T Recommendation P.800, Methods for subjective determination of transmission quality, August
1996, at http://www.itu.int/rec/T-REC-P.800.
86
See ITU-T Recommendation P.862, Perceptual evaluation of speech quality (PESQ): An objective method for
end-to-end speech quality assessment of narrow-band telephone networks and speech codecs, February 2001, at
http://www.itu.int/rec/T-REC-P.862.
87
See ITU-T Recommendation G.107, The E-model, a computational model for use in transmission planning,
March 2005, at http://www.itu.int/rec/T-REC-G.107.
88
See ITU-T Recommendation P.562, Analysis and interpretation of INMD voice-service measurements, May
2004, at http://www.itu.int/rec/T-REC-P.562.
89
See ITU-T Recommendation J .144, Objective perceptual video quality measurement techniques for digital
cable television in the presence of a full reference, March 2004, at http://www.itu.int/rec/T-REC-J .144.

57

90
There are various useful discussions of the factors influencing drive around tests. See:
quality of service parameter definitions and measurements; Part 3: quality of service parameters specific to
Public Land Mobile Networks (PLMN), April 2005, at http://pda.etsi.org/pda/queryform.asp.
aspects for popular services in GSM and 3G networks; Part 5: Definition of typical measurement profiles,
November 2005, at http://pda.etsi.org/pda/queryform.asp.
91
For example, 1537 tests, under conventional assumptions about the distributions, are enough to give 95%
confidence that a measured percentage of 99.0% corresponds with an actual value between 98.5% and 99.5%,
a measured percentage of 96.0% corresponds with an actual value between 97.0% and 95.0%, and a measured
time is within 10% of the actual value (if the observed standard deviation is at most twice the observed mean).
92
The term best effort refers to IP networks that make their best efforts to deliver IP packets but do not
guarantee to do so. For the services such as file transfer for which IP was originally devised, these best efforts
are satisfactory, because another protocol (TCP) is placed on top of IP to make delivery reliable. However, TCP
adds delay to communication, so real time services usually depend on a different protocol (UDP) instead of TCP.
As UDP does not make delivery reliable, if too many IP packets are lost real time services become unusable. In
this paper the term suitable quality refers to IP networks that allow assurances of quality for real time and other
services, by ensuring in particular that packet delay and loss are tolerable. Though best effort is a conventional
term, suitable quality is specific to this paper.
93
However, getting to a state where retail price controls might be abolished needs time and effort: in the UK it
took 22 years and entailed separating the management of the retail services and the wholesale services. See
Ofcom, Retail Price Controls Explanatory statement, J uly 2006, at http://www.ofcom.org.uk/consult/condocs/
retail/statement/rpcstatement.pdf.
94
See ITU-T Recommendation G.1000, Communications quality of service: A framework and definitions,
November 2001, at http://www.itu.int/rec/T-REC-G.1000.
95
See ITU-T Recommendation I.350, General aspects of quality of service and network performance in digital
networks, including ISDNs, March 1993, at http://www.itu.int/rec/T-REC-I.350.
96
See ETSI TS 102 250-2 V1.4.1, Speech Processing, Transmission and Quality Aspects (STQ); quality of
service aspects for popular services in GSM and 3G networks; Part 2: Definition of Quality of Service
parameters and their computation, March 2006, at http://pda.etsi.org/pda/queryform.asp.

QOS Bkgpaper

Uploaded by

Copyright:

Available Formats

QOS Bkgpaper

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

QOS Bkgpaper

Uploaded by

Copyright:

Available Formats

INTERNATIONAL TELECOMMUNICATION UNION

TELECOMMUNICATION DEVELOPMENT BUREAU

You might also like