advanced search

Search

11 February 2006| Member Edition

Feature Article

Measuring performance in services

Services are more difficult to measure and monitor than manufacturing processes are but
executives can rein in variance and boost productivity—if they implement rigorous metrics.

Eric Harmon, Scott C. Hensel, and Timothy E. Lukes

2006 Number 1

increasingly demanding customers, high labor costs, and,

Faced with stiffening competition,
in some markets, slowing growth, service businesses around the world are trying to boost
their productivity. But whereas manufacturing businesses can raise it by monitoring and
reducing waste and variance in their relatively homogeneous production and distribution
processes, service businesses find that improving performance is trickier: their customers,
activities, and deals vary too widely. Moreover, services are highly customizable, and
people—the basic unit of productivity in services—bring unpredictable differences in
experience, skills, and motivation to the job.

Such seemingly uncontrollable factors cause many executives to accept a high level of variance—
and a great deal of waste and inefficiency—in service costs. Executives may be hiring more staff
than they need to support the widest degree of variance and also forgoing opportunities to write
and price service contracts more effectively and to deliver services more productively.

As with any task or operation, to improve the productivity of services, you must apply the lessons of
experience. Consequently, measuring and monitoring performance (and its variance) is a
fundamental prerequisite for identifying efficiencies and best practices and for spreading them
throughout the organization. Although some variance in services is inescapable, much of what
executives consider unmanageable can be controlled if companies properly account for differences
in the size and type of customers they serve and in the service agreements they reach with those
customers and then define and collect data uniformly across different service environments. To do
so, it is necessary to bear in mind a few essential principles of service measurement.

 First, service companies need to compare themselves against their own performance rather
than against poorly defined external measures. Using external benchmarks only compounds the
difficulties that service companies face in getting comparable measurements from different parts of
the organization.
 Service companies must look deeper than their financial costs in order to discover and
monitor the root causes of those expenses. This point may seem self-evident, yet many companies
fail to understand these causes fully.
 Finally, service companies must set up broad cost-measurement systems to report and
compare all expenses across the functional silos common to service delivery organizations. The goal
is to improve the service companies' grasp of the cross-functional trade-offs that must be made to
rein in total costs.

None of these principles is easy to implement. Top executives are likely to face resistance from
managers and frontline personnel who insist that services are inherently random and that service
situations are unique. Managers who have grown used to the protection that lax measurement
affords may be reluctant to view their operations through a more powerful lens. But only by
adopting these principles and implementing rigorous measurement systems throughout the
organization can service executives begin to identify reducible variance and take the first steps
toward bringing down costs and improving the pricing and delivery of services.

Why variance is difficult to measure

Executives who launch variance-measurement programs in a service business are often surprised at
the level of difference they discover among similar sites and groups within their own organization,
let alone when they compare one company with another. In general, a company's metrics are not
uniform across its business units, so that, for example, one group in a call center may regard all
calls on a given issue as a single case, while another logs every call separately. A top executive with
a background in consumer goods (where items are similar and thus comparable) assumed control of
a service business and was shocked to find that the variance of key metrics among similar sites
ranged from a factor of 2 to 30. Site managers explained this vast range by asserting that every
site was different—and, according to their metrics, they were right.

Services are different

To make meaningful comparisons, companies have to identify the sources of difference in their
businesses and devise metrics that compare these businesses meaningfully. The considerations that
show up frequently include the most obvious differences among jobs and groups, such as regional
variations in labor costs, local geographies and difficulties in reaching accounts, the workload mix
(for example, repairs versus installation), and differences in the use of capital (whether equipment
is owned or leased by the company or owned by the customer). Several other major issues come
into play as well.

Service-level agreements. The more types of services a business offers, the more variability it can
expect in its agreements. The metrics for a help desk that provides customer support for 5,000
users in a 9-to-5 office are very different from those for a help desk that supports logistics in a
round-the-clock industrial environment. Even when offerings are similar, variance can be introduced
locally through the way contracts are interpreted. In one IT-outsourcing company, two desktop
support accounts with service-level agreements that specified an eight-hour response time had very
different cost metrics. When asked why, the manager of the poorly performing account said that,
despite the contract's limits, "If we don't answer within the hour, our client goes ballistic." The
written service-level agreement had been trumped by an unwritten one that was costing real
money.

Environment, equipment, and infrastructure. Each customer's environment has unique aspects that
are difficult to measure. A logistics provider will see huge differences between managing a big,
automated warehouse and a small, simple one. Field services that support industrial systems must
contend with many generations of equipment and upgrades at customer sites. Some clients have
their own on-site staff to support service, while others may be difficult even to reach. Given the
range of possibilities, it's usually not very helpful simply to measure the average cost of a service
call.

Work volume. Size is a major reason for the wide variance among accounts and business units.
Interestingly, managers of both small and large accounts claim that size makes their particular
metrics worse. Both have a point: large accounts should benefit from scale, but in general they are
also more complex, and that drives costs back up. Volume needs to be considered, but only in
tandem with other patterns (including scale benefits and the breadth of work) that help explain
costs.

The data problem

Underlying all of these problems is an inability to identify what must be measured and how to
normalize data across different environments. Even when companies know what to measure, they
struggle to achieve accuracy. Data are rarely defined or collected uniformly across an organization's
environments. A service call involving the installation of two elevators, for example, could be
measured as a single installation in one part of a company and as two in another.

Contributing to this ambiguity is the fact that data collection is usually driven by the requirements
of financial cost reporting, which often fails to shed light on ways of boosting performance.
Accountants for an IT services company may need to know the cost of each server, for instance, but
an executive looking to reduce variance would also need to know the number of service incidents by
server type and the time spent on each incident. Variance in demand drivers is also important: did
the number of calls to a help desk rise because more users bought a product, for example, or
because it changed? Financial metrics might fail to detect this important distinction.

Principles of service measurement

Many executives don't understand how to measure and manage what appear to be unique activities,
and they confuse correctable performance variance with irreducible environmental variance.
Embracing three principles that identify variance and allow for meaningful comparisons can help
executives overcome these difficulties.

Use internal benchmarks

While a company must know what its peers are achieving, it's a mistake to measure its performance
against the competition: these benchmarks are typically just samples of data with little explanation
behind them. Companies that use external benchmarks are often frustrated to find themselves off
by a factor of five to ten, positively or negatively.

Using external benchmarks compounds the internal difficulties that service companies face in
normalizing activities and the data that define them. Consider a measure such as costs per unit of
information processed: some companies include allocated costs, such as corporate overhead and
salaries; others don't.

Internal benchmarks deliver more detailed metrics, allowing a company to find its own best
practices and to see where and how they are achieved. It can then have access to all relevant
information to assess differences among business units and accounts. In defining internal
benchmarks, for example, a company can determine which costs are included or how asset costs
are allocated—details that get lost in external benchmarking. A company can see what's really
possible within the organization by using its own benchmarks.

A cost tree with detailed metrics is an important tool to help companies define internal benchmarks
(exhibit below). External metrics might deliver numbers on the top level of the tree, but only by
developing internal trees for each service line can a company begin to understand its true cost
drivers. A tree allows a manager to compare the performance of different accounts against similar
metrics and also to calculate which improvements will have the most impact on the top-level figure.
Once a team has gathered cost data throughout the tree, for example, it could target opportunities
to cut costs and calculate which efforts would have the most impact on the bottom line. Creating
cost trees can also help companies write better service agreements that exclude unprofitable
activities or generate more revenue where service costs warrant it.
Measure cost drivers

Even after service companies begin to define and capture the detail that lies beneath the top level
of the cost tree, they still need to discover the underlying cause of each expense. Measuring only
the cost of repair calls, for instance, probably wouldn't reveal whether they all stem from a single
poorly built product, which could be improved or sourced differently for less than the cost of the
repairs, or from factors such as variability in the performance of repair teams. Better measurements
look at cost drivers, such as cost per employee (a resource metric), incidents per employee per day
(a productivity metric), or—in a product-based service business—the number of incidents per
product (a volume metric).

Of course, companies must also omit allocated costs, which can confuse the issue. A business unit's
support infrastructure, for example, could include human resources, physical plant, and product
engineering, all of which must be considered from a financial point of view. But such costs do little
to determine productivity and are something of an obstruction when companies try to spot variance
and waste. Once these obstacles are removed, managers can stop trying to cut costs that may be
beyond their control and instead address the drivers they can improve. Before measuring the
financial costs, it's often helpful to measure the items and events that drive costs, such as people,
machines, incidents, service calls, and change orders.

Measure deep and broad

When service companies try to measure only their selected costs—rather than taking a
comprehensive approach—they are often surprised to see that their teams hit every budget target
while still losing money. That's because services are fungible, and it's easy to measure the wrong
things or to shift costs, intentionally or not, to unmeasured areas.
Consider the case of a cable company that was trying to reduce the resolution times of its help desk
and service calls. After setting goals, managers saw resolution times shrink, but total service costs
were rising. In this case, help desk representatives, eager to meet their goals, spent less time
trying to resolve problems remotely. After asking only a few questions, these employees referred
cases to field service reps, who were happy to have a series of fast and easy calls to boost their own
metrics. Unfortunately, the number of field service calls, which are far more expensive than help
desk calls, rose dramatically. To resolve this problem, management combined call centers and field
services into a single cost tree and monitored the percentage of calls passed from the one to the
other, as well as the time spent on each type of call. Managers then encouraged the call center reps
to spend more time trying to resolve difficult calls before passing them along to field services,
thereby increasing the average call time but helping to reduce total costs. Thus a critical purpose of
any cost tree is to yield insights about how better (or worse) performance in one area of the tree
might affect another.

Setting up measurement systems

With these principles in mind, executives can begin to define their metrics, collect data, and
implement processes that will drive their efforts forward.

Build the tree and choose your metrics

Cost trees should be detailed enough to spot efficiency problems and broad enough to be
comparable across operating units. Once companies have identified the allocated costs and cost
drivers, they can begin to build the cost tree. Broad input from the field (line managers, engineers,
field and service reps) is vital, along with input from senior executives, who are generally better
able to focus on the total costs required to deliver a service to customers. In this way, the tree
captures all the costs of (and details on) the most important cost drivers. The tree should also be
constructed to compare key metrics across a range of environments—for example, all call centers,
whether they operate 24 hours a day or 9.

As the data arrive, management will want to monitor the top level of the tree as well as the key
metrics below. In most cases, we find, three to five metrics monitor 80 percent of the variance in
costs.

Collect with care

Without clearly defined metrics and knowledgeable people to support the gathering of data
throughout the organization, companies can spend too much time cleaning up messy data. Training
and improved processes can alleviate this problem.

Managers should review the data collection rules and templates with the people who develop them—
usually employees from different regions or accounts. Even with new procedures in place, however,
there will be much room for interpretation. It's therefore helpful to show not only how data should
be collected and entered but also how users occasionally misinterpreted these processes in the past
—an approach that sheds light on gray areas the rules might not address. Guidelines for identifying
problems early on can save time later. It's also important to establish boundaries beyond which
suspect metrics should be investigated. One service company, whose teams handled from two to
five service calls a day, wondered why one of its teams was reporting an average of only a single
call. It found a good reason: the account belonged to a prison system, where security procedures
made each visit a daylong affair.

Reviewing data collection in the early stages of implementation can help to ensure that procedures
are followed. Equally, sharing reports with regional and account leaders gives them an early view of
their standing and can help identify unusual patterns in the data.
Institutionalize measurement

Managers accustomed to tracking costs in accordance with accounting needs will have to
understand these new metrics and make them consistent throughout all levels of the organization.
Periodic reviews, whose frequency should be based on the availability and shelf life of data, are
essential for individuals and work groups. Visible interest from senior management—such as
sending an executive vice president to attend a regional metrics review—promotes a strong
message to everyone that a company is intent on identifying variance and improving service
performance. Compensation should be tied to these metrics.

What's measured can be managed

For tools managers can use to price and manage service contracts, see "How to make after-sales
services pay off."

Once executives have learned to measure the variance inherent in service companies, they can
begin to manage processes to eliminate waste, to improve the delivery of services, to price services
more accurately, and to write better contracts. Although a company can do many things to control
the variance of its service delivery, most of them fall into three main areas: managing demand,
standardizing environments, and applying appropriate resources to tasks.

Managing demand offers the biggest potential for improvement. Cost trees help managers identify
the sources of demand for services—sources that might include faulty products, poorly performing
service units, or any number of other causes. Some fixes must be made within the organization
(better training, better products, automated-response systems); others depend on shaping the
behavior of customers (for instance, by offering tools and guidance to help them resolve problems
themselves).

Standardizing operating environments requires the most discipline, since salespeople are strongly
tempted to sell as much customization as a client wants. Standardization can yield enormous
results: in addition to raising productivity, it helps the workforce become more flexible because
people can transfer with less retraining. Where possible, companies should standardize not only
service product lines and tasks but also the work environments of employees and the equipment
they use to deliver services. Scripted routines help eliminate errors and allow employees to emulate
high performers. Furthermore, clearly defined programs limit overdelivery, a common problem in
service companies.

What's more, identifying cost variances can help companies allocate their human resources more
effectively. In general, it's more productive to handle problems with the least expensive resources
that can resolve them: calling in experts or sending out field technicians increases costs and slows
response times—and therefore makes customers less satisfied. Metrics on costs per call or device
demonstrate the benefits of using less expensive labor, thus encouraging companies to keep
requests upstream and to place first responders (often a call center) in less costly regions to further
increase savings and productivity.

Finally, companies that have a better picture of where costs are incurred can price services more
accurately to avoid losing revenue on unprofitable activities. They can write better contracts that
take into account cost drivers hitherto written off as inescapable variance.

As services become an ever larger part of the global economy, managers are rightly looking for
ways to improve productivity and efficiency. Services may be more difficult to measure and
standardize than the manufacture of products, but executives should not abandon hope. Adopting
the principles set forth in this article will help companies improve the delivery, pricing, and sales
and marketing of services.
About the Authors

Eric Harmon is an associate principal in McKinsey's Dallas office, Scott Hensel is an associate
principal in the Stamford office, and Tim Lukes is a consultant in the Miami office.

The authors would like to thank Byron Auguste, Ken Davis, Travis Fagan, Ozan Gursel, and Greg
Neubecker for their contributions to this article.