Customer Analytics in Telecom

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 47
At a glance
Powered by AI
The key takeaways are around customer lifecycle management (CLM) in the telecom industry including segmentation, offer design, execution and evaluation of test campaigns.

The different phases in a typical telecom customer lifecycle are: 1) New joiner phase 2) Stable phase 3) Churn phase

Some of the key parameters used to measure the performance of predictive models are hit rate, penetration and lift.

Customer & Marketing Analytics:

TELECOM

Agenda
Need of Segmented Offers
CLM test campaign Segmentation approach
CLM test campaign Offer design
CLM test campaign Offer Execution
CLM test campaign Communication Execution
CLM test campaign Evaluation
Process Flow

Agenda
Need of Segmented Offers
CLM test campaign Segmentation approach
CLM test campaign Offer design
CLM test campaign Offer Execution
CLM test campaign Communication Execution
CLM test campaign Evaluation
Process Flow

Telecom: A snapshot

Operators

Subs figures (in


million)*

Bharti Airtel

188.2

Vodafone

152.35

Reliance

122.97

Idea

121.61

BSNL

101.21

Tata

66.42

Aircel

60.07

Teledensity in India stands at 74.02%


Monthly 5.5 million subs are added

*Source: trai.gov.in
QE Mar13

WHERE DOES

CUSTOMER
&

MARKETING
ANALYTICS
FIT IN
5

Objectives and
Deliverables
To target increase in revenue and contribution through targeted and
segmented programmes, reduce churn and manage CLM

To engage customers by offering segmented products and retailers


by providing earning opportunities

To create processes to gain insights and drive performance & profits


across products/segments

To convert this expertise into competitive edge and improve overall


market share (RMS & CMS)

How is it different from traditional Usage &


Retention programs?
1

Dont guess,
Test !

Test several offers against each segment to find positive Return on


Investment (ROI).. Let customer decide the best offer..!

Segments of 50100K

Micro segments (50-100K) based on usage through Intensive data


mining.. Different offers to different micro-segments prevent
value destruction ..!

3 Scale & Speed

Customised
campaigns

Scale tested offers to millions of subs through multiple parallel broadlaunches and analyze results on real time basis.. Build a library..!

Target customers at all stage of their lifecycle.. Upsell, crosssell,


retain and train through creative offerings ..!!

Use ROI w.r.t. control group to measure real


contribution
7

The objective of Prepaid CLM campaigns is different based on the lifecycle


stage of each customer
Typical lifecycle of a telecom customer

ARPU

Phase 1: New joiner phase


Objective
Reduce high churn
Phase 2: Stable phase
Increase ARPU
ARPU stimulation
Reduce downgrade

Focus Area
for Static
programs

Focus Area
for Dynamic/
ZU Programs

Phase 3: Churn phase


Objective
Prevent churn
Stimulate upgrade

Time (AoN)

Agenda
Need of Segmented Offers
CLM test campaign Segmentation approach
CLM test campaign Offer design
CLM test campaign Offer Execution
CLM test campaign Communication Execution
CLM test campaign Evaluation
Process Flow

Basic Segmentation
Approach

VLR days: Visitor Location Register (No of days a subscriber is


latched to a network in a month)
MoU: Minutes of usage in a moth
10

Basis of Segmentation
Framework

Broad segmentation basis MOU, VLR and AON


Further segmentation basis current stage on CLM

11

Micro Segmentation
Approach
The strategic segmentation is used to categorize the customers based
on main usage and behaviour categories

The micro segmentation is used to identify specific behaviours and


usage patterns to make campaigns targeting specific value levers
Current focus

Primary dimension
MOU/VLR AON

Primary usage

xx

xx

Data/ text
users

>3
mths.

Sub-usage
xx

Mr. Local onnet

xx

Mr. Local offnet

xx

Local balanced

xx

Mr. STD

xx

Mr. VAS

xx

Mr. Balanced

Retention

Type (SMS, Voice, VAS)


Time (Weekend, Night))
Value
levers

Usage
stimulation

Volume (MoU, Count)


Duration/Frequency

All
subs

Super

xx

xx

<3
mths

Data/ text
non users

xx

Mr. Local onnet

xx

Mr. Local offnet

xx

Local balanced

xx

Mr. STD

xx

Mr. VAS

xx

Mr. Balanced

Outgoing/Incoming

Share of
Wallet

12

Every
micro
segment
has
different

NOT EXHAUSTIVE

Microsegmentation

Secondary variables are chosen depending upon


campaign intent

Local on-net
boost

2 Local boost

3 STD boost

4 SMS boost

5 Night MoU

6 Data

7 VLR days

Mandatory

Optional

Local I2I OG MoU

Local I2I OG MoU /Total local OG MoU


Local on-net OGRR

Local OG MoU

Local OG MoU /Total OG MoU


Local OGRR

STD OG MoU

OG SMS count

STD OG MoU /Total OG MoU


STD I2I OG MoU
STD OG RR
SMS revenue/Airtime revenue ratio

Night OG MoU

Night OG MoU /Total OG MoU


Night I2I OG MoU

Current data spend or


Current data usage (No of KB)

Data revenue/Airtime revenue ratio

Current months VLR days

VLR days of last month


Recharge /Usage behaviour current vs last
month

The core of Prepaid CLM is the test campaign


methodology
What and why?
What is a test?
A test is conducted on a sample
to determine effectiveness of
various campaign offers
Why test?
Testing enables us to use targeted
campaign offers to
a) increase revenue
Exploit elasticity of microsegments
Use repricing
b) Reduce churn
Increase saves

How?

Analysis and data mining


Offer design
Test to select target and control
ROI results

Segment 1
Segment 2
Segment 3
Segment 4
Offer
A

Offer
B

Offer Control
C

Every CLM test campaign has four core components


A combination of necessary test components

Intent
1

Segments

Offers
2

Offer
execution
3

Offer
evaluation

results in a campaign grid used for


evaluation

What do we want to address

through the campaign? E.g.,


Local onnet boost, STD boost or
SMS boost
Who are we targeting? E.g.,
segmentation by ARPU,
SMS/MOU, time of zero balance

What offers do we use:


Top-up bonus
Bundle offers
Tariff discounts
How do we execute the offer?
Target and control group
Hold outs and multiple offers
Imagine configuration etc
How do we execute

communication?
Mode, language, time of day
How
should we evaluate the
etc.
campaigns?

Target 1
Target 2
Target 3
Target 4
Offer A Offer B Offer C Control

Execution details:
Offer grid identifies successful
offer for each target through
testing
Offer is then executed as a
broad launch

Agenda
Need of Segmented Offers
CLM test campaign Segmentation approach
CLM test campaign Offer design
CLM test campaign Offer Execution
CLM test campaign Communication Execution
CLM test campaign Evaluation
Process Flow

Offer design

Offer design is a 3 step process


Activities
1
Decide on
product
concepts
2
Decide
price
points
3
Decide
script
variants

Based on the chosen intent, identify 3-4


relevant product concepts (e.g. bulk minute
bundles, rate cutters, rate cutters with talktime
etc)

Do a detailed profiling of the micro-segments


around spend, usage, recharge pattern
Based on profiling design 3-4 price points for
each product concept at varying levels of
discount and varying levels of stretch

For each price point, design 1-2 variants in the


communication script where applicable (e.g.
communicate tariff or communicate discount)

Offer Design

1 Intent decides the product being offered to the customer


Objective for the various bands of usage
Cross-sell

Up-sell

Do Not Touch

Number of subscribers1

Adoption

Low

Medium

High

Usage on the selected secondary usage variable


1 Assuming a near normal distribution of the subscriber base

Offer design

2 Deciding the price points for testing is a function of 3 key


factors stretch /ticket size, discount /benefits and validity

Guiding
principles

Stretch
should be
sufficiently
aggressive
The
discount
offered
should not
be greater
than the
stretch

Key driving factors

Supporting analysis

Affordability

Recharge frequency

Ticket size

Tariff discount

RR (on specific leg)

Benefit

Validity/
eligibility

Available price points

Minutes

Modal recharge value

Usage (leg wise)


Usage per decrement day (leg
wise)

Talktime

ARPU

Duration of benefit

Decrement days (on specific leg)

No. of times
customers can take
the offer

Offer profitability

ILLUSTRATIVE

WHICH PRICEPOINT IS BETTER?? MRP 14 OR MRP 19

Offer design

3 Script variants can be designed by modifying the Hook and


Value elements

Offer

Hook

RC 11: STD
calls at 30p/
min for 15 days

STD rates slashed

FTT on RC
200

Full talktime

Recharge for Rs. 200 and


get full talktime

Free talktime

Rs. 20 free talktime


credited to your account.
Recharge for Rs. 200 to
start using free talktime

Value
Recharge with Rs11
and call STD at 30p/min
for 15 days

Upto 50% discount on


STD rates

Agenda
Need of Segmented Offers
CLM test campaign Segmentation approach
CLM test campaign Offer design
CLM test campaign Offer Execution
CLM test campaign Communication Execution
CLM test campaign Evaluation
Process Flow

Offer execution

CLM offer execution comprise of the following elements


Element

What

Why

Target control
1
philosophy

Comparison of target group of a test


campaign with a control group not
exposed to the campaign

Ensure exact evaluation of effect


of a campaign

2 Sampling

Number of subs in target and


control group and approach to
generate the list

Statistical rigour to ensure the test


is statistically significant and
applicable for entire population

Hold out
3
principles

Exclusion of target/ control group of


a test from other campaigns for a
defined duration of time

Ensure exact evaluation of effect


of a campaign over time and on a
sustainable basis

4 Exclusion list

Exclusion of subs from a campaign


due to different reasons DnD,
VLR (only for test)

Ensure subs who were not


contacted do not result in misinterpretation of campaign results

Period when offer is run and


communications are sent. The offer
period is defined upfront. Analysis period
is the duration over which target and
control groups are studied & compared
to evaluate campaign

To ensure subs are given


sufficient time to respond after
receiving a communication
Ensure exact evaluation of effect
of a campaign over time and on a
sustainable basis

Duration of
5
test/analysis

Offer execution

Net take rates determined by target and control take rates

Control take rate


Defined as the share of control
group subscribers who would have
qualified for the offer or availed of
the offer (even though they were
not exposed to the offer)

Relationship between target, control and


net take rates

Net take rate


Defined as the share of target group
takers who under normal
circumstances would not have
availed of the offer

Target (gross) take rate


Defined as the share of target group
takers who availed of the offer

7/23/15

7/23/15

7/23/15

Agenda
Need of Segmented Offers
CLM test campaign Segmentation approach
CLM test campaign Offer design
CLM test campaign Offer Execution
CLM test campaign Communication Execution
CLM test campaign Evaluation
Process Flow

Send :
Whom should we Scrub from the base?

DND
scrubbing

VLR
scrubbing

Hold-out
scrubbing

Guidelines

Rationale

DND Scrubbing to be done for all

Compliance to Regulatory Guidelines &


not to disturb DND customers.

Push Modes used for


Communication like SMS and
OBD.

Scrubbing will be based on VLR

Required to optimise SMS/OBD

latched subscribers
Campaign manager to decide on
need for VLR scrubbing.

Scrubbing of Target and Control

base(Holdout Base) from Circle


Segmented Offers Base during
the Holdout Period (currently 15
Days)

capacity
For high VLR segments, previous days
data may be obsolete, and they may be
back on VLR today, so no VLR scrub

For Revenue Calculation.

Receive:
Preferred mode and time maintaining discipline

Element

Mode of
1
communication

Time of
2
communication

Frequency of
3
communication

What

Why

Choice between SMS, OBD, IVR


etc. to communicate the offer.
Choice of mode over a day and
over the offer period

Ensure target group is contacted


in a mode of their preference

Choice of the exact time to send


the communication at. Choice of
any variation in time during the
campaign

Ensure the subs are contacted at


a time with maximum possibility of
action

Choice of number of times a user is


contacted in a day and over the
offer duration

To strike the right balance


between reminders/ recall and
communication spam and
dissatisfaction

Receive:
Mode of communication?
Modes of communication used?

SMS
SIVR
OBD
Push SMS from Call Center.
Push SMS from SIVR
USSD
Insta alerts- SMS & OBD
Recharge confirmations- Retailer & customer

What validity and retry policies are used?

Receive:
Mode of communication selection
What mode of communication is to be used?

Customers mode preference to be updated based on proxies Customers with outgoing SMS usage >=10 in last 30 days will
be SMS profiled .
More than 3 successful OBD in last 3 months.The customers
with >3 OBDs heard for >= 10 secs or >= 50% of the length of
the OBD.
Default customer mode preference to be SMS

What validity and retry policies are used?

OBD retries: 2 retries to be attempted for all unsuccessful OBD


calls.
SMS retries: 1 Retry (automatically happens in BMG).

Receive:
Time of communication?
When should the communication be sent?

Time of day
For all communication
Recharge based offers: Communication to be sent 1 hour before
peak recharge window for corresponding MoU/VLR segment
Use & get offers: Communication to be sent at start of day
success of communication: OBD success & SIVR hits

Event based triggers


Recharge
Latch on to network (Insta alert) for Zero Segment (Customers not on VLR for >=30 days)
ZU >=4 Days &
Lower ARPU bands of Regular)

Receive:
Frequency of communication?
How much to
communicate?

Illustration

Communication not to exceed 2 SMS (U&R, VAS) + 1 OBD push per


unique sub per day to VLR latched subscribers
All Communication to go between 9 AM to 9 PM as per TRAI Guidelines

Day 1
9:00 AM
SMS

Day 1
6:00 PM
OBD
Local Idea Call Rates
Slashed To 20P/Min For
30 Days. Recharge With
Rs.26 And Enjoy Cheaper
Idea To Idea Calls. Hurry!!
Limited Period Offer
Specially For You.

Incoming call
+91 95432 12345

Understanding Language:
Language of communication
What language is to be used to communicate?
All communications to customers (text and voice) will be in the customers preferred language
Customers language preference to be updated based on proxies
Static IVR language preference
Call centre IVR language preference
CRBT language
Default customer language preference to be the regional language
Vernacular script to be used for compatible handsets only (IMEI based check)
Is stated preference
known?

Is English the selected


language?
Yes

Yes

No

Use stated
preference

Is handset vernacular
script compatible?

Are proxies
available?
Yes
Use language
deduced from proxies

No

No
No

Yes

Use default language


English script used

Vernacular script used

Understand offer:
Scripts should hold customers attention & have clear
call to action
Learning

Example

Must capture attention in the:


First 10 seconds of voice communication
First line of text communication
Linked to campaign type, not to specific microsegments

Free Mins!
STD rates slashed

Clarity
Simple explanation of offers
Perceived value
Rate cutters, bulk minutes:
Quantum of benefit stated
Use & Get:
State incremental usage required
Use of personalised messages e.g., Especially for
you
Trust
Do not use caveats e.g. Avoid T&C apply

500 minutes for STD


calling
Now make STD calls
at only 30p/min

Explicitly stated call to action


Ease of availing offer e.g. through balance deduction
Creation of sense of urgency
Reminders on expiry date

Hook

Value
proposition

Call to action

Recharge for Rs.26


today!
Offer valid only for 2
days more!

Understand Offer:
Script variants can be designed by modifying the Hook and
Value proposition elements
Offer

RV11: STD calls at


30p/min for 15 days

Full talk time on RC


200

Value

Hook

ILLUSTRATIVE

Call to action

Recharge with Rs11


and get STD Calls at
30p/min for 15 Days!

Recharge Today!

Get STD calls at only


30p/min for 15 Days
on Recharge of Rs11.

Offer till 30 Mar only!

Full talktime!

Recharge with Rs200


and get full talktime

Last 2 days Left!

Free talktime!

Rs. 20 free talktime


credited to your
account. Recharge
for Rs. 200 to start
using free talktime

Hurry! Recharge
Now.

STD rates slashed!


Upto 50% discount on
STD rates!

Rs. 20 free!

Understand Offer:
How crucial is Way of Communicating?
Sample SMS Scripts used for
testing
Script 1

Script 2

Eng

Complete value for money!Get


FULL Talktime of Rs22 valid for
20 Days on Recharge of
Rs22.Recharge today!Limited
Period Offer Specially For
U.Call tollfree 53533

Full TT Only for U Recharge = 22 ; Talktime = 22


; Validity = 20 Days. Recharge
Today! For Details,dial tollfree
*147*99#

Hin

Har Recharge Poora


Recharge!!Recharge Kare
Rs35 Se Aur Paaye Poore
Rs35 Ka Talktime.Validity 30
Din.Seemit Samay Offer Khaas
Aapke Liye.Call Tollfree 53533

Full TT Sirf Apke Liye Recharge = 35 ; Talktime


= 35 ; Validity = 30 Din.
Aaj hi Recharge Kare!
Adhik Jankari ke liye,dial
*147*99#

Agenda
Need of Segmented Offers
CLM test campaign Segmentation approach
CLM test campaign Offer design
CLM test campaign Offer Execution
CLM test campaign Communication Execution
CLM test campaign Evaluation
Process Flow

Campaign evaluation

Campaign evaluation using ROI helps us to understand which


of the different offer is performing better
All I2I local calls @ 30
paisa per minute for a
recharge of Rs 11 is
totally the best offer!

Which offer is better? Campaign evaluation


using ROI helps answer by considering
different factors:
How many customers are you going to get to top
up who would not have topped up otherwise?
How many customers are you going to be giving
the offer to who would have topped up
otherwise?
What will the customers who take your offer do
after they take the offer? Do they spend more on
the network?
What happens to the customers who dont take
the offer? Do they top up on their own later?

No way! Recharge for


` 11 and get 30 local I2I
minutes is way better!

Key Indicators used for campaign evaluation


A

Revenue Uplift
Net Take Rate
Gross Contribution
Net Contribution
Net Contribution Margin
Benefit to Cost Ratio

Revenue Uplift: 1%
Net Contribution:
100
ROI: 25%

Revenue Uplift: 2%
Net Contribution:
200
ROI: 35%

Revenue Uplift: -2%


Net Contribution:
-100
ROI: -

Revenue Uplift: -5%


Net Contribution:
-300
ROI: -25%

Revenue Uplift: 1%
Net Contribution:
100
ROI: 25%

Revenue Uplift:5%
Net Contribution:
1000
ROI: 25%

Revenue Uplift: -7%


Net Contribution:
-600
ROI: -

Revenue Uplift: 4%
Net Contribution:
500
ROI: 20%

Revenue Uplift: 1%
Net Contribution:
100
ROI: 25%

Campaign evaluation

The total impact (ROI) from a campaign is determined by 5


factors

Outgoing
usage benefit
Loyalty
benefit

Total impact

Inbound
usage benefit

Offer cost

Execution
costs

What is it?

How do we calculate it?

Incremental contribution

Compare usage growth Vs

Monetized value of the

Compare expected

Revenue earned from

Compare inbound usage

Cost of making the offer to

Calculated based on offer

Direct costs of executing

Execution cost multiplied

because of increased
outgoing usage

improvement in TTC
distribution at the end of
the campaign

inbound calls

all the takers of the


campaign

the campaign (e.g., cost of


outbound calls, IT
changes)

control during campaign


period

lifetimes in target and


control, based on TTC
distribution

growth Vs control group


during the campaign period

type

by the number of targeted


customers

Campaign evaluation

Post launch campaign evaluation provides a wealth of campaign data

Campaign insights and implications


Take rate: How many people took the offer?
Realized rates: How did the realised rate change after the offer?
Usage uplift: Did the customer who took the offer increase their MoU usage?
Revenue uplift: Did the customers who took the offer spend more on the network?
Deep dive on specific offers: Which offers yield the best net revenue stimulation?
Are there new insights that drive revenue?
Reprice: How did usage patterns change with the offer?
Bro-Test-Launch: Is there sufficient learning to proceed to broad launch?

Agenda
Need of Segmented Offers
CLM test campaign Segmentation approach
CLM test campaign Offer design
CLM test campaign Offer Execution
CLM test campaign Communication Execution
CLM test campaign Evaluation
Process Flow

End to end process for campaign management

BROAD LAUNCH

Entire customer base

TEST
Macro Segmentation

Product 1
Segment 1

Micro Segmentation

Product 2
Product 3

Segment 2

TEST
Revisit Segmentation

Segment 3

Script 1
Script 2
Script 3

ROI Analysis

Agenda
Need of Segmented Offers
CLM test campaign Segmentation approach
CLM test campaign Offer design
CLM test campaign Offer Execution
CLM test campaign Communication Execution
CLM test campaign Evaluation
Process Flow
DIU and other Analytics Models

CONCEPT: Dynamic Trigger


1st May
to
7th May

8th May
to
14th May

15th May
to
21st May

22nd May
to
28th May

29th May
to
4th June

5th June
to
11th June

12th June
to
18th June

19th June
to
25th June

W1

W2

W3

W4

W5

W6

W7

W8

W1

W2

W3

W4

W5

W6

W7

W8

W1

W2

W3

W4

W5

W6

W7

Base
Month

Trigger
Week

26th June
to
2nd July

3rd July
to
9th July

W8

Predicted
Period

Dynamic Trigger Algorithm ensures that new set of potential droppers


are dynamically identified based on their latest usage pattern
47

CONCEPT: Performance Measures


Actual Drop

Predicted
Drop

No

Key parameters

Yes

No

Yes

Hit rate

= (B) / (B+D)

Penetration

= (B) / (B+C)

Lift

25
Random Target

20

Modelled Target

Assume Base showing >50% Drop in M2 = 20%


Thus Prediction Accuracy achieved by Random Base
Selection = 20%

15
10

If Model Hit Rate = 60%

Then, Lift = 3x

0
0

D1 D2 D3 D4 D5 D6 D7 D8 D9 D10

(Hit Rate) / (Prediction Accuracy achieved through


Random Selection)

48

ANALYTICAL MODEL : Evolution of DIU Models


Linear Business Rule:

Single Variable mid-month / 7-day profiling

Decision Tree (CHAID)

Weekly KPIs considered for Model Development

Captures interplay of variables

Logistic Regression

Daily KPIs with statistical transformations

Subscriber level Churn Scores to handle seasonality

Provides better performance compared to Linear Model

49

ANALYTICAL MODEL : Dimensions Considered


DAILY LEVEL
USAGE DIMENSION

TRANSFORMATION

USABLE DATA
RAT_REV_F7M3_M2

RCHG_CNT_PAPER

SUM

PROP_HOM_REV_L7

CUST_REV

COUNT

RAT_REV_M1_M2

TOTAL_MOU

PROPORTION

MAX_REV_L7

DECREMENT_DAY_FLAG

AVERAGE

DS_RCH_L37

SLOPE

SLP_MBL_L7

RATIO

CNT_OG_ANY_L7

RCHG_CNT_V_TOP_UP

MAIN_BAL_INR
TOT_OG_HOME_REV
TOT_OG_ROAMING_REV

PROP_IC_MOU_ALL_L7

AA Mart
(360 variables)

DIU relevant
(93 variables)

ABT
(~460
variables)
50

ANALYTICAL MODEL : Featured Variables


Variables

Description

Impact

Ratio of extrapolated ARPU in last 7 Days


compared to M2 ARPU

Negative

Standard Deviation of customer revenue of 12


groups with mean of 3 days each for last 36 days

Positive

Maximum of the MAIN_BAL_INR in the last 4 days

Negative

CNT_DEC_INS_M1

Number decrement days in M1

Negative

SUM_OG_MOU_L7

Total outgoing minutes of usage in last 7 days

Negative

SLP_DEC_L7

Rate of change of decrement in last 7 Days

Negative

DS_RCH_L37

Days since last recharge in the past 37 days

Positive

DS_USG_L37

Days since last usage in the past 37 days

Positive

RAT_REV_F7M3_M2

SD_REV_3DB_L36
MAX_MBL_L4

51

You might also like