Modelling Customer Behaviour in a Competitive Environment
Ahmed Mohamed & Frederic Lagacherie
BT Laboratories, Martlesham Heath, Ipswich, IP5 3RE
Tel: +44 1473 647 663. Fax: +44 1473 649 421
e-mail: ahmed.mohamed @ bt.com
Abstract
As telecoms markets around the world become increasingly competitive, customers are exercising their
right to move freely between operators in order to secure the best possible deals. Consequently, operators
are having to work harder to retain their market share.
This paper tries to understand the drivers behind customer behaviour in such a competitive market place,
through the use of a Systems Dynamics model. This model investigates the relationship between customers,
an incumbent operator and the largest national cable operator. The drivers behind a customer joining an
operator (take-up) and a customer leaving an operator (churn) are explored. The model also investigates
the impact of Quality of Service (QoS) and pricing strategies on customer behaviour.
A Comparative Performance Indicator Chart has been created for the incumbent and for the cable
company. This chart is used as a benchmark for their QoS and to create a QoS profile for each operator.
Both operators are assessed using 6 separate criteria: Bill Accuracy, Customer Complaints, Reported
‘aults, Faults Rectified/Time, Orders provided and Customer Satisfaction. Attractiveness for each operator
is also calculated using the QoS profile and different pricing strategies.
All these factors are used to try to understand what makes a customer choose one service over another
similar service. This paper does not try to give a definitive reason why a specific customer would switch
operator, rather it shows the trends in customer behaviour in migrating from one operator to another.
Introduction
Increased competition has forced firms to become more competitive in a rapidly changing technological
and regulatory environment. This industry "free for all" will do much to expand the potential range of
services a telecommunications company can offer, this may introduce additional competition into a
company's markets whilst allowing the same company to expand into other markets. It is therefore crucial
for such companies to understand current and prospective customers.
Telcos are constantly improving their service offerings, with new services and ideas to capture the
customers’ imagination. Although their products may be leading the market in terms of innovation, often it
seems that the customer is dissatisfied with some aspects of the service and therefore, decides to churn to a
competitor.
Trying to predict customer behaviour has always been a difficulty for telcos, many of whom have a re-
active, rather than a pro-active strategy to customer churn. Trying to understanding how customers think,
is not the practical answer to solving these problems, rather improving the conditions under which a service
or product is nurtured is one way of minimising churn and maximising customer retention.
This paper describes the methods used to develop a model that focuses on the residential telephony market
and simulates factors leading to the migration of customers from one operator to another.
The model assumes that there are two operators in this market:
© Telco The market leader
© Cable The largest Cable operator
The simulation runs for a period of 20 years. In that time customers may move freely between operators.
Each simulation run depends upon inputs from a customer profile.
The operators start with a percentage of the market depending on their position in the market place. They
can attract new customers by changing the quality of service (QoS), attractiveness, tarrifing and service
availability. All the variables are linked to a Comparable Performance Indicator (CPI), which calculates
the operators perceived QoS, from the customers point of view.
The purpose of the model is to assess which factors influence churn and which attributes of customer
behaviour have a positive effect, i.e. reduce churn and those which increase churn.
The Customer
For the purpose of this simulation it was decided that it would be beneficial if we could create individual
customer profiles. It is also possible to create group profiles by segmenting customers according to wealth.
Slider bars are used to set up the customer profile. The default profile is that of a middle income individual
who is price sensitive, that is to say a 20% or more increase in the cost of his/her service package will
increase their propensity to churn.
Figure 1 shows a typical customer profile ready for a scenario run.
Figure 2 shows the customer profile input screen.
Customer Profile
istomer ensitive QoS. oss_Churn
action
Figure 1: Graph of a Generated Customer Profile
Customer Profiling
Customer Awareness Price_Sensitive
q a i >
00 02 04 06 08 1.0 G@)
QoS_sensitive
Figure 2: Inputs to Customer Profile
Customer Profile Input Screen
1.
Customer awareness. Used to set the level of awareness the customer has about the operator or any of
the operators’ products. A high score here is a good indication that the customer is willing to stay loyal
to the operator.
Customer satisfaction. A high score here indicates that the customer will be more likely to churn if the
operators’ performance indicator falls below this threshold.
Attractiveness. This slider bar shares links with price sensitive and QoS sensitive sliders (4 & 5). The
settings on sliders 4 & 5 effectively sets how sticky the customer is likely to be to a operator who has
high QoS. The slider can then be adjusted to fine-tune the attractiveness (stickiness).
Price sensitive. This determines how much of a price change will affect the customers' decision to
churn. A high score is here indicates a customer is more likely to switch operators.
QoS sensitive. This works in the same way as Price sensitivity, again a high score here could result in
customer churn.
Once the profile has been set, a scenario run can take place. The model takes the customer profile and
feeds it into the Comparable Performance Indicator (CPI).
Comparable Performance Indicator (CPI)
Increasingly, households in the UK can now choose the operator from which they can buy
telecommunications services. This r s that customers need to be able to compare the performance of the
different telephone (or telecommunications) companies offering services in their area, so as to check that
they are
receiving the best deal available.
To help customers make their choice, the telecommunications companies and several consumer
organisations, with support from OFTEL, the industry regulator, have developed a set of comparable
performance indicators for a range of telecommunications services.
The performance indicators currently used are described briefly in the following table:
Indicator What is Measured
Service Provision Operators ability to provide services.
Customer-Reported Faults The reliability of the company’s network
Fault Repairs The ability to repair faults within target times
Complaint Handling How promptly complaints are dealt with
Billing The accuracy of billing information
Customer Satisfaction The perceived QoS by the customer
Table 1. Description of CPI’s
CPI Profiles
Figures 3 & 4, below, show a run generated by calculating the CPI for both the operators.
SS,
E=—_ > 2 =
80}
="
3
—;—Cablet_Bill_ Accuracy
oo 3 _y— Cable1_Customer_Complaints_Resolved
4
—g~ Cable1_Mean_time_Reported_Faults
ae = 47 Cablet_Reported_faults
40) s _,— Cable1_orders_provided
e—=——3" 5
‘4g Cablet_Customer_Satisfaction
—
4
204 ——
F
1
1,990 1,995 2,000 2,005
Time
Figure 3. Cable CPT
100. :
2 3
7 = "_L
—— ——— —
5 6
» i __ om
6
|e _Telco_Bill_Accuracy
60. E _g- Telco_Customer_Complaints_Resolved
—g~ Telco_Mean_time_Reported_Faults
_,- Teleo_Reported_faults
40 4
4 _g— Teleo_orders_provided
er 5
——— : _, Telco_Customer_Satisfaction
-——__,
os —
———
4 1
1990 1995 2000 2005
Figure 4. Telco CPI
Careful examination of figures 3 &4, shows that Telco has a superior CPI rating. This is shown more
clearly in the cumulative CPI rating chart below Figure 5.
Cable Scores
300
100
200
Telco Scores
370
300
400
Poor
Average
Good
High
Figure 5. Cumulative CPI Rating for Telco & Cable
Generating a Scenario Run
The next step in determining the migration of the customer, is to generate a scenario run.
Firstly we must set-up the rules under which the scenario run will take place, this is done by configuring
the Telco & Cable profiles.
The user can flex key parameters in order to alter the outcome of the simulation run. Parameters are flexed
using slider bars and outputs are in the form of graphs and other visual indicators.
Key Parameters
The model has a 20 year simulation period and the key input parameters for Cable are shown below in
Figure 6.
Figure 6. Key Input Parameters for Cable profile
The Cable Input parameters determine the share of the market they control and therefore, the number of
customers they can target. This is a flexible of method getting values, as the cable operators’ profile can be
changed to reflect their position in the market place.
The key input parameters for Telco are shown in figure 7.
Figure 7. Key Input Parameters for Telco Profile
Global Variables
The following inputs affect the climate under which the model is run.
Barriers to Entry
Barriers to Entry
CABLE res rey ir»
00 02 04 06 08 1.0
TELCO q SSS »
00 02 04 06 08 1.0
Figure 8. Barriers to Entry.
This slider can be set between 0 & 1. The higher the value, the more difficult i
up any new customers.
Market Inertia
is for both operators to pick
Market Inertia
ere
00 02 04 06 08 10
Figure 9. Market Inertia
This slider decides customers’ reluctance to change operators. Here 0 represents that customers will readily
change operators; 1 represents that the customers will not change operators no matter how favourable a
competitors services may appear.
Model Outputs
Once all the profiles have been created the model is put through a simulation run. The simulation period
can be varied from 2 years (minimum) to 20 years (Maximum).
The model dynamically simulates a competitive environment, using historical, (therefore unchangeable)
data, for years 1985 to 1998. It gives a visual alert once the thresholds have been reached for your
customer profiles. The model then migrates (churns) the customers to the operators.
Figure 10 & 11, below, show final outputs from a simulation run.
0:00]
Figure 10. Outputs for Customer Take-up Vs Churn for Scenario Runs
We can see a typical output here for the middle income_price sensitive customer profile. All customers
with this profile have contributed to an average churn rate of 16% from Telco, for the simulation period,
and 22% churn from Cable. Although Telco has lost more customers in the same period than Cable, their
customers have simply churned away to a default ‘Other operator’ , and not migrated to Cable.
This simulation run shows that Telco's CPI ratings helped to reduce churn for them, and similarly the Cable
CPI ratings help them to gain more customers in the simulation period, although their churn rating was
higher.
Figure 11. Market Share of Telco & Cable after Simulation Runs
The graph shows the market position of the two main operators, compared to the rest of the market.
¢ Telco 65%
© Cable 20%
¢ = Other 13%
¢ — Unprovided Customers 2%
The Importance of Customer Behaviour
Telecommunication markets around the world are becoming more competitive, and with technology
improving leaps and bounds, telcos are finding it hard to hold onto their market share. The most important
factor in the increase in competition is the removal of barriers between switching customers from one
service provider to another.
Customers in the telecommunications market are no different to customers in any other commercial market
place; they both expect a high level of service and reliability from their suppliers. It is becoming
increasingly difficult to differentiate between the technologies the telcos use. The only indication the
customer has that the operator they choose will provide an adequate level of service, is by advertising or
word of mouth. Therefore a move towards some form of standardised performance ratings between
operators will do much to improve customer choic
Future Developments
One of the most important aspects of this modelling work is trying to understand the drivers behind
customer churn, and preventing it.
This model can be used to help reduce customer churn by:
¢ — Identifying the most profitable customers and what can be done to maintain their loyalty
e Decreasing promotion costs by using only those promotion packages that are best at attracting those
customers which will remain loyal after the promotion is over
¢ Targeting the most profitable segments of the market for the most aggressive promotions
Telcos can greatly reduce churn costs by isolating characteristics of customers likely to remain loyal to a
product. When targeting new customers, a company can then choose the "right" customers to target. Also,
models might suggest ways to treat current customers differently to induce them to remain loyal to your
product. With a wealth of customer and market information, telecommunications companies can employ
information they already have to reduce churn in their customer base and thereby lower customer
acquisition costs.
SUMMARY
A model has been developed to try to understand the reasons for customers joining a particular operator.
The model does this by setting a typical customer profile and then setting varying levels of service for the
operators, in order to determine which mix of factors persuade a customer to churn.
Tf telcos can understand the reason behind Churn it will enable them to retain their customer bas
proactive in servicing the needs of their customers, and in the long term to reduce the costs incurred.
and be
ACKNOWLEDGEMENTS
We wish to thank Ann Matthews for reviewing this document and Mike Matthews for authorising its
release for publication.
CC BY-NC-SA 4.0