|
Towards a better understanding of customer behaviour
 In
the latest instalment of his series on CRM, Khalid Sheikh explains some other
techniques for prediction of customer behaviour
RFM Analysis is based on the assumption that the customer response to promotions
depends greatly on when specific customers last made a purchase, how often they
buy, and how high their purchasing value is.
Association discovery
Association discovery detects groups of similar items or events and, hence,
can be used to detect sets of items or events that occur together. In CRM, association
discovery is used to do market-basket analysis to help businesses find out what
products are purchased together by finding out associations between the different
items that customers place in their shopping baskets. The discovery of such
association rules can help in the development of marketing strategies. Retail
chains can use affinity groupings to plan arrangements of items on store shelves
or in a catalogue so that items often purchased together will be seen together.
Association discovery can also be used to identify cross-selling opportunities
and to design attractive package or groupings of products and services. The
discovery of association rules is mostly dependent on the discovery of frequent
sets, and algorithms attempt to determine the set of frequent item sets in a
given set of transactions. To reduce the number of discovered frequent sets,
the user can specify constraints to indicate which combinations of items are
the subject of interest.
Sequence mining
Sequence mining identifies combinations of activities that occur in a particular
order. It is part of a broader data mining field called temporal data mining.
Temporal data mining is an extension of data mining that involves a non-trivial
extraction of implicit, potentially useful, and previously unrecorded information
with an implicit or explicit temporal (time-phased) content from large quantities
of data. It can help a business predict customer behaviour from business events
captured in the transaction processing system.
Examples of sequence analysis include the following rules: ‘Any person
who buys a car also buys a steering lock after that;’ or ‘Some patients
tend to develop reactions after two months with this combination of drugs.’
Events occurring at different points in time might be related by causal relationships,
that is, an earlier event might appear to have caused a later one. Such relationships
can be discovered by analysing sequences of events to discover common patterns.
From a given set of data sequences, sequence mining attempts to discover subsequences
that are frequent, that is, they occur as many as or more than a user-defined
percentage of times.
Sequence analysis is used to determine whether customers are doing things in
a particular order. Sequence mining has important applications in retailing
where it can be used to predict the next sequential purchase of different customers.
In the telecommunication industry, customers might display a behavioural pattern
indicative of impending churn, which can be captured.
Customer value assessment
Customer valuation helps companies to focus limited resources most efficiently
on the best and most valuable customer relationships. A fundamental principle
of marketing is that not all customers are the same. On one extreme, there are
customers that cost a great deal to attract, require a great deal of service,
and offer little in terms of profitable purchases; on the other extreme, there
are customers who seem to be anxious to know about your company and products,
require very little service, and make sustained, high value purchases. Thus,
different customers represent different levels of profit for the firm.
Successful organisations attempt to define characteristics
of the best customers, to then estimate the lifetime value of such customers,
and to adjust marketing strategy accordingly. Customer value assessment involves
assessing customer profitability and representing it through the following parameters:
- Customer profitability
- Customer lifetime value
- Customer rating (ABC analysis, Overall attractiveness
or satisfaction, etc.) Profitability analysis requires an integrated business
model for contribution margin analysis that combines various types of revenue,
product costs, and sales costs to produce a coherent picture of customer profitability.
- Most CRM solutions integrate activity-based costing (ABC
costing) with customer profitability with a minimum manual effort. This makes
it extremely simple to allocate customer-related costs (such as customer visits,
customer support, or campaign costs) to respective customers
- It should be noted that customer profitability does
not replace product profitability, which is still indispensable to a company’s
success.
A customer can be unprofitable but could have referred three
high-value customers to your firm, thereby rendering himself very valuable.
Despite not being currently profitable, a recent college graduate shows several
signs of emerging profitability and thus might be considered valuable over her
lifetime.
Customer Lifetime Value (CLV)
Customer lifetime value is the net present value of the total profits that a
company could realise with the average new customer within a given customer
segment during a given number of years. CLV is the true value of a customer
that can be considered as the most appropriate measurement of how much an organisation
would or should be willing to invest to acquire/retain a customer.
Customer Portfolio Optimisation
For strategic decisions in marketing, sales, and service, companies usually
don’t examine the characteristics of individual customers. Instead, companies
examine what forms their customer base—the customer portfolio. CRM performance
indicators, such as customer lifetime value, customer ratings, or strength of
customer relationships, are used to assess customers or customer groups more
effectively and to determine the most appropriate policy for winning over customers,
providing them with the right kind of service, and retaining them for future
business relationships.
The analysis of the customer base with a suitable classification of the customer
profile, the customer portfolio, is an important tool for optimising the make-up
of the customer base. For example, customers can be sorted into different categories
based on two key figures—customer attractiveness and strength of customer
relationship to divide customers into a portfolio of four categories of customers
as shown in Figure 1.
| Objective |
How the results are used |
Analytical Techniques Used
|
| Identification of homogeneous customer segments |
Used as the basis for decisions in marketing, sales,
and service functions |
- Clustering
- Scoring,
- RFM* (Recency, Frequency, and Monetary) Analysis
|
Customer profiling
- ABC analysis
- Customer portfolio analysis
|
Profiles of existing top customers can be used to
target prospects with similar profiles. There is a high probability that
these customers would eventually become best customers. |
- Scoring
- Decision Trees
- Subjective judgement
|
Purchasing behaviour prediction
- Propensity-to-Buy Analysis (Which products is a customer is likely
to buy)
- Next-Sequential Purchase
- Product-Affinity Analysis (also called Market Basket Analysis to
predict which products are likely to be bought together)
- Price Elasticity Modelling and Dynamic Pricing (Determining the optimal
price for a given product for a given market segment)
|
- Make offers that are best suited to the needs of individual customers.
- Exploit cross-selling and upselling opportunities
- Package certain products together and offer at a discounted price
|
- Association discovery
- Decision Trees
|
- Churn Prediction (Determining customers who are likely to churn [leave]
based on attributes similar to those who have already left)
|
To plan personalised marketing interactions to
motivate customers who are 'likely to churn' to stay. Examples of such
actions include pre-emptively offering discounts or fee waivers.
|
- Association discovery
- Scoring
- Decision Trees
|
The author is associate professor of Supply Chain Management
at S P Jain Institute of Management & Research, Mumbai. He can be contacted
at khalid_sheikh@hotmail.com
|
