Wednesday, January 25, 2017

28. CUSTOMER LIFETIME VALUE 2

OBJECTIVE

Estimate the retention and future spending amount of customers.


DESCRIPTION

In the previous chapter I explained the principles of CLV and its calculation. However, one of the problems was the estimation of the retention rate (for which the simplification was to apply the average retention rate of similar customers) and future spending amounts (we assumed that the average spending amount of each customer will not change in the future). Despite the facility of the implementation of this approach, it can be much too simplistic and fail to estimate CLV reliably.

There are several methods for estimating retention and spending amounts, but some of them can be far too complex. The method that I will propose has a good balance between accuracy and implementation simplicity and is based on customer segmentation and probability.

The first step is to take the customer data of year -2 and segment them based on their value and their activeness. For the value we can use the amount spent in a specific year (which is a mix of the average amount spent per purchase and the frequency of purchases), and for activeness we can use the recency of the last purchase (for example the number of days between the last purchase and the end of the analyzed year). For activeness we can also use a mix of recency and frequency. In the second step, we have to define a certain number of customer segments. The segmentation technique can be either a simple double-entry matrix or a statistical clustering technique. We can for example end up with six clusters:
  • -          Active high value
  • -          Active low value
  • -          Warm
  • -          Cold
  • -          Inactive
  • -          New customers

The idea behind this technique is to estimate the retention and spending amount using the probability of a customer remaining in the same segment or changing segment and by applying to this customer the average spending amount of the new segment. To calculate the probability of moving from one segment to another, it is necessary to segment the customers into year -1 and create a transition matrix (transition among different segments from year -2 to year -1) in which probabilities are calculated for each combination of segment groups.

Customer Lifetime Value Transition Matrix

Transition Matrix of Customers’ Segments

With the probability transition matrix we can simulate how the segments will change in the future and maybe realize that we are dangerously reducing active customers in favor of inactive ones and that we need to acquire a slightly bigger number of each kind of customer to avoid a decrease in profits. In any case with this matrix we can simulate several years ahead and estimate how many customers will still be active. We can also estimate their value by multiplying the average value of each segment by the number of customers of the same segment in a specific year (year 0, year +1, year +2, etc.).

In the proposed template, I have added an estimation of new customers acquired each year to simulate the total number of customers and their value a few years ahead. However, to calculate the CLV of the current customers, this value should be set to 0 and then the total value of each year discounted by the discount rate.


TEMPLATE


No comments:

Post a Comment