Problems Calculating Retention Rate

Jim answers questions from fellow Drillers
(More questions with answers here, Work Overview here, Index of concepts here)

Topic Overview

Hi again folks, Jim Novo here.

What is your customer retention rate? Well, that kinda depends on how you define the customer. Have you had an internal discussion, and more importantly, solidified agreement across divisions / functions on the definition of an (active?) customer? Please do.

For example, is someone who hasn’t interacted with your company in any way for over 5 years still a customer? You see, if you don’t specifically define a customer, then you can’t have discussions around topics like reactivation, retention, Lifetime Value (LTV) and so forth. Where to start? With segmentation. Create segments of similar customers, then try to decide which segments are still customers; this exercise will get you going down the right track. The Drillin’?


Q:  Seasonality has great effects on customers’ purchasing activities in the retailing industry, as you may easily understand.

A:  Yes…

Q:  Furthermore, what you call Latency has also great effects on their purchasing activities, (I mean, for example, the customer who purchased a coat in one winter season are not expected to purchase another until the next winter season and so forth.)

A:  Yes, but you are profiling customers, not products, right?  The customer who bought the coat may also buy a dress, shoes, pants in other seasons?  Your approach so far sounds a bit too product centric…

Q:  Here is the problem, how these issues of seasonality and Latency must be taken into consideration for calculating retention rate?

A:  Well, you can take it into account or not, depending on your objectives.  What is the objective of the analysis?  If the objective means you should take these issues into account, then you probably should segment the customer base to do so.

For example, a customer who only buys winter coats – and nothing else – probably is not your most valuable customer.  So is this segment important to track by itself?  How many customers are in this segment?  What kind of winter coats do they buy – the most expensive available or the cheapest?  The answers to these questions determine the value of the segment, and then based on your objectives, decide on whether the segment is worth tracking uniquely.

Let’s say this segment is fairly large and they primarily buy very expensive winter coats with high profit margins.  This makes the segment worth tracking by itself.  So rather than including them in the RFM or retention scoring, track them separately using the Latency metric.  If these seasonal coat shoppers usually buy their expensive winter coat every August like clockwork, and then you identify customers in the segment who do not buy the coat in August, you have potentially lost this sale.  Depending on the objectives of the analysis, you can act on this information immediately to try to get a sale of a coat (or accessories, if the coat sale has been lost) or act the following year in a “pre-emptive” way to increase the chances a coat will be bought from you.

Now, let’s say this segment is fairly small and not worth a lot of money.  I would simply ignore them, perhaps even exclude them from any specific analysis.  At some level, the data doesn’t justify the effort, particularly if the segment is so small that it is not efficient to take any action on.

For example, many companies eliminate 1x buyers or 1x visitors from critical analysis, because the skew they create makes the analysis less actionable.  A good example of that is RFM scoring – if 50% of the population are 1x buyers and the population is small, the result of the scoring is meaningless.  

That’s not to say you don’t track the size of the population and so forth, but these populations can be so large and so unresponsive they simply are not worth any kind of specialize marketing approach.

The point is, when you are modeling a customer base, if you have reason to believe that segments exist with behaviors outside the norm, find out if they indeed exist, how big they are, and what their value is.  Then determine if this value justifies handling the segment as a unique group. If it does, then proceed with the best model for that segment, regardless of what model you are using with the other segments.

This is an ongoing process.  The more segments you discover, the more models you will use, and the more accurate your modeling will become! Good luck with the effort!


Jim

Get the book at Booklocker.com

Find Out Specifically What is in the Book

Learn Customer Marketing Concepts and Metrics (site article list)

Download the first 9 chapters of the Drilling Down book: PDF 

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.