Blog: A stepwise approach in customer retention

What is the best retention strategy?

Mustafa Bozkurt

Last month I cancelled my membership at the gym where I used to go at least once a week in the recent past. The last couple of months I skipped my regular visits to the gym for different reasons or excuses, resulting in the cancellation of my membership. I did not have the feeling that my irregular visits fully justified the monthly fee of almost € 40,-.

Marketing literature on customer retention acknowledges the fact the best strategy to grow your customer base is not to lose any customers. But what is the best retention strategy? Let’s discuss a data-driven approach for a successful retention strategy in the following sections.

Segmenting your customers

One of the keys to a successful retention strategy is segmenting your customer base. Creating different customer segments will help you to understand your customers who all make use of your product/ service, but may have different needs and behaviors. There are many ways to segment your customer base, the different segmentation techniques and input variables are out of the scope of this article, but please find here a brief background on segmentation.

Let’s illustrate the above with the gym example in the introduction and some fictional segments in table 1 below.

Table 1. Fictional gym segments & profile

The main purpose of segmentation is to summarize a lot of information about customer groups that have similarities which then can be used for tailored strategies. Segmentation can help you to identify the most valuable customers of your company and find out what they need most. This is important because these customers are naturally the highest priority for your retention campaign(s). Most important criteria for segments are they should make ‘sense’ and be relevant to the business. Next to that, the segments should be targetable and statistically different from each other.

Creating these segments can be done on internal data that is already available or data collected by market research on your customers.

Churn model

Once you have segmented your customer base, the next step is to create a churn model to identify potential churn candidates. The benefit of creating a churn model lies in the fact that it can discriminate between customers with a high churn probability versus customers with a low churn probability. In the end this will result in better allocation of scarce resources for customer retention.

Let’s briefly summarize the steps in developing a churn prediction model which are presented below in figure 2:

  1. Data sources: Collection of all data about customers from different sources
  2. Join & merge: Joining all data sets into a single source
  3. Data preparation: Recoding, aggregating, collapsing variables & outlier handling
  4. Dimension reduction & sampling: Removing variables with high correlation and low variance
  5. Partitioning & model training: Partitioning of the data into test set vs validation set & training of different classification models
  6. Model selection & deployment: Comparison of different model predictions & selection and deployment of winning model on total customer base

Figure 2. Data steps in churn prediction model per segment

As mentioned already, figure 2 gives a brief overview of all steps in developing a churn prediction model. More specific details for each step are out of the scope of this article, but please find here a detailed approach, CRIPS-DM, for model development that has a similar setup as described above.

Retention strategy per segment

Once we have the customer base segmented and build a churn model, the next step would be to combine the result of both analysis together with the customer value to identify customers that would be valuable to include in a retention campaign.

Table 3 below gives an example of the calculation to identify customers for a retention strategy, where the cutoff point is ≥ 20.00 €. This cutoff point is arbitrary and might be increased or decreased for each business case depending on expected costs and revenue of a retention campaign.

Table 3. Customer selection for retention campaign

The retention strategy should be mainly focused on customers with high churn probability and high value for the company in each segment. As we can see already the number of customers to be included in the retention campaign might vary for each segment.

Figure 4 gives an overview of all the findings combined in the previous sections: segment sizes, segment values, proportion of customers to be contacted for retention campaign and potential recommended retention strategies for each segment.

Figure 4. Customized retention strategy per segment

A customized retention strategy per segment has a couple of benefits which are as follows:

  • Tailored retention strategy per segment makes the offer more relevant for the customer
  • Measurement of effectiveness of different retention strategies
  • Allocation of resources to high value segments or high value consumers within a segment
  • Increase of overall retention rate


This article summarized the steps to develop a segment specific retention strategy and explained the benefits over a non-tailored retention strategy for the total customer base. Starting with a segmentation followed by a churn model, we demonstrated an approach to decrease the churn rate of your total customer base.

Going back to my introduction- was there anything the gym could have done to keep me as a member? Usually I visited the gym at least once a week, but the last two months prior to my cancellation I did not visit the gym at all. With a segment specific retention strategy (I would be an “Occasional visitor”), the gym could have recognized an unusual drop in my visit-pattern and pro-actively make me an offer that might keep me as a member. Now it’s too late, I already made up my mind and cancelled my subscription. Thinking about it, there is only one question left out there: what subscription am I going to cancel next?

Stay up to date with our industry insights
Receive our insights on big data, predictive analytics and marketing automation once every two months.

Follow us on: