With a decade of experience in customer retention, we construct the analytical background for customer retention according to the logical structure summarized below:
Churn Models and the refined Conditional Churn Models represent the basic pillars of analytical support for customer retention.
Scores provided by Churn Models represent, in percentage form, the probability rate of customer attrition.
Uplift Modeling can be used to demonstrate the extent to which a retention offer is likely and able to change a customer’s intentions to moving over to the competitor.
Having accurate knowledge on the effects products have on customer lifespan, as well as identifying both the rational and less rational aspects of customers – together with assessing the savings opportunities made available by the given products – provide assistance in the selection of client-tailored retention offers.
Retention Campaigns can be optimised with the selection of appropriate target group and suitable offers.
A systematic Campaign Evaluation helps in fine-tuning retention strategies.
Churn Models
Churn (or attrition) models are widely used predictive models aimed at defining the likelihood of individual customers moving over to a competitor. These models represent the fundamental pillars of analytics behind retention activities. Our company is one of the most experienced suppliers in churn prediction for the telecommunications sector. Throughout our projects, we have worked with various service providers and created systems that made systematic forecasting possible for a range of different churn scenarios. Based on the predicted event, we can distinguish between
Customer Churn Models meant to specify the probability of losing a customer,
Turnover Churn Models that forecast the discontinuation of customer activity, and
Product and Service Cancellation Models
Also closely related is the prediction of customer account suspension due to unpaid bills, which will be covered in more detail when we discuss our solutions for collections management.
Conditional Churn Models
Churn Models are generally run on a monthly or weekly basis and so churn probabilities are updated at a similar frequency. Conditional Churn Models can supplement the basic model in cases where the regularity of forecasting is not frequent enough to ensure the possible fastest response to events likely to exert a significant influence on customer churn (for instance, the conclusion of a loyalty contract). Pre-defined conditional probabilities take effect immediately after such specific events occur, thereby updating the churn forecast without having to wait for the results of churn models that might only come in much later.
Churn Prediction
Well-constructed and appropriately maintained Churn and Conditional Churn Models allow for the identification of customers at risk. The customer-specific churn risk can be easily transferred, in a simple-to-use format, to customer service representatives’ CRM interface. In addition to calculating simple churn probabilities, the profiling of customers at risk and the behavioural segmentation of various motivations can help in identifying the reasons behind attrition.
“Uplift” Retention Modelling
Once a churn forecast has been prepared, another challenge for retention programmes is ensuring that the right offer is delivered to the right target audience. In other words, although Churn Models can be used to identify customers at risk, they are useless when it comes to selecting those customers deemed worthy of being targeted. Uplift Modelling is a technique that takes into account that customers react differently to the same offer. It can therefore be very helpful in narrowing down the list of recipients of retention offers to those known to have a positive response, which can translate into huge savings to the company.
Instead of simply calculating the attrition rate, Uplift Models also specify the probability of a retention offer being able to change a customer’s intention to move over to a competitor (see retention offer optimisation and sales uplift models).
Retention Offer Optimization
Successful retention needs appropriate retention offers. To assist with the offer selection process, we offer the following support analytics:
- Customer Lifespan Impact Analysis. Shows how long and to what extent individual product purchases, plan migrations and other customer events affect customer loyalty.
- Rationality Check. From a marketing standpoint, a key customer trait is their financial and service use awareness. Classification of customers based on their rationality can help in the selection of the most suitable offer.
- Next-Best Offers. Taking into consideration the customer’s previous behaviour and the available product range, the savings prospects behind the acceptance of an alternate offer, as well as the consequences thereof to the service provider, can both be estimated. This can help service providers in fine-tuning appropriate offers in their strategy.
Retention Campaign Optimisation
Attrition forecast models based on appropriate churn definitions, as well as prospective uplift and retention offer models, serve as the basis for establishing a successful customer retention programme. Customers listed in the system can be segmented into different target groups in the retention strategy by means of scoring and alongside various other criteria.
Modelling can ensure the long-term optimisation of campaigns, both ad-hoc and recurring. This method is suitable for optimising outbound campaign offers and for appropriately serving inbound customers alike, provided that churn scores set by the system are transferred, in a suitable format, to CRM interfaces as well.
Campaign Evaluation
Following up the quality of retention activities and drawing the right conclusions are of vital importance for planning subsequent campaigns. Our effectiveness analyses and reports are always carried out by means of comparison with corresponding control groups. Our solutions allow for retention activities to be monitored from various aspects:
- With the effectiveness analysis of retention offers, we compare the churn rates and behaviour of treated customers to those of members of an appropriately selected control group.
- By the efficiency analysis of campaign targeting, we specifically gauge the performance of models on the basis of which the target group has been assessed (churn, uplift and offer models).
- Measuring the success rate of customer relations associates helps compare the effectiveness of campaign activities between individual associates.
- The results of the effectiveness surveys and comparisons are both made available to campaign managers and customer relationship managers in the form of regular reports. A diminishing rate of offer effectiveness could indicate changes in the business environment that could be avoided or eliminated by reconstructing or updating the models.
