5 Barriers to Churn Prediction and How to Address Them

6th April 2017

By Lynn Heidmann

Given that it costs five to 10 times more to acquire a new customer than to retain an existing one, it seems obvious that all businesses should be engaged in some level of churn prevention. But many businesses’ strategies result in addressing customer churn only once it’s too late, which is unfortunately an expensive (and generally ineffective) method.

The alternative is churn prediction. If you can predict with a high level of accuracy which customers are likely to leave, you can capture them at that critical decision period before it’s too late.

Sound too good to be true? It’s not! Using predictive analytics to anticipate customer churn is actually a very accessible way for even smaller and less experienced teams to get started in the world of machine learning. Yet many businesses still aren’t doing it.

Here are the top five barriers cited as a hindrance to starting a churn prediction project and tips to overcome them and get started.

  1. Admitting that churn is an issue. The first step to resolving a problem is admitting that there is a problem. And for some businesses, this is a hurdle to overcome if they are gaining customers quickly – it’s hard to see why churn should be addressed when the overall number of users is rising. But the reality is that churn has the power to plateau the growth of any businesses even if that business is gaining customers quickly.
  2. Settling on the right definition of a user or customer. Before deciding what it means to churn, you need to decide when users or customers are eligible to even be considered users or customers. If all they do is sign up for an account but never use it, have they churned? Or were they not engaged enough to even be considered a customer in the first place?
  3. Settling on the right definition of churn. The next step is deciding after what period of time of inactivity (or waning activity) constitutes churn. Defining a churn period that is too long risks creating predictive models with artificially low churn rates, not capturing enough people and defeating the purpose of predictive modeling. But defining a churn period that is too short makes it difficult for marketing teams to evaluate churn prevention campaigns because they ultimately can’t distinguish between organic actions (users or customers who would have come back anyway without intervention) and effective campaigns.
  4. Getting data. This step is actually much easier than most businesses think, because the minimum data required for a churn prediction model is simply some form of customer identification and a date/time of that customer’s last interaction. Almost all businesses already have this data available. Of course, you can develop more accurate and robust predictions by adding features with more data, but this shouldn’t be a barrier to getting started.
  5. Getting predictive. If you’re a beginner when it comes to machine learning and algorithms, it shouldn’t stop you from predicting churn to improve your business. You can use a tool like Dataiku Data Science Studio (DSS) to run open source algorithms to predict churn in a clickable interface without having to write any code.

With these barriers out of the way, you’re ready to really dive in to a churn prediction project. Refer to our how-to guidebook for complete step-by-step instructions for beginners as well as more advanced tips.