Share this content

Customer churn modelling & the likelihood to lapse

18th Oct 2021
Share this content

Retaining customers is generally cheaper than attracting new ones, and data shows that long-term customers tend to be more profitable, both in terms of a lower cost to serve and a higher take-up of new products and services. These long-term customers are also less susceptible to competitors’ attempts to lure them away.

One powerful tool in a business’s toolkit for retaining customers is churn modelling. This involves identifying a customer’s “propensity to churn,” the likelihood that a given customer will soon stop doing business with a brand. If a business uses modelling successfully, it can identify customers who are at risk of lapsing in time to influence their behaviour.

Churn modelling is particularly effective for businesses that rely on subscription models – which these days can mean anything from SaaS and news subscriptions to “meal kits” and toothbrushes – as customer cancellations are easier to quantify than a person’s waning interest. However, with clever use of the data, businesses of all kinds can use churn modelling.

Identifying churn risk

Determining whether a person is likely to churn – and effectively changing that likelihood – requires that a business understands the key drivers behind churn. This involves comparing data on customers that have already churned to find common characteristics and pre-churn behaviour. These traits can then be compared against the same data from people that have remained customers to find the tell-tale attributes of impending churn.

The type of data that commonly identifies churn risk often includes transactional information, such as what products a customer has purchased, the price of the product and the frequency of purchase. Customer tenure and lifetime value are also key factors.

However, transactional data alone is rarely enough to differentiate customers that lapse from those that a business can retain.  Transactional data must be combined with demographic information – such as age, affluence and lifestage – as well as engagement data – including website visits and email interactions. Having all of this information in a unified, accessible repository is a prerequisite for successful churn modelling.

Through the process of understanding churn risk, the business also needs to be cognisant of the difference between voluntary churn and involuntary churn. Voluntary churn occurs when a customer actively cancels their subscription. Involuntary churn refers to cancellations that were not driven by the customer, such as those stemming from the expiration of a credit card or a missed renewal notification. Businesses should exclude involuntary churn from their churn model build. These risks can be addressed through tightening up procedures.

Not for everyone

Propensity modelling isn’t suitable for all organisations.

First, businesses must ask whether they have the requisite data to create a model that can distinguish a customer likely to churn from one that you are likely to retain. This comes down to the data that a business has access to, as discussed above.

Next, businesses need to determine whether the data is centralised. Modelling, like any other statistical process, depends on having clean, well-sorted data available all in one place. This is important both for the initial analysis to identify likely churn characteristics and the ongoing work of feeding and refreshing the model. Effective modelling can only happen if a business already has a robust approach to handling data.

Businesses also need to consider the type of product or service being offered, as product mix can have a significant impact on modelling. Retailers of high-volume, low-cost items and fast-moving consumer goods will have a harder time creating strong propensity models – when the product mix on offer changes so quickly it’s hard to get a model to work effectively. To combat this, high street retailers, well aware of the challenges of the cross-product purchase element, create propensity models based on category spend and use this to apply more general incentives, such as a flat discount when spending a certain amount or additional incentives to make a purchase in-store.

Avoiding the pitfalls

Implementing marketing interventions off the back of churn modelling can deliver tangible benefits to a business, just so long as you are able to avoid the pitfalls along the way.

Simply identifying customers likely to churn is not enough to prevent churn: the business needs to be able to influence the customer. This means reaching them in time and having ways to change their mind – the specifics of this will vary between businesses and often involves employing a test and learn approach.  In addition, businesses should focus their efforts on customers that will have a long-term value to them, and they must learn how much they are willing to invest to get the desired return.

Market context is also crucial (if there’s a new competitor in town for example) as some factors might not be addressable with an incentive, however good.  In these instances, other tactics such as reinforcing benefits and highlighting customer satisfaction ratings may be more effective at preventing churn than an offer-based strategy.

Finally, models aren’t set and forget, and businesses must be committed to maintaining them over time. As customer behaviour changes, the models need to be updated and refined.

This is a lot of work, but it’s all worth it if a business can significantly cut churn and keep customers surprised and delighted.

Replies (0)

Please login or register to join the discussion.

There are currently no replies, be the first to post a reply.