Marketing

Prediction, Renewals and Big Data

(this is a repost of a post written by me for PAKRAgames. It is part three of a series of four.)

Do you know why your customers renew their subscriptions or services? Do you know how to predict whether any given customer will renew? I suspect you probably have an answer something like, “Well, yes, but it could be better.”

So let’s make it better.

And let’s make better marketing investment decisions by doing so.

The big marketing shift

One of the most important shifts in marketing in the past decade is the combination of the exponential increase in the amount of data available to us and the advent of tools to analyze that data. This trend is generally referred to as “big data.”

Those of us who relied on data-driven marketing in the era before all this data became available used statistical models to attempt to predict behavior based on the few items we could measure (does anyone remember when the measure of success was the order percentage from direct mail … yes, postal mail?)

Big data is an improvement, but it presents us with a very big issue: We now have more data on our hands than we can possibly handle, and, while dramatically improved in the past few years, the tools available to handle this data are barely in their infancy — in fact, they are barely keeping up with the exponential growth of data.

This leaves us with only one way to get the most out of all the data at our disposal: Ask good questions. I cannot emphasize the importance of this enough. And what constitutes a good question isn’t always what you think it should be.

The key question

To answer the questions I posed above, we want to ask not why customers renew, but rather what predicts whether a customer will renew.

To answer, we must use both statistical and data-mining techniques. Historically, we’ve been pretty good at looking at this question and answering with either subjective measures (such as attitude during on-boarding) or objective measures (such as number of successfully resolved support/service calls).

But we need to go farther. We need to look at actions taken during the course of the use of the service. Are your customers adding new users? Are they putting specific kinds of data in the system? Are they completing a full cycle of whatever your service is supposed to help them do within a certain period of time after signing up? All these and many, many more items are potentially significant.

How do you know which ones are? Here we use good old-fashioned statistical techniques to correlate the data to renewals.

What will we learn? We don’t know until we’ve done the analysis, but we might find out that customers who take one series of specific actions within the service always renew, and customers who take a different series of specific actions never renew. Whatever we find out, we’ll have a very good method for predicting the likelihood of renewal.

Now what do we do with that?

To me, this sounds a lot like lead scoring. We can assign grades to customers based on their likelihood to renew and take different actions based on that. For customers we think are 100% likely to renew, maybe we just make them a really good offer. For customers 5% likely to renew, maybe we make a rescue offer. For customers on the bubble — maybe they will, maybe they won’t — we might assign a renewal rep to learn more about how they value the service and turn them to renewal.

Which decisions you make for your particular service depend largely on the economics of the renewal and the specific relationship with and value of that customer (we’ll discuss this in the next part of this series). You can then make good investment decisions and create renewal programs that will help you maximize renewals for your most valuable customers.

Leave a Reply

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