Online-Offline Integration for Retention Marketing

(This post is part of a series on the state of multichannel metrics today, one year after the book came out.)

In a subscription based business model, e.g. mobile phone service, you have two ways to make a customer:

1. gain a new one
2. renew an existing one, i.e. retain them

When Unica’s clients from the mobile phone carrier industry report about their work they invariably start by describing how vital #2 is to their industry.

The market for mobile phone services is so saturated that the only way to gain a new client is to take them away from a competitor.

Sounds like a place for marketing innovation.

A few innovative Telcos have hit on ways to combine online and offline worlds in order to improve their success rate with retention marketing.

As I learned from one of my colleagues last year, here is how one company went about it.

Online-offline integrated analytics for detecting attrition signals

At a large US mobile phone carrier, traditional retention marketers were working on predicting which customers were about to leave for the competition. These clients would be included in retention marketing efforts.

Originally, the statisticians had been going after this job the old fashioned way, i.e. trying everything from a customer’s contract details to transactions (i.e. usage) and demographics details to find something that would predict attrition.

But the only variables that showed any influence were the age of the subscriber’s phone device and the amount that they paid on the last bill.

Not exactly enough to catch someone in the hot act when they are about to walk out through the door.

Yet, it was going OK.

To put it in numbers, the marketers were able to reach 70% of customers at risk of leaving by contacting 40% of the possible audience. So their predictive models were giving them some amount of lift.

But wait a minute … If someone is thinking about switching would they not likely be coming to the web site and doing something on there that deviates from their usual click behavior?

Might they not be checking available promotions or upgrades or ways to strike a deal?

The idea seemed so promising that the statisticians gave it a go.

They took a chunk of historical web data for registered clients. They paired that up with the same customers’ historical churn data in order to train a predictive model (along w/ the offline data).

And what they found was impressive

Indeed there were predictive click behaviors on their web site but it wasn’t intuitive.

• Clients on a low subscription contract would have one kind of online signal that revealed their intention, e.g. address change.
• Clients on a higher rate plan however turned out to send a different signal with their clicks.

The numbers rewarded them.

Now, when contacting 40% of the potential list they were able to reach an extra 15% responses for a total of 85% of potential responses.

That doesn’t just means lots of stamps and mailers saved.

It means foremost saving the cost of special discounts that they would have extended unnecessarily to clients who weren’t thinking about leaving anyway.

Highly worthwhile.

Real time?

Most of the online-offline integration case studies that you may have read about in this series were of interactive nature, i.e. online click behavior would prompt action within a short period of time.

Here we have an example of how a company first took a historical chunk of data to train their model. No real time needed here.

But now that the model has been trained, fresh web analytics data would be fed to it regularly in order to keep predicting current customers at risk.

The morale

This is yet another strong business case for integrating online and offline analytics.

No wonder the case is strong. After all

• The case for competing on analytics is strong.
• The case for using behavioral data is strong
• Click data is a rich lather of behavioral data

It is time for the lollygaggers to stop acting surprised and jump on board!