We hear it every day: That one of the biggest issues for multichannel retailers is customer attrition.
But saying it is easy. It’s a little harder to come up with a solution.
One obstacle to understanding defection has been the inability to predict which customers are most at risk to leave. And it’s a even tougher in retail because of the voluntary nature of the relationship.
Retailers’ traditional response to this has been to wait – sometimes a long while – before trying to reactivate customers. Quite often this might be as simple as having a business rule—say, that a promotional offer will be mailed to a customer who has not shopped in twelve months.
But there are a few problems with this approach. First, if customers have really left the brand, then waiting many months to try to win them back from is unlikely to succeed.
They may have been spending their money across the street. Or maybe some just didn’t need your products for a while.
Loyalty and frequency programs may be part of the answer, as are win-back programs. But what has been missing is a way to proactively identify potential defectors and intervene before they are gone.
First, though, it is useful to consider the customer lifecycle. Some buyers may be one-time shoppers—that’s often true of people who bought for the first time during the holidays. Others will buy only for holidays or other events.
The rest come to shop when they need something—or simply feel like it. In online retailing, since there are no anonymous buyers, we can gain more data to analyze.
In order to predict which customers are most likely to leave, we cannot employ traditional statistical methods used in contractual settings. Rather, we need to use a tool called a stochastic model.
What’s that? A stochastic model uses the patterns of past purchases to provide two predictions: Is a customer still a customer? And if the person still appears to be active, how many times will he return?
The information in a stochastic model gives a solid basis for predicting future attrition and purchasing frequency. But you can add additional variables to the model to get even more useful results. For example, you can incorporate past promotional response history to segment our customers even further.
You may have one segment of likely defectors that is relatively insensitive to promotion, which would suggest that spending heavily to retain them is unlikely to be effective. On the other hand, another group that appears to respond to promotions could be a better audience for retention marketing.
And other variables can be included in the stochastic model, such as product affinity or demographics, to provide even more differentiated segments.
So, although retailers will also need to create and deliver effective marketing programs aimed at attrition, they have now gained an effective tool to identify at-risk customers and prioritize those worth fighting to keep.
David King is chief executive officer of Fulcrum, a vendor of analytics, technology and program services.