Bring 'Em Back
While there is no hard and fast rule as to when a marketer should develop a reactivation program, it usually becomes an issue when the "net flows of people in an organization start slowing down," according to Steve Briley, chief statistical officer, Merkle Direct Marketing. That is, when customer acquisition efforts are compromised by the number of customers "falling out the bottom of the funnel," he explains.
Smart marketers will plug the bucket beforehand by monitoring defection rates. Marketers need to have an "ongoing review of the sheer volume of those lapsing or defecting from an organization," notes Briley, adding that it might be at the point that represents from 2 percent to 5 percent of your acquisitions where you really have to start paying attention to your attrition rates so you can get "in front" of the problem before it affects your net growth.
Dan Rubin, vice president, analytics, Harte-Hanks CRM, also recommends making attrition a key metric that is frequently tracked and updated, but not so often that you can't differentiate "noise from reality." He adds, "Know what's normal for that time of year and be able to understand the cycles you see."
A reactivation model is one potential tool that can be used to implement a reactivation plan. The decision to build a model is determined by the number of inactive records or if the data set simply is large enough to support a statistically valid model, according to Ruf.
Another consideration is how you house your data. "If you have a relationship database and all your reactivation modeling is being implemented there, and campaigns are being executed there, developing a model is a great idea," says Briley. However, he explains, if you need pieces of data from three or four different legacy databases to implement an effective reactivation model, you're probably going to need to take some smaller steps before tackling the challenge of building the "ideal" model. One way to do this, offers Briley, may be to focus on the "historical purchase history to estimate a 'good' model that gets you 75 percent to 80 percent of the way there."