StoneMor Partners' Dan Shin on Leveraging Predictive Analytics for Customer Acquisition
For the younger crowd, it may be a shock to learn that funeral preplanning accounts for 60 percent of business for cemetery and funeral home owner and operator StoneMor Partners of Levittown, Pa. Much of that is due to predictive modeling that StoneMor—the steward of 232 cemeteries and 59 funeral homes in 28 states and Puerto Rico—uses to figure out which consumers would be most interested in settling their earthly concerns prior to death.
StoneMor already had been sending consumers direct mail, bearing the original names and return addresses of the neighborhood cemeteries and funeral homes familiar to the prospects—not "StoneMor." But the targeting efforts primarily consisting of profiling and preselecting prospects for funeral preplanning weren't producing much of a response lift.
So StoneMor began preplanning its predictive analytics push in January 2008. Then, in November, the company started purchasing data from companies like KnowledgeBase Marketing (KBM) of Richardson, Texas, and plugging the information into marketing analytics software from SPSS of Chicago.
Soon, consumers began buying more burial lots, lawn and mausoleum crypts, burial vaults, caskets, memorials, and related services. StoneMor Market Statistician Dan Shin explains why.
Target Marketing: What type of data does StoneMor Partners purchase from list companies?
Dan Shin: We would like to do predictive analytics in all of the sales channels that we have for our marketing department, [which are direct mail, Web, e-mail and telemarketing]. We were trying to accomplish some of the business objectives. And one of our major business objectives was lifting response for our core product line, which is a direct mail plan for preplanning. So we first implemented and tested it on this channel before we went ahead and implemented it on various channels that we have. The reason … is the other channels require transactional data, or in-house data, historical data—anywhere from response data to purchase behavior. But for the direct mail campaign, because it was prospecting, we didn't have that information in-house. So a lot of the prospect data that we obtained was actually from a third-party vendor like KBM or Equifax, Experian, some of the larger data providers.
TM: What predictive model does StoneMor Partners use to figure out who is most likely to be receptive to preplanning a funeral?
DS: To be specific, our current, in-house predictive model is a logistic regression algorithm. The variables that we use to create this model [are] primarily based on demographic data, ailment data, buyer data and survey data, which is supplied by our third-party vendors …
TM: What process does StoneMor Partners employ in order to optimize response from the 2.7 million pieces of direct mail it sends out each year?
DS: After we pull our specified population from the prospect pool, we run the records through the model algorithm and place them into ranking deciles. And then, based purely on the cutoff points of each decile, we take all the records and mail out to as many people as we can that fall into each of the top deciles, while taking into account the required numbers of records we need to drop a week.
TM: What results has StoneMor Partners seen from this predictive modeling initiative?
DS: From November 2008 and into the beginning of 2009, we have effectively integrated predictive modeling and data mining and have been able to increase our response rate by approx[imately] 96 percent ([from] 0.175 percent before modeling to 0.343 percent after modeling) even through such financially difficult times. This has allowed us to receive 80 more leads a week for our inside sales team.