One of our clients, a multichannel merchant of high-end home decor that uses the co-op databases and vertical files aggressively, had an all-too-common problem: Mailing about 5 million-6 million prospecting names annually, it was seeing performance fall off as well as shrinking prospecting universes on continuations.
We were challenged, as most brokers are, to find enough continuation names that met the performance requirements for mailing. Many files were falling just short of acceptable performance. But though we refined the selects, tested enhancements, and applied zip models, most of these changes failed to deliver the necessary lift in response needed to get the files back onto the plan. Clearly we needed a new tool.
We began working with a partner to develop a way to build very affordable response models. The model we used in this instance was built on past mailings using data from 2,000 responders and the mailing tapes that generated those buyers.
The model indicated that a response lift of 21%-56% was possible using the top five deciles. We validated the model by applying it to a previous mailing not used in the model construction. We then compared the modeled names to the actual responders from that mailing. The response lift was validated with a lift of 15%-49% (Performance variance is typically+/- 25% from model projections.) The process, from receipt of data through model build and validation, took five days, and the mailer incurred no additional costs.
The merchant then tested the model in holiday mailings in a variety of scenarios: marginal files just below acceptable performance levels; going deeper into top files where performance would indicate a need for response lift on older segments; exchange files with available balances that were not being used due to performance issues; and against the house file to reactivate names not selected for mailing. List owners sent the ordered names to our modeling partner to apply the response model. Our modelers then sent the selected modeled names to the mailer’s service bureau and supplied documentation to the list owners of names selected for the merge. From receipt of names at the modeler to running the model and shipping to the service bureau took no more than two days.
To date, the response lift generated from the model is ranging from 26% to more than 300%. Names that would have been left out of the plan are now performing above breakeven, allowing the opportunity to go back and use more names from top files, put marginal files back into the plan, and take advantage of old exchanges successfully. By using response models, it is now possible to identify those names most likely to respond to the offer at a very reasonable cost on vertical files, with no loss of recency, no loss of list selectivity, and no risk to the mailer.
Susan Darling is senior vice president/director of list brokerage and Wendy McLaughlin is vice president, list brokerage for Hackensack, NJ-based list services firm Mokrynskidirect
Do you have a case study or success story you’d like to share? If so, contact Mark Del Franco at [email protected] or Tim Parry at [email protected].