Response Modeling Revisited

Marketers are beginning to look at response modeling outside of the co-op databases, according to Chris Montana, senior vice president at Hackensack, NJ-based list services firm Mokrynskidirect.

“The economics of modeling have changed,” Montana says. In the past, building a response model cost $10,000-$25,000. “Typically the cost of the model outweighed the benefits of derived from the model,” he notes. That’s not necessarily the case today.

“One of the modelers we work with extensively actually builds the model at no up-front cost and charges a reasonable fee per 1,000 on the model output,” Montana says. What’s more, the model is prevalidated based on earlier campaigns: Mailers can identify the potential strength of the model before putting anything in the mail.

Direct marketers pre-validate by applying the model against a past mailing. In this process, the mail tape from a previous mailing is sent to the modeler with no response data. The modeler will then apply the response model and identify the buyers vs. non-buyers. This information is returned to the mailer, based on what we have seen with surprising accuracy and model confidence.

If they like what they see they can apply the response model to go deeper into strong lists in addition to raising the level of performance on marginal files. The new hurdle, Montana says, is educating list owners and managers about the value of allowing mailers to apply models to the names they rent. Generally an overage of names from the list owner are sent to a third-party firm; the model is applied, and the mailer pays the list owner running charges on the names dropped from the model and the agreed-upon rental price for the names being sent to the service bureau going into the merge. The list owners can generate incremental usage by getting existing users to go deeper into their fils as well as increase the number of overall mail dates for which these continuation mailers take names.

“In addition,” Montana says, “we have seen numerous situations where a list that has fallen out of a schedule due to declining performance is able to be put back in when the model’s performance is factored in. It really is a win for all parties.”