Predictive Modeling Charms Halloween Store Sales
A national retailer of party goods operates seasonal stores to sell Halloween costumes and decor. Store locations vary year-by-year, so the company relies on direct mail and digital marketing to drive demand to new locations during a short seasonal shopping period. Prior year customers are an obvious choice for direct mail, but response from existing customers is not enough. To increase demand, the company relies on list brokers to procure lists of consumers who shop with other catalog retailers as a likely source of additional customers.
Building Predictive Models to Yield Better Results
List marketing can be challenging as lists ‘wear out’ over time due to saturation. The same customers get targeted over and over to the point that even though they may be ideal prospects, they’re also tapped out. This retailer needed a new list source to offset wearout from the other lists and hopefully identify a new scalable marketing universe.
Rather than hunt for another vertical list from a marketing co-op, we used past customer order data to develop a look-alike model using a national opt-in consumer file. This way we could score all individuals in the country with a probability indicating how well their profile (demographic, attitudinal, behavioral) matches this client’s existing customer base. National consumer files tend to respond at lower levels than vertical lists as there’s no unique singular consumer behavior that indicates membership on a national consumer file. Furthermore, look-alike modeling is not a unique process, but we believed our model development process paired with specific audience segmentation was unique and would identify a more highly-responsive target audience than others.
Sending Direct Mail to a Custom Audience
Our modeled direct mail list became the top-performing list.
Our modeled list generated a 26% higher response rate and 28% more revenue per mailing than the remaining vertical lists. The other lists had been sourced from specific catalog marketers, tested and tweaked over time to produce their best possible result. Yet sourcing names from a national consumer file won. List testing is still popular for many direct mail marketers, but many times a better performing and more scalable audience can be produced through predictive modeling. Overall, direct mail can be a huge factor in successful holiday marketing.
The company has continued to source modeled lists from us for three consecutive years now and we look forward to many more to come!
