Cover Story: Prospecting for Smiles
Ma notes that she knew this option of list optimization was possible when she came to Smile Train, because she has a direct marketing agency background in the "for-profit world." Those agencies included Ogilvy, McCann Erickson and Saatchi & Saatchi. That work meant overseeing accounts such as General Motors, Motorola and Nestlé.
She just wasn't sure how to tailor that agency experience for this nonprofit until discussing it with Infogroup and starting the program in the fall of 2010.
"We started just testing it, initially," Ma says. "So how we built it is we took response files—people who'd responded to our direct mail acquisition programs—and, obviously, our house files and our suppression files, to make sure we removed anyone from that pool.
"After all the de-duping and removing names that were on our suppression files and house files," she continues, "we were left with an available pool of prospects of 117 million households." Smile Train then:
- Appended data to those prospects (demographic, transactional, lifestyle);
- Ran regression models to identify key predictive variables;
- Scored the prospective pool of 117 million records;
- Identified approximately 29 million households that would be most likely to respond to direct mail; and
- Segmented those households into 20 tiers for targeting.
The process helped Smile Train find prospects who were most likely to become donors—people who looked a lot like responders to the charity's past direct mail acquisition efforts.
"We initially tested into it, back in the fall, using a mix of compiled data (prospect database) and external lists," Ma says. "We were doing about 50 percent from the prospect database and 50 percent from the continuation/external lists. ... And, obviously, we did some tracking to test, to make sure that we were able to track the results coming from the prospect database." She adds, "when we actually compared results, we found that the compiled names from the prospect database performed on par with our average external continuation lists and outperformed the middle to bottom third of a typical list plan."