Masterpiece Refines Modeling TechniquesClaiming current models lack recency and timely feedback, Masterpiece Corp. has introduced a system that uses artificial intelligence to help marketers refine campaigns in real time and mail to smaller, more efficient lists.
Current models use past buying patterns to predict future response. While they are helpful in determining why a certain demographic buys a certain product, models work off information that is a few months old, offer only a single snapshot of data taken before list selection, and can't adapt to existing market conditions based on incoming responses. As a result, response rates remain low, typically under 5 percent.
Masterpiece, Bensalem, PA, promises its first group of new clients improvement in response rates of 100 percent or more, or modeling charges will be waived.
Self-adaptive modeling works off the premise that prospects can be better targeted with real-time rather than static data. The new model uses artificial intelligence -- complex computer programs based on genetic algorithms or neural networks that measure hundreds of variables and learn over time -- to alert marketers of trends in incoming responses.
Bruce Ratner, a statistician and consultant who designs models at David Shepard Associates, Dix Hills, NY, and who led a discussion on modeling techniques at the DMA's List Day last week, said the use of artificial intelligence (AI) for modeling is not a revelation, and plenty of companies use it in some form. Its use has picked up as computing power has increased to levels where AI modeling can be implemented at a desktop.
Artificial intelligence from Masterpiece helped a telephone company reduce churn among cell-phone users by detecting that those who purchased voice mail tended to stay with the company. As a result, the company started bundling voice mail with its phones and attracted more profitable customers.
Self-adaptive modeling also can react to competition in the market. A bank, for example, could improve the interest rate on a loan in mid-campaign to match or beat a competitor's new product.
"The system remembers responses," said Dan Mittelman, director of sales at Masterpiece. "You can refresh it or devise different campaigns when one isn't responding.''
The self-adaptive model campaign is conducted with smaller list segments over a longer time period than a single, large-drop campaign. Instead of mailing 100,000 pieces all at once, Masterpiece suggests that marketers mail 10,000 pieces, evaluate those responses, then mail the next 10,000 to a more targeted list.
Masterpiece claims artificial intelligence will improve response rates for each mailing until it determines that all good prospects have been exhausted. The computer tells marketers to stop the campaigns when response rates and profitability have been tapped out. This staggered approach eliminates waste from mailing to nonresponsive audiences and ensures that marketers are adequately equipped to handle the expected volume of responses.
"We can tailor a campaign to produce close to the exact number of responses the client wants for a day, week or month,'' said Mitchell Cohen, vice president of operations at Masterpiece. "This avoids a barrage of responses and allows staffing to handle it. You won't waste the [customers] that want to do business with you.''
Ratner advises marketers to proceed with caution and said AI models require more testing and validation than normal techniques. He doubts that AI alone can dramatically lift response rates.
"There are no real new mousetraps,'' Ratner said. "The only thing new is implementation. Refinements could be significant but it takes a lot of minor improvement to make a significant difference [in results]. The Catch-22 is that artificial intelligence finds things traditional techniques have not found but what it finds are spurious relationships. What it sees is not necessarily a real finding.''