November 20, 2008

Catalog And E-Mail Drive Retail Sales Via Matchback Analytics In Some Instances

Jim Wheaton brings us an article that those of us who've analyzed the data firmly believe in --- that after controlling for prior customer frequency, multichannel customers are not fundamentally better than single-channel customers. The article explores how analysis of customer behavior can trump established marketing beliefs.

I am reminded of a meeting in 2003 at Nordstrom. Our sales rep from Abacus visited, suggesting that matchback analytics would prove that our catalog marketing efforts drove a billion dollars or more of retail volume (our retail business was maybe $5 - $6 billion at that time). Abacus clients were using matchbacks, learning that retail customers who received catalogs were likely to buy in stores in the days after receiving a catalog. Our Abacus rep took a leap of faith --- the mailing of catalogs caused a billion dollars in retail sales --- the catalog caused the sales to happen.

And Abacus was right, based on the analytics available to them. You run a matchback analysis, and you see that retail customers buy in the days following the mailing of a catalog.

Multichannel Forensics, however, suggest a subtle distinction that we must keep in mind.
  • Matchback algorithms are reasonably accurate when the retail channel is in Acquisition Mode.
  • Matchback algorithms are somewhat accurate when the retail channel is in Hybrid Mode.
  • Matchback algorithms are highly inaccurate when the retail channel is in Retention Mode.
If you don't have the bandwidth to run a Multichannel Forensics analysis, at least run mail/holdout groups through your matchback algorithm, subtracting the difference in matches.

There is a faction of the marketing community that sees matchbacks as a religion, and this is ok, because it is sometimes better to do matchbacks than to do nothing. That being said, subtle performance differences and profit increases happen when we combine Multichannel Forensics, Test/Holdout Groups, and Matchback Algorithms --- for catalog mailings and especially e-mail campaigns.

Back to Nordstrom. We did the testing, we did the Multichannel Forensics analysis, and we did Matchback Analytics. Our findings ran contrary to the best practices suggested by the marketing establishment.

We killed our catalog program in June 2005. From July 2005 - June 2006, comp store sales were positive. In other words, without $36,000,000 of catalog advertising that matchback algorithms suggested were driving maybe a billion dollars of retail sales, we were able to increase retail sales. Multichannel Forensics and Test/Holdout groups told us that a retail brand in Retention Mode yielded highly misleading Matchback results.

So this is the deal. The vast majority of the marketing establishment believes that catalog and e-mail marketing drive sales to other channels, as evidenced by matchback analytics. And these folks may be correct, especially if channels operate in Acquisition Mode. When channels operate in Hybrid Mode or Retention Mode, the rules change --- the organic percentage overrides matchback analytics.

No comments:

Post a Comment