Database Marketing is about Correlations not Causality

A few articles coming out recently from Forbes and NYTimes have drastically overstated the intent and generally the capability of large retailers’ ability to analyze customers and their shopping behavior.

Having some experience in Retail and specifically with the company and people mentioned in the articles below I can say that database marketing tactics are not nearly so glamorous (or devious) as they have been represented in these articles. Statisticians don’t sit around cooking up ways to invade someones privacy. They sit around thinking up ways they can get another percent of incremental productivity, usually response rate, out of an email or direct mail campaign.

The data used in database marketing sometimes can be perceived to be invasive and certainly these articles are written with a tone which is geared primarily shock value. More often than not the data available to retailers is inaccurate for a variety of complex circumstances and it is a part of the database marketers job to “control” for those inaccuracies by executing marketing campaigns with proper experimental design techniques. Personally, in the decade or so of practicing database marketing I’ve never been able (or interested) in pinpointing an individual by name, address, or any other private information that your average consumer would consider invasive.

Database marketers can not determine causality. We do not know the actual cause of customer behavior. We are only left with correlations, complex subjective and objective algorithms and pesky margin of errors.

As much as I’d like to tout being apart of sophisticated database marketing projects in the past the honest truth is, as a database marketers, we sift through a lot of complex data, statistics and resource to come up with marketing programs and campaigns that many times don’t work. And when a program or campaign finally does work it is not based on the identification of causality rather it is most often just correlations between various data.

– Josh Smith –