The Dog Ate My Analysis

December 2012 JOURNAL OF ADVERTISING RESEARCH 395 the groWth of analytics in the past few years has been nothing short of phenomenal. Sparked by the birth of digital commercial platforms and fueled by the emergence of “big-data” streams and cloud computing, analytics has penetrated nearly all aspects of decision making in most Global 1000 companies. As is true in any “bubble,” however, people—in this case, managers—fall prey to “irrational exuberance” and have a tendency to under-, over-, or mis-apply the tools to often ridiculous extremes. The problem is particularly acute in the marketing and advertising business. It is astounding to me that, with all the tools at our disposal, 99.9 percent of the ads I see or hear every day are actually meant for someone else. For example, while recently listening to Pandora on my iPad, I heard a commercial for a weekend sale at a car dealer located about 50 miles from my home in New Jersey. The experience was a little creepy, but mostly amusing, as I was in China at the time. Could Pandora’s ad servers not bother to check whether I was actually anywhere near my home (or at least in the same country) before choosing the ad they played? Maybe this was just Pandora’s way of enticing me to sign up for the premium, ad-free service. If the phone company mis-routed 99 percent of calls, however, we would all just disconnect our phones and go back to the old reliable post office. More recently, the escalating level of analybabble in the marketing community about “digital attribution” and “social chatter” are on a trajectory to become such an enormous waste of company time that could be much better spent on the softball field or in quality improvement process training. Though some of the core principles and ideas under pursuit are sound, the execution is flagrantly illogical. For example, if you really want to know what proportional impact a series of digital advertising exposures have on my behavior, do you not need to actually see what I do when I am not online? Otherwise, are you not just over-analyzing a very small portion of my shopping and buying behaviors? Are you not optimizing within a small box just based on what you can see? Yet, marketers are transfixed by this concept of fractional digital attribution—applying rocketscience math to a rich but limited data set in the hopes of squeezing out some insight or evidence that will stand up to scrutiny. Extending this logic, you might attempt to attribute my health to what I had for breakfast this morning. This is precisely the sort of myopic approach that killed earlier generations of media mix models in favor of the more holistic strategic allocation methods in place today. In that case, the very narrow framing led to over-reliance on short-term tactical promotions and eroded brand loyalty—a very dangerous bias that went unrecognized for years. I suspect there is a lesson to be learned here. Similarly, if your analytics is pointed at trying to predict what I will do next by monitoring my social comments online, let me save you the effort: I am most likely to stop posting and turn off my computer. What do you infer from that? That I am no longer interested in computers? (BTW, does it strike anyone else as amusing that we are spending so much time and energy applying analytics to monitoring aptly named social “chatter?”) As social measurement moves beyond the “volume and polarity” phase into semantic analysis and natural language processing, we run the risk of once again being seduced by precise instrumentation into a false confidence that we are reading reality. Wrong. We are reading a thin slice of reality at best. Now, if that slice happens to reliably predict behavioral outcomes, we celebrate having caught lightning in a bottle. In the vast majority of cases, however, social voice is showing up as a great lagging indicator, not so much as a predictive tool (except, of course, for the extremes of negative or positive viral reactions). The Dog Ate My Analysis The Hitchhiker’s Guide to Marketing Analytics