Can Your Data Find Sarah Connor?March 18, 2013 By John Haake
Fast-forward to 2009, when nascent demand-side platforms made relying on editorial content as a proxy for in-market audiences a thing of the past. Seemingly overnight, it became easy for media agencies to leverage tracking cookies to locate humans who visited their brands' Web properties and target them with cheap remnant inventory. With the addition of third-party data, media agencies began to scale up their laser-targeted campaigns with "look-alike" audience segments.
The problem was, these display programs didn't perform any better than traditionally purchased media because the algorithms powering the campaigns were fueled by data that was misused or too unreliable to be an effective predictor of consumer behavior—especially when compared with data that can be gleaned from a live shopping session. As it still stands, we entrust our media programs to automated buying platforms that leverage suspect data to optimize the past to predict the future, and are completely blind to the opportunity of the here and now. This threatens the very existence of humanity. Locating Sarah Connor is our only hope!
The Signal and the Noise
It's easy to dismiss data fails as anything less than life-threatening—after all, who hasn't gotten a chuckle from auto correct fails in our texts, or when Google tries to guess our search terms before we finish typing. But it does get serious when misinformed algorithms bring down financial systems or airliners.
Still, algorithms have quickly achieved superhuman status in ad tech circles. Every CMO I know is busy retooling teams to ensure data scientists and marketers transform the cascades of first-party consumer information that brands are sitting on into value-creating insights. But it's good to remember algorithms fed a diet of unreliable, stale, unactionable and too-granular data will most certainly not deliver optimum results. For best results, you must go directly to the source of the data.