Leveraging Big Data to Make Cross-Channel Value Measurement a Reality
Advances in multivariate modeling are paving the way for advertisers to mine this existing data in ways that allow them to not only better optimize within a given channel, but also connect the dots between what's happening in one channel and another, bypassing the well-known issues with cookies in addressing cross-channel, cross-device measurement.
Advanced multivariate modeling techniques enable advertisers to use algorithms that help predict future outcomes and solve the complex pricing, placement and allocation issues surrounding everything from paid search to display to Facebook advertising, including taking into account how an ad in one channel is influencing purchases in another.
Tapping big data through advanced modeling techniques to optimize individual channels isn't new. In recent years, this approach has helped advertisers solve complex bidding challenges for individual keywords that number into the tens of millions. Multivariate models factor in historical cost and conversion data, in addition to hundreds of variables influencing performance — everything from day of week and seasonality to weather forecasts and market factors — for every single individual keyword. The resulting insights enable individual bids to be set accordingly to maximize an advertiser's results toward their chosen goal.
This ability to tap the most granular or "atomic level" data to drive dramatic results at the channel level now also has the potential to measure cross-channel value. As with the paid search scenario described above, the requisite data is already at most advertisers’ fingertips. The trick is applying multivariate modeling to uncover the insights hiding within this sea of data.
For example, a goal that still eludes many a digital advertiser today: measuring the cross-channel value of brand advertising (e.g., display). To understand the cross-channel impact of display on paid search, for example, a particular display ad's impression volume becomes yet another variable fed into the multivariate models. With the ability to analyze at the atomic level (individual ad and keyword), the models can identify whether that particular display ad's impression volume is a top variable driving the predicted performance of a particular keyword.