This started with Marketing Mix Modeling (MMM) years ago but now has a second coming with the data exhaust coming off of digital.
One fly in the ointment is that in service of quick wins the protagonist in these stories often just uses digital data because it is handy -- which of course leaves out TV, in-store and other powerful marketing influences. When you leave out stimuli in analytics of stimulus-response, the response caused by the missing stimuli gets credited to the stimuli you have included. In a recent case one of the largest research companies in the world presented a massive analysis of digital data, admitting that the ROI credited to digital was 1500% off the mark established by MMM (with MMM firmly believed to be the right numbers by client management).
This post summarizes a short course I taught on April 8 on “The Holistic Use of Big Data.” You can watch the whole webinar here. The high points I advise:
Don’t just look at online sales, digital ad cookie exposures, direct mail and other easy in-house data; if you look you will find people who can provide TV set top box and offline purchase data, in-store stimuli, the ability to measure store traffic linked to cross-media ad exposure, more depth of household level demographics and many other important stimuli. TiVo Research/TRA is one example. They are already direct matching at the same-household level all of the following types of data and more.
Check out all the companies you know have set top box and/or offline sales data and compare what they offer in terms of making a grand Big Data match with all of the Big Data you already have.
Because display and feature in-store stimuli tend to be more powerful than media (not always), make sure to compare suppliers based on the percentage of stores that are audited weekly for these measures.
Compare suppliers in terms of the projected weighted sample versus universe estimates by demographics and geographics.
Make sure that households with DVRs are either included with full program episode-level actual DVR playback measurement, or are excluded.
Use the safe harbor method to direct match by household name/postal address and yet blind the research company to all such personally identifiable information (PII). The methods of doing this are now well known and practiced by Experian and Acxiom/