The Promise of True Digital Measurement — and Its Betrayal

Digital advertising exploded partly because the public swooned for digital unlike for any new media in history. But it also promised something the old-fashioned type of advertising could not. It promised to be measurable. Verifiable. It was supposed to be the end of guesswork and the beginning of numerically accurate advertising science.

Advertising science is not new. We already know that audience acceptance has been measured by industries for decades, and with increasing emphasis on data over guesswork. The recent success of the Mad Men series brought the venerable “focus group” into millions of non-advertiser flatscreens. As illustrated in the series, until the advent of digital, advertising science was of a sociological variety. It was the study of people themselves and what they said, and extrapolation of data was an enormous component of its utility.

Nielsen and Anti-Nielsen

Extrapolation of data is required when volume is lacking. The prime example of extrapolation in the delivery of advertising data is the science of television audience measurement developed by the Nielsen company. For decades, its word was law in broadcast media. In television, low ratings meant almost instant cancellation. This was because Nielsen purportedly “knew” how many millions were tuned in (or not) to a broadcast event.

The method by which Nielsen would claim to “know” was by having installed devices on the televisions of several thousand selected homes – a vanishingly tiny percentage of the general population. From this tiny sample, they deployed extrapolation techniques to “enlarge” the data onto a bigger screen. They created mathematical formulae that would take the miniscule sample and play out a thousand into a million and a million into 10 million. For all anyone knows today, it may have been reasonably accurate. Or it may have been wildly incorrect. Without knowing what the many millions of non-Nielsen households were doing, it would be impossible to determine true accuracy.

Validity of extrapolation aside, Nielsen has been a success for two generations because they’ve been able to demonstrate substance. But digital made substantial claims to be, in effect, the anti-Nielsen. It was supposed to do away with the need for data modeling and extrapolation. It was supposed to tell you how “users” were in fact interacting with digital media with much, much greater accuracy. This new level of accuracy was to be accomplished by measuring actual behavior in unsampled data packets that would result in verifiable “true” counts. If, as Rand Schulman has said, “Creative without conversion equals zero,” then digital analytics was supposed to provide unadulterated evidence of how creative drove conversions. This tends to be fairly electrifying stuff for folks who know the code. It means “you can create beautiful commercials in cyberspace, but if your commercials don’t get anybody to buy anything, then at best you’ve gained nothing and probably have wasted your money.”

Creative vs. Data – Battle to the Death?

Sufficient budget for digital analytics and data analysis in general are not in place at most enterprises. The need for it is in too many cases only vaguely apprehended by senior management that grew up on “hunches” and “people I know.” Analytics success is thwarted by big-advertising “creatives” who build digital campaigns and then purport to accurately self-measure as a courtesy. Self-measurement is not quite an oxymoron, but it does set up a natural conflict of interest. Content creators want to measure themselves the same way a ball team would love to call its own balls and strikes. Except there really wouldn’t be any game to play if that’s the way it were umpired.

Ask any number of in-house agency analytics specialists and they will likely tell you they often feel like a voice crying in the wilderness. Facts discovered by agency analysts are, if an industry scuttlebutt is to be believed, fungible. They fall victim to the opinion of the senior ad wizard about what worked and what did not. In the end, the enterprise that hired the agency gets little untainted information about how much of its advertising dollar was wasted.

This concept of conversion/non-conversion lies at the heart of any critique of digital analytics. And as we move toward a multichannel measurement model, it is perhaps an even more potent indictment of the amped-up promises of digital analytics in general.

Why is digital analytics failing its promise? How does that contribute to its own destruction – and ultimately to an unhelpful distrust in the verity of an entire class of nonfinancial business data?

Related reading