Don’t Grind Advertising Data into Dust

2017 07 10   Peter Feinstein   Advertising    

Wow … this long and meandering article is a stellar example of reduction to the ridiculous in an effort to determine the indeterminate.

The idea of parsing out viewers on a per-second basis with the intention of getting really granular is pretty funny. It falsely confers a false sense of authority to a process where none is deserved.

The whole notion of establishing a definitive value to viewership based upon second-by-second data is truly absurd; yes, it does give so-called data scientists more to talk about, but it does nothing to establish absolute, predictable modeling of consumer behavior. Only ad-tech geeks (not a compliment) benefit from this kind of nonsense.

The human being is complex far beyond any algorithm. Haven’t you ever stopped to wonder how something so simple as an algorithm could accurately predict the vagaries of human behavior? It can’t. It doesn’t. It offers insight into correlation, not causation. And that’s a potentially dangerous, mistake-ridden and expensive path to follow.
Correlation identifies that when “A” happens, “B” sometimes happens too. Causation establishes that “A” actually causes “B.”

So if you’re all gaga about algorithms and the power of data to prove when to do what on TV, just remember that all you’re seeing is deeply researched correlations – kind of like: “when the moon is full, I always have to blow my nose.” You might think that was a pretty big deal – an indicator of importance – until you found out that I blow my nose every night. The moon’s presence is merely coincidental, from which you can build a great correlation model, but nothing approaching a causation model.