Commentary

Is Accurate Pre-Bid Valuation A Pipe Dream? Discrepancies In Measurement Make It Difficult

Viewability remains a controversial topic among advertisers. Is the MRC’s standard enough? Why aren’t rates rising? How important is viewability, anyway? A handful of people have reached out to Real-Time Daily over the years to offer their thoughts on the topic.

But things came to a head when Tom Triscari, CEO of Labmatik, and Max Knight, VP of global marketing science services at Turn, each wrote op-ed pieces that seemed to disagree on some key points. Both Triscari and Knight are smart thinkers with valid points, and their arguments on the matter go much deeper than conversations about viewability usually do.

So over the course of several weeks, Real-Time Daily facilitated a quasi-panel with Triscari and Knight via email to discuss viewability and the challenges associated with the important metric.

The debate hinges on a single hypothesis: That bidders should consider impression quality along with conversion probability. In simpler terms: “Not all impressions are created equal, and calculating the value of an impression should be part of the pre-bid process." An ad being viewable does not instantly make it $1 more valuable no matter what, for example. Obviously, Knight and Triscari agree on this general fact.

advertisement

advertisement

But Knight notes that this inherent inequality is, more or less, an insurmountable problem. That’s because the different viewability measurement vendors aren’t all on the same page. Vendors A and B might both be measuring the same site, but there’s a relatively high chance that their recordings will differ.

With mixed agreement between the measurement providers, it’s virtually impossible to reach an equilibrium. Outside of partners using the same measurement vendor -- which only avoids the problem -- there's no obvious solution. 

“Currently there is no reliable mechanism for deterministically labeling impressions as viewable or human,” noted Knight. “Different viewability technologies will measure across these dimensions differently, so while we can all agree what viewability should be, different measurement providers won’t read the same set of impressions the same way, and thus all these labels are inherently probabilistic.”

In short, “not all impressions are created equal” comes with a caveat of “and we still can’t deterministically measure said inequality.”

Knight noted that there are several formulas that can be used to make the best use of the probabilistic data -- including an equation referred to as the “union of events,” which assigns a value to every possibility (such as a non-human viewable ad) and spits out a weighted average -- but it still doesn’t get around the fact that “probabilistic” is just a fancy way of saying “this is our best guess.” It also doesn’t get around the fact that the “union of events” formula will vary depending on whose data you use.

“This, coupled with the guarantees that buyers are demanding, means that inventory valuation is fundamentally flawed due to a lack of standardized weights and measures,” Knight said.

Triscari didn’t disagree with Knight, but countered by pointing out that the discrepancies between viewability vendors are getting smaller and smaller (or, at least, that the industry is working toward bridging the gap).

Perhaps they don't agree on the whole spectrum of requirements, but I'm guessing when you open up and look inside there are some things that are pretty even throughout those measurement vendors,” Triscari said. “I’m thinking of it like a Nielsen panel -- it's not perfect, but if you can't measure everything without it turning into a mess, then pick a subset that projects out.”

Knight conceded that over time, as vendors close the gap on their respective measurements, “the inclusion of impression quality valuations in bidding [will become] less problematic.”

But he warned that factoring in these differences is not a “silver bullet,” because the problem is not in the difficulty of (probabilistically) accounting for the discrepancies -- the problem rests in the “effect those discrepancies have on the bid.” Knight called it a “snowball effect” that is a “value killer for publishers.”

The conversation continued, with Knight and Triscari discussing what steps the industry can take to improve viewability. That’s for another column.

Next story loading loading..