Commentary
Calling All Video Ad-Buying Humans: You Should View This
- by Tyler Loechner , Staff Writer @mp_tyler, May 11, 2015
The
quality of digital advertising has been top of mind of late, but which key metrics -- such as viewability, click-rates and completion rates -- give marketers a better shot at buying quality
video advertisements? TubeMogul, a programmatic video ad platform, teamed with media measurement and fraud detection firm Integral Ad Science to find out.
The two companies studied over 127 million impressions from over 400 video ad placements bought via TubeMogul's programmatic platform and measured by Integral. The companies say they compared each placement against one another and assigned a rank. The rankings were based on human, in-view completions.
These human, in-view completed views were deemed by TubeMogul and Integral to be the “correct” type of video ad, and these ads were then ranked again based on other metrics such as viewability, click-rate and standard completion rate.
“We looked at how often a marketer would choose the less effective placement when optimizing on other metrics and called the frequency of incorrect decisions ‘percentage wrong,’” TubeMogul's report reads.
In other words, TubeMogul and Integral found the “correct” ad placements -- i.e. ads that were completed, in-view, by humans -- and then determined whether or not optimizing for other key metics (viewability, click-rate, etc.) would cause the marketer to buy “incorrect” ad placements.
To some degree, this compares apples to oranges. Viewability, click-rate, standard completion rate and the like all optimize based on one, single metric. “Human, in-view completions” optimizes for three metrics -- verified humans, viewability and completion rates -- all bundled together. Of course it will be better.
But the purpose of TubeMogul and Integral’s report was not to simply prove that optimizing for “human, in-view completions” was better, but rather to find out just how much better it was.
The verdict: Marketers spend an extra 15% to 22% per human, in-view consumption when optimizing on other metrics, such as viewability or click-rate.
“Given that our sample is drawn from advertisers already investing in monitoring viewability and non-human traffic, there is reason to think the broader market may stand to gain even more.”
TubeMogul and Integral found that when marketers optimize for “standard completion rate,” they make the wrong decision 38% of the time. The second-worst standalone optimization metric is “click-rate,” which led to wrong decisions 30% of the time, followed by “viewability” (17%) and “unfiltered in-view completion rate” (10%). “Unfiltered,” in this case, means not accounting for suspicious activity -- i.e. potential non-human traffic.
The full report can be found here.
Tyler Loechner is a MediaPost Reporter covering real-time, programmatic, RTB and more. You can reach Tyler at tyler@mediapost.com.
A most interesting report, Tyler. Thanks.
I think that it is very important for everyone who is concerned to carefully read the full report---it's not that long--and, especially the definitions on the last page. What they are using for "viewability" is the "standard definition"---at least 50% of an ad must be on the screen for at least two consecutive minutes. The term "human in-view completions" does not mean that they actually know whether a user "watched" an ad. Rather, it reflects the fact that if one takes an ad and divides it up by quartiles, each quartile was in-view, using the same standard definition. I assume that this is meant to reflect a measure of ad exposure intensity and I have no quarrel with it, if used as a relative benchmark.
@Ed thanks for reading.
I think it's a useful bit of research. I understood the quartile bit of the report to simply show that what TubeMogul/Integral found in terms of "what's the best metric to optimize against" rang true even if the ad was not fully completed.
@Ed, thank you for reading so carefully! Let me address part of what I understand to be your concern:
I'll defer to Integral for the specifics of their technology, but human, in-vew completion in this research means that the ad was in-view from beginning to end, or to the quartile referenced. As soon as an ad drops out of view, it no longer qualifies as being in-view beyond that point. So, if I watch 50% of an ad, kick it out of view and come back at the 75% mark, it's only in-view complete to 50%.
I hope that clarifies things.
For what it's worth, Tyler is also correct. The intent of the quartile analysis is simply to say that optimizing against partial completion (ignoring viewability) is no better than optimizing to 100% completion (again, ignoring viewability).
This matters a lot as you set goals for yourself, for your team, for your agency and for your media providers. One thing doing this research made clear to me: fraud and viewability have to be part of your internal and external KPIs. Otherwise you might not get what you really want.
@Taylor, thanks for that clarification.
Hi Ed.
Sorry for picking up late on the thread but I was moving house.
Did you mean two consecutive minutes - or seconds.
I also have issues with defining that as "human", as it implies that the human was there at the device and attentive. Seeing an interaction 75% of the way through a video could represent a person returning after a 'nature stop'. Put simply, a browser has little idea whether a person (or even multiple people) is present. OS based software has a better chance (i.e. can detect clicks, mouse-movement etc, other software outside of the browser, but is still subject to arbitrary (but logical) validation rules.
Hi, John. I meant seconds. Another typo. Sorry. Sigh!