With Two Sides To The Viewability Equation, Discrepancies Persist

davemarquardThe Sell Sider” is a column written for the sell side of the digital media community.

Today’s column is written by Dave Marquard, director of product management, publisher products, at Integral Ad Science.

Now that the MRC has defined that for a digital display ad to be viewable, half of the creative must be in view for at least one second, advertisers want to transact their media buys on the metric. Consumers, they argue, won’t convert if they never see an ad.

Publishers are also interested in transacting based on viewability. Advertisers now demand it, and many publishers hope advertisers will pay more for impressions that are in view.

Yet many wonder why discrepancies occur when the MRC definition seems so straightforward. Different vendors for advertisers and publishers often report different viewability rates. There are also time and location components to viewability: Advertisers are more effective at measuring the time aspect, while publishers are better suited to measuring the location of an ad.

Discrepancies occur because viewability is often measured from one of two places: the advertiser’s ad server or the publisher’s ad server. Each implementation confronts limitations that affect measurement, which in turn, affect the reported viewability and measurement rates.

When viewability is measured on the buy side, the viewability solution sits with the advertiser’s ad server. Since the ad server is responsible for serving each and every creative, it’s very easy to know exactly when to start the viewability clock and determine when the creative is rendered for at least one full second.

But due to ad environment challenges, like unfriendly, cross-domain iFrames, advertisers can’t measure every ad unit in every environment, which means some percentage of ad impressions is simply unmeasurable. If a vendor reports that 60% of the ads were in view, with a 70% measured rate, what value do the remaining 30% have? The 70% rate must be extrapolated for the entire campaign, which can lead to confusion for both sides.

Meanwhile, the publisher’s viewability solution is also integrated with its ad server. Since publishers are measuring fully owned inventory and not dealing with foreign ad environments, they have no difficulty determining whether the location of an ad unit is in view. Put another way, publishers can reliably determine the location of all ad units throughout their web properties virtually 100% of the time.

What they can’t do, however, is measure the time an ad creative loads with the same level of certainty of advertiser-side technologies. Publishers know the exact moment an ad container loads, but that’s not the same as the actual creative, where rich media and video ads often take a longer time to render or start to play. I’ve seen discrepancies between times reach up to 20%. While 200 milliseconds may seem insignificant, when the goal is merely one second in view, it’s a huge difference, and one that will certainly have an impact on the overall viewability count.

Clearly, the current dual system of measuring viewability leaves too much room for disagreement and opens the door to mistrust. What’s needed is a measurement number that both sides can trust, and until that happens, transacting on viewability will be fraught with contention.

Follow Integral Ad Science (@Integralads) and AdExchanger (@adexchanger) on Twitter.

5 Comments

  1. Dave, can you clarify this section: If a vendor reports that 60% of the ads were in view, with a 70% measured rate, what value do the remaining 30% have? The 70% rate must be extrapolated for the entire campaign, which can lead to confusion for both sides.

    Do you mean that the 60% in-view rate should be extrapolated across the entire campaign, across the 70% measured and also the 30% unmeasured? I'm not sure what you mean otherwise.

    Reply
    • Correct, that's what I meant. Assuming the campaign has a non-trivial measured rate, you'll want to extrapolate the in-view rate from the measured portion over 100% of the delivered impressions.

      Reply
  2. Part of the issue is the extrapolation part. If you can only measure 70%, then that's all the campaign can be based on.

    There's usually a clear reason why the other 30% can't be measured:

    1) There's a tech issue - this should be fixed by replacing whatever vendor is serving those impressions. There are enough alternatives today that there's no excuse to get this done.

    2) The ads are just not in a spot that can be measured - there's no getting around this and the placement is effectively worthless.

    Extrapolating means randomly attributing value to impressions that cant be measured and probably dont have any real value. I've never seen a situation where the unmeasurable impressions are truly just random (and why would that ever be the case)? Things like these are what erode trust. If we want to stick to a standard, both advertisers and publishers should agree to only use actually measured and verified impressions without any extra guesstimates. This will also increase scarcity and raise rates to somewhat decent levels for pubs while giving advertisers verified value.

    Reply
    • Mani, thanks for the response.

      1) A publisher is not able to "replace vendors". The campaign runs on who the agency uses, end of story.

      2) Video and mobile creatives have difficulty being measured. Are those placements worthless? What about large ad units that are not 50% in view? Or creatives served programmatically in cross-domain iframes?

      What happens when only 50% of a campaign's impressions are measured? Or 20%? Is this the fault of the publisher? The viewability vendor? Neither? How do you tell?

      Reply
      • Mani, on your point on extrapolation: extrapolation certainly should not be *random*. In the example from the article, if you sampled ("measured") 70% of the impressions in a campaign and found there to be an in-view rate of 60%, you'd extrapolate that 60% of the total campaign was in-view. In other words, we'd expect the remaining 30% of impressions that we did not observe to also show a 60% in-view rate.

        The reason why < 100% of ads can be measured is largely attributed to browser and environment limitations. Measurements from inside cross domain iframes -- which often happens in programmatic ads and/or ads served from a buy side ad server -- are harder to do than those done from directly on a publisher's site. Measurement in a cross domain scenario often requires the presence of Flash, which is why you see mobile browser measurement rates being far lower than desktop.

        Varying measurement rates generally aren't anyone's "fault"; rather they are the consequence of the fact that ad campaigns will be served to many different environments and to many different browsers.

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>