New Challenge for Marketers: Preventing Non-Viewable Ads From Getting Credit

Ask a random digital ad buyer how she feels about the shift to a Viewable Impression standard and you’re likely to get an upbeat response. But she may not realize the efficiency and cost savings that come from buying only in-view ads can mask a significant problem on the reporting side.

The issue is this: Even if an advertiser is paying only for in-view impressions, those junk ad placements can continue to fire view-through cookies, earning credit for the ad in post-conversion attribution. In a sense, it’s no different from the display ad view-through debate that has been raging for years (Did the user really see my ad?). But with viewable impression tracking, buyers may not realize the below-the-fold ad that was never seen or paid for could still generate a “false positive” in their attribution models.

C3 Metrics talks about this in a white paper out today. COO Jeff Greenfield writes, “Just because brands and agencies might not pay for an unviewed impression doesn’t mean that cookies won’t be set, and when cookies are set, display ads not actually displayed will receive credit. Thus, expect that the post-viewability world will focus on a new question: how do we ignore unviewable impressions and assign credit only to viewable impressions?”

Wells Fargo Securities analyst states in the paper, “Unviewable impressions fundamentally distort ROI analysis…taking credit for consumer actions, despite never providing opportunity to be viewed.”

But wait, you say, isn’t this already being solved by the availability of “pre-bid” data feeds from ad verification companies? In theory such feeds could prevent advertisers from ever bidding on inventory unlikely to be in view. No-bid-placed means no-ad-served, and no-ad-served means no-cookie-fired. Right?

Not exactly, according to Greenfield, who says there will continue to be a significant number of non-viewable impressions. He tells AdExchanger, “All you’re looking at with ‘pre-bid’ is what just happened 30 seconds ago” and the accumulation of impression factors including publisher, ad type, page position, and so on… At least over the next two years there’s going to be a lot of unviewable inventory out there.” (More on ‘pre-bid’ verification)

It’s a complicated problem, since agencies are working with comScore showing tens of millions of daily impressions with cookie IDs  — and trying to map those impressions to conversions. To solve it, ad viewability must be baked into the ad serving and tracking system – something that (no surprise) C3 says it can address.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Matt Anthony

    The focus on technology here is smoke-screening a larger issue. Simply, the vast majority of the attribution engines out there today are built on less-than-acceptable algorithms and mathematics, which is why this is such a dire tech need.

    An unviewed ad, by its very nature, has no causal/incremental effect, so a well-built statistical analysis should find that to be the case. Position and assumption based models (“funnel” models, manual fractional models, etc) are incapable of detecting this because the values that they assign to ads are done by arbitrarily assumed models, rather than from models derived and fitted to data to maximize explanatory power.

    Is having a flag or indicator for whether an ad is viewed or not a potentially valuable data point? Sure! But anyone who thinks that’s the only thing standing in the way of the current fleet of attribution models with respect to accurate measurement needs to bone up on their math and stats a bit more.

  2. Adam Smith

    Publishers using async ad calls and only firing ad requests when the ad unit is visible will eliminate this problem. Visibility in ad request needs to be handled client side, prior to the ad request being made. Any technology based on server side visibility metrics are b.s. .. my 2 cents.

  3. Greg Hills

    The only way to draw a causal link between exposure and conversion is an A:B study.

    Agree with Matt: By definition, a non-viewed impression do not generate sales. Attribution based on manual assumptions are dumb. However, statistical regression models are still based on assumptions, albeit complicated ones, which are non-scientific.

  4. I’m glad to see more awareness around this issue, and always enjoy the spirited debates in the reply section. We’re starting to see requests from brands and agencies to exclude out-of-view impressions from the attribution analysis and are working with AdSafe and MediaMind to address this growing need.

    While there is a clear need to exclude out-of-view impressions, I hope / believe it will decrease over time as the bad-actors (publishers and ad networks) are forced to clean up their act. As with ad verification, the more transparency there is, the less prevalent the problem will become.

  5. Sharon Park

    I want to bring attention to viewability technology. Only ad networks that have DIRECT relationships with publishers can accurately detect viewability. Due to iframing of ad units, many third party verifiers are not able to properly detect viewability. Thus, reports on viewability by third parties like Comscore or Double Verify may not provide significant technology to accurately detect and report viewed impressions.

  6. Sharon – I think you raise a great point. There needs to be more discussion around how 3rd parties are going to “solve” the cross domain iframe issue without direct relationships with publishers.