Viewability Guarantees: Missing the Forest for the Trees

adrian-t-dataxu“Data Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Adrian Tompsett, Vice President of Business Development at DataXu.

It’s been a busy couple of months for the emerging viewability metric.

Here’s a small sample of headlines:

Many marketers are also jumping on the bandwagon to dictate that they only pay for viewable impressions. This appears to be an entirely rational move, right? If my ads are not seen by consumers, then I won’t pay for them. With this policy in place, my marketing spend by definition will be more effective. Or will it?

I’m going to make a provocative claim: You’re asking your media-delivery partners for the wrong thing.

This might sound odd coming from me. Earlier this year, I wrote an article in praise of incorporating viewability as one of the metrics to assess your media spend. That said, changing your approach to viewability without simultaneously addressing attribution will lead to bad ROI calculations, mistaken insights and, on top of that, it will create new margin opportunity for your media partners.

Don’t get me wrong: There’s little question that media sellers will be able to deliver exactly what these marketers are asking for, which is “guaranteed viewability.” That viewability will not, however, mean that marketers will improve the effectiveness of their digital media dollars, nor will they recognize that they’re inadvertently asking their media delivery partners for the wrong thing.

Let’s look at how your typical ad network would respond to this request. Today I charge my client a $2.00 CPM for the “quality” media they are purchasing. Once the marketer shines the light of viewability on this buy, she discovers that 40% of the impressions are served in-view. With this report in hand, the marketer walks in and says, “Ad network, I will only pay for viewable impressions.” But, Whac-A-Mole style, as we see one problem solved, another pops up to replace it.

How would the ad network respond?

With simple math.

The ad network responds with a proposal for “guaranteed viewability,” a campaign that will run at a $4.00 CPM, with the client only paying for those impressions validated as in-view. And therein lies the new arb opportunity. With some viewability experience and some minor campaign tweaks, an ad network can get that in-view percentage up – without too much difficulty – to, let’s say, 50% or 60%.  With some major tweaks, they could boost that number even higher. The net result here would be that the ad network actually improves its margins, but the marketer would not actually gain improved efficiency or effectiveness for its marketing dollar, which is what drove the move toward viewability in the first place. In this scenario, viewability just becomes the new CPA, and one step forward drives the industry ten steps back.

The Simple View

Imagine five impressions measured against viewability. Under the new regime, the advertiser says it will only pay for impressions 1, 3 and 5.






Now factor in last-viewed attribution modeling. Sounds like a win: The media buyer is incented to find viewable impressions. The marketer gets better viewability.






You know that can’t be right. An impression that wasn’t viewed can’t impact consumer behavior, can’t cause a conversion, right? Right. Impression No. 4 was not viewed and you didn’t pay for it. Yet you gave that vendor attribution credit when calculating post-view CPA.

Then why would you give it credit for one?

A Better Approach

One reason viewability is so attractive is because it’s such a simple concept. For any new construct to hold up and get widely adopted in our space, it must be easily distilled and understood. We have seen many of the complex attribution vendors make some progress, but it is a slow and painful slog. In my opinion, changing attribution models is hard and isn’t going to happen any time soon. Conceptually, I’m on board with smarter attribution models, but I’m also painfully aware of today’s operational realities and the fundamental misalignment among industry players that is preventing widespread adoption.

Instead, let me propose a baby step that can help agencies and marketers better assess the mix of media sellers and technology platforms on their plan. Run your attribution model the same way you always have, but with a twist. Stop automatically crediting the last impression before the user converted. Why? Because that model incentivizes your media partners into a game of lower-funnel cookie-bombing that has been proven extremely effective at winning last-view attribution…and totally ineffective at delivering incrementality and real-world marketing effectiveness.

Instead, credit the partner who served the last viewable impression before the user converted.  


I’m the first to admit that by no means is this a perfect attribution model. However, I do believe that the results of viewing campaign results through this lens will be incredibly illuminating. This perspective will separate the wheat from the chaff in your media plan, and will represent the first major step in the journey toward optimizing your digital marketing effectiveness.

This attribution filter changes the whole formula for how your media partners deliver value. In the old view-through game, better described as “last-served,” exploitation of the attribution model’s flaws is the winning formula. Victory in that game is driven by high frequency, low CPMs, tool-bar inventory, and a host of tried-and-true industry bad practices (e.g. “cookie-bombing”) that successfully place the last cookie into a user’s browser to win the attribution of a nonincremental conversion.

Victory in the new game will be about creating real, incremental value for the marketer. It’s time to stop talking about viewability as a standalone KPI or the newest performance-billing model, and instead use it as a new filter to make each digital marketing dollar much, much more effective.

Follow Adrian Tompsett (@atompsett) and AdExchanger (@adexchanger) on Twitter.


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. Mr Performance

    Sometimes I worry that too much effort is spent on finding a perfect pricing solution. 80% of the effort is trying to get that last 20% perfect.

    Say a man is trying to sell a used car. He is powerless to write the contract for anything other than “$500 for a Hyundai”. He will not comment on how much gas is in it, that the transition has been replaced with bundle of corn from an ad exchange. All his boss ever told him is sell the car, we have our technicians check it, and fib the rest.

    Despite the rust and dents you’re well aware by the “My D Student beat up your Honor Roll kid” bumper sticker that the previous owner loved Fox News, wore tinfoil hats and was a big buyer of gold bars that he stored the trunk.

    So you buy the car knowing you’ll also get $20k in gold bars and a rusted Hyundai with a little gas for $500 and you’ll most likely have to buy a razor for $1 to take the sticker off.

    Often pricing don’t necessarily explain the total value of the deal. I worry that as an industry we hurt ourselves focusing on having everything too precise.

    Spending $500 is $500. Did you like the results?

    Often vendors need to simplify things to have their whole and complete system work out for their own reporting, objectives and optimizations. Sales will chase the next new pricing thing if the other lemmings have a good quarter and for quick deals, cause havoc on their own back end.

    On the buy side, it’s best and easiest to know the ins and outs and just bake in “30% of that is crap, but it’s still a good deal”. Over pay on the crap knowing you’re sneaking some good things out overall.

    There never will be a perfect solution, don’t let that bother you so much you never get around to making money today.

  2. Hey Adrian, this is an excellent point and speaks to the need for integration between viewability and attribution technologies. While I imagine that your idea wouldn’t completely fix our attribution situation, it would totally improve it.

  3. If last-ad-loaded attribution is the bread and water model of attribution, last-ad-viewed is the bread and bottled water. Barely better at all. If an advertiser is going to make the effort to do better than last-ad-loaded, expend the extra two kilojoules of energy and use one of the many attribution models out there that gives you real insight.

  4. Great article Adrian, I agree that working with viewability metrics requires changes to attribution models, rather than just CPMs.

    The next real question for viewability seems to be; how can the demand side put it to use as an optimisation metric without simply reporting it back and adjusting CPMs accordingly?

    There is huge potential for online display to benefit from optimisation towards viewability as it could provide legitimacy to the view conversion, also getting the ball rolling for wider attribution discussions. The knock on effects of viewability are also far reaching; informing smarter creative and site design as well as combatting fraud. The bad news is that this can only start by getting the advertisers to buy in to viewability at enough scale to get the industry moving and that media buyers currently need custom data integrations to make optimised viewability a reality.

    However, once your impression level feed is in place, the good news is that you can make massive strides in campaign performance when measured against post-viewable conversions instead of post view conversions.

    By optimising towards post-viewable conversions across a handful of our more digitally mature advertisers, we found that even the most basic of manual optimisation techniques could increase the proportion of post-viewable conversions by between 60 and 80% compared to the start of the campaign.

    Good post-viewable CPAs and transparent pricing can help to encourage advertisers to switch away from the last view and, with the potential to optimise algorithmically towards post-viewable conversions quickly becoming a reality, I’m confident that the technology is both scalable and exciting enough to lead the charge against poor measurement and even poorer media buying decisions.

    And as you summed up, viewability shouldn’t be just a standalone KPI, rather a tool for unlocking more effective media. I would like to add that it’s happening already; viewability will soon be available via many larger buying platforms and early adopters are already optimising towards post-viewable conversions as part of standard attribution models. So I would say if you’ve lost faith in post-view, then it’s time you had a look at post-viewable.

  5. jeremy


    if you can get attribution correct; viewability is meaningless.

    oh by the way, it’s not that hard to get attribution right. hint: see incremental lift. it’s been around for awhile. i think it’s how they cured scurvy actually.

    consider the following two scenarios:
    A: serving an ad to someone moves them from being 80% likely to convert to now being 90% likely.
    B: serving an ad to someone moves them from being 10% likely to convert to now being 50% likely.

    Which is better for the advertiser? It’s obviously scenario B; yet 99% of the commercially available attribution models favor scenario A. Who cares about viewability when no one is willing to tackle the real problem?

  6. Jeremy – obviously that is what everone is trying to figure out with attribution modelling, the problem is actually making that calculation and proving causation, not correlation!

    in your example, it could just be that the type of person who visits the publisher your ads are on, is already 50% likely to convert, compared to the type of person who doesn’t visit that publisher and has a 10% conversion likelihood. In this case, viewability is key as the ad being seen at least has a chance of influencing the user where as an unseen ad on a highly correalative site is just a misdirection.

    Viewability may not take attribution modelling to a causation state BUT it will take correlative attribution modelling to the next level