Viewability Is Not A Tax On The Ecosystem

andrewshebbeare“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 by Andrew Shebbeare, Founding Partner & Chief Strategist at Essence Digital.

Ad viewability is technically and logistically bewildering. If you need proof, read Jeremy Stanley's excellent two-part overview of the space on this site.  Today though I'd like to set aside the discussion about nested cross-domain iframes, how many seconds are enough, and so on.

Instead I want to explain why viewability is important, and why it will eventually be good for everyone in the advertising value chain. I firmly believe that 1 million viewable impressions are worth a great deal more than 2 million impressions of which 50% get seen on average. That might sound strange, but it's based on my real-world experience of the way brands make decisions.

Sure the tech is confusing and it's hard to know whose iframe-busting miracle solution to buy into, but I've heard many other nasty things said about ad viewability lately. Things like: "A tax on the ecosystem" or "Just another way for brands to squeeze publishers," even "Another metric, when we have too many already." I disagree.

Let's start with the uncontroversial. The point of advertising is to make people think or act differently. That happens when we get the right ads in front of the right people at the right time. If we do, we expect to see measurable change. Based on that measurement we refine our message and our investment strategy.

Obviously ads need to be seen for them to work. Assuming they are good ads, those that get seen more often will work better. Everyone agrees: Viewability is a good thing. What many don't agree on is the value of its specific measurement. Critics say the metrics are redundant. In the end, the argument goes, effective ads will deliver results, and budgets will follow. The extent to which ads get seen is wrapped up in that effectiveness along with a bunch of other dimensions like targeting, creative, and the rest. We don't need to grease the palm of the ad viewability man, when we have the Invisible Hand to guide us.

This would be fine indeed if measurement were easy and quick. It's not. Today most digital marketers rely on metrics that include some sort of statistical error. Once you accept that last-click isn't telling you the whole truth, you're suddenly in a probabilistic measurement world and you need a model. The success of any model is predicated on the data you feed it. Algorithmic attribution, Attitudinal surveys, Incrementality experiments, Panel-based measures – these all rely on adequate samples and variables of adequate power to help us make decisions. Until those models have given us an answer with a degree of statistical confidence we're happy with, we continue to invest, aware that some of our budget is less well spent. We all know the quote attributed to John Wanamaker: "I know half of my marketing budget is wasted, I just don't know which half".

Some of the most valuable data, then, is that which accelerates measurement and with it investment optimization. If I can tell which half is wasted in half the time, that's going transform my business. The very best such data is a salient variable that helps you split wheat from chaff more accurately and quickly. Far from being inundated, marketers are endlessly seeking those variables among a soup of video completion rates, rich media engagement rates, search uplift, and small survey samples. What could be more salient than an indication of what ads were seen at all? Here's a datapoint that is close to free and highly predictive. I'd think very hard before turning it down to save 1% on my media plan.

Unseen impressions are noise. They dilute the impact of our advertising and the power of our models. Worse, they have a real cost – measured in bandwidth and CPU, page load times, privacy and data collection/leakage risk, data storage for everyone and so on. This cost is paid by brands. Brands who measure their return on investment by dividing results by spend. If we can withdraw those impressions, we'll get more impact for the same budget. Thus, fifty viewable impressions are worth more to an advertiser than 100 impressions of which 50% are seen. Not only do they deliver the same lift, you'll detect it faster, optimize sooner and spend less money on ad serving.

Less waste is good. Good for brands and consumers, but in the long run good for everyone in our industry. Smart advertisers will pay more for placements that deliver more. If we can pay more for fewer ads, that's less consumer intrusion to fund the same content. That means more receptive audiences, better results for advertisers, more elegant content earning the same or greater yield, and fewer good ads having to cross-subsidize bad ones.

This might sound idealistic, but it isn't entirely without precedent. Most brand plans still get reviewed at a placement level, and position above the fold is hugely important to marketers. Impressions that are guaranteed (position, pacing, frequency) trade for a disproportionately high premium for all the same reasons.

"Ok, but I've got a finely-tuned last click based DR machine, so what do I care?" If you really place no value on people seeing your ads without clicking on them, then all this is moot. The truth is though, there aren't many cases where clickers are 100% incremental or view-based conversions are 100% incidental.  Your measurement plan should include testing your Retargeting for incrementality almost above any other channel – and that means probabilistic models, and understanding the value of ad exposure. I'm not sure there is such a thing as a finely-tuned last click based DR machine.

One last hitch. If you want a really powerful model, you're going to need your viewability data at the impression level.  We can prove the relationship between ad viewability and brand lift; it shows up in the numbers just as common sense would have it. But because viewability typically gets reported back in aggregates, we have a hard time knowing which individual impressions were actually seen and working that data into our models. This is likely to be a challenge until the ad serving companies catch up to the viewability leaders, or the viewability leaders start sharing impression-level data in a convenient form. In the end, my money is on the ad servers; I'm hoping to see "seconds in-view" showing up in impression logs.

Getting to viewability nirvana relies on suppliers and buyers of ad inventory alike asking the question and offering up the data. Site designers and brands will need to ask these questions of everyone in between to get viewability on the agenda. Granted - in the near-term this probably might include some pain – particularly for those who are compensated on a CPM-like basis. In the long run, it is a step toward better media experiences, at the expense of the cookie stuffers, the data leeches and the ad stackers.

Follow Essence Digital (@essencedigital) and AdExchanger (@adexchanger) on Twitter.


  1. This article misses the reality of the current environment, specifically in the exchange RTB space. There is currently no accurate measure of viewability as confirmed by the IAB

    The default attempt to control viewability is by restriction to above the fold inventory, however this is a poor solution. ATF is self- and under-reported by Pubs in RTB – most pubs don’t report it at all, and even those that due, most exchanges don’t verify. So a very large portion of exchange inventory is not classified on way or the other. Restriction to AFT only in RTB can throw out up to 80% of bid volume and much is this inventory can be ATF but isn't classified.

    The other key point that this article misses is that viewability is more of an issue for the branding and the CPM model. When the model is CPM, clearly advertisers don't want to pay for impressions that are not viewable. But when the model is CPA, as long as the performance is above goal, viewability shouldn't matter. A current solution for DR is to run on a CPA pricing model where the advertiser is not paying for any impressions, viewable or not.

  2. Hey Arjun,

    Thanks for your comments. I absolutely agree that definitions and technology have some way to go.. in fact that's why I started out with the preface. But let's assume they DO get sorted out.. my point is only that they are important.

    As for your comment about DR vs Brand, as I wrote in the article - if your entire path to conversion is exaclty one click long, then fair enough; let your publishers eat what they kill and forget about viewability. On the other hand if you have place any sort of value on exposure in driving conversion, or on how multiple engagements contribute to a sale, then viewability is an input with enormous value. More value for Brand than DR? Well sure. But for most big smart advertisers the interplay of the two is important.

    For what it's worth, the industry may need to reach standardisation for viewability to sit comfortably on IOs, but you can still augment your measurement model and shorten you measurement cycle today.

    • Andrew, you raise a good point that the viewability issue becomes an attribution issue for CPA, and that remains a key challenge. Just as on CPM, where advertisers don't want to pay for impressions that might not be viewable, on CPA, advertisers may not want to attribute conversions credit for impressions that may not be viewable. And clearly the last click model is becoming widely accepted as broken. The flip side is that with the rise of FBX, it's become much easier to slip in the last view via retargeting, given how often people go on FB, and thus even the last view attribution model can be gamed. That said, determining accurate attribution, while challenging, is still currently easier and more realistic than a solution to viewability.

  3. Alejandro

    Great article! For all the reasons you cite above a) viewability metrics will be ready for most marketers after the technical issues are fixed and after we can see data at the impression level and b) pubs should jump on board because these metrics will ultimately benefit content-producers that add value (as opposed to the cookie bombers).

  4. Andrew: This is a good article full of well thought thru and correct reasons why viewable is important. However, I think there is a key point towards the end which you have minimized and which represents a great challenge for this important metric. The issue is impression level predictability -- be it in the hands of the ad servers or vendors.

    A critical problem is that, viewable impressions by their very nature are not predictable. Being viewable is a function of place, user action, etc. This multi-varient metric immediately starts to break down under the after-the-fact "campaign model" deployed by vendors. And while they, and ad servers can (and will) ultimately develop mapping technology to create highly viewable "zones" based on previous user behavior, the results will still create challenges in the marketplace.

    The first challenge comes from the very thing you correctly love about viewable (at the start of your article.) Specifically, unlike overlays for a particular demo or behavior, viewable is binary. It is or it isn't. Publishers and agencies are okay (in the CPM world) charging for and paying for the 10% of impressions that miss the target as defined by a post campaign report. In fact, all sides will claim that is a good result. But for the 10% of impressions that are not viewable? Different story to be sure.

    And then the second challenge -- one that is truly daunting (albeit at this time somewhat speculative.) Experienced ad ops professionals know that their is an unnerving inverse correlation between demographic delivery and CTR and that this reverse correlation becomes more profound in the presence of a high percentage of CPA campaigns. Our research at Acceleration has shown that the reason for this is the CPA campaigns are being optimized towards CTR which comes (as you point out) from being viewable and that the satisfaction of many targeting overlays comes from impressions that are not viewable. Therefore, the demand of a brand advertiser to have only viewable impressions is a good thing, as you pointed out. But the level of disruption for Yield Managers and Operational leaders is about to get profoundly more powerful than most have predicted. The impacts will be on the roles and pricing of CPM, CPA and remnant inventory. It will affect major client/pub negotiations, price modeling, relationships with aggregators, etc. At Acceleration we have worked with a number of pubs around the world on these issues and we (like you) see viewable as the most meaningful development in years. But we believe that we are only seeing the tip of the iceberg on impacts right now. Anyway, thanks for a good article and conversation starter.

  5. Hey Steve,

    Thanks for your points - really interesting perspectives there.

    I actually started to run into some of these questions when I drafted this column. They made me brain hurt and my writing tortuous!

    This definitely isn't one of those problems with a neat beginning, middle and end, but I'm hoping, maybe idealistically, that if the demand is there the knots will figure themselves out.


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=""> <strike> <strong>