The Industry Grapples With Mobile Viewability Growing Pains

mobileviewabilityMobile viewability is going through its awkward adolescent phase.

There’s no official standard yet – the Media Rating Council (MRC) is planning to circulate a document for public comment in the next few weeks – and vendors are still tinkering with their methods.

Which can lead to confusion around implementation, measurement-induced headaches and wonky reporting.

“Everyone is always looking for a source of truth in the digital space,” said Joe Barone, GroupM’s managing partner of digital ad ops and a member of the MRC’s Digital Committee. “But it’s easy to forget that under the surface there’s an iceberg. There’s a tremendous amount of work that goes into technical integration – and in any case, you’re never going to get two measurements that are exactly the same.”

Viewability solutions aren’t out of the box. A discussion last Monday on the adops subreddit around how to properly implement Integral Ad Science (IAS) tags on mobile epitomizes the problem.

As a Redditor put it to the group: “We’ve been having major issues getting IAS tracking to work properly for our mobile campaigns.” Turns out, that user wasn’t aware that, for in-app inventory, the IAS solution requires MRAID-compliant inventory.

When he/she was told, the response indicates what seems to be representative of a breakdown in communication between ad ops and verification vendors: “That’s the 1st I’ve heard of this. is this documented anywhere … ?”

Integral’s in-app viewability solution hinges on the use of MRAID, the Interactive Advertising Bureau’s standard for programmatic mobile rich media creative, in order to function properly. The solution is not yet accredited by the MRC, but Integral is seeking its stamp of approval. At the moment, Moat is the only vendor accredited for mobile viewability. That will certainly change when the MRC comes out with its official mobile viewability standard.

MRAID was originally developed by the IAB to make sure mobile rich media ads, which include elements like video and animation, could run successfully across all environments. When an MRAID tag is pinged, it responds with a binary answer – either “I am in-view” or “I am not in-view” – which acts as a signal to the piece of mobile rich media creative to start playing.

Integral hooks into that functionality through software development kits to measure viewability within in-app environments. A lot of publishers and most of the major SDKs, including MoPub, Google and Millennial Media, are MRAID-compliant, but there isn’t necessarily consistency in how the container is deployed.

“The overarching challenge with the MRAID standard is that it’s just a PDF document on the IAB’s website, and anyone can download that and build an MRAID SDK,” said Jason Cooper, general manager of mobile at Integral Ad Science. “But you would be building it based on your interpretation of the guidelines.”

That could result in “idiosyncrasies and slightly different behaviors” because “different developers have different interpretations of the document,” Cooper said.

The IAB Tech Lab did release a quality assurance testing tool for MRAID in January, but there’s no requirement to use it.

Then there’s the added complication that the current iteration of MRAID (MRAID 2) wasn’t developed to support viewability as defined by the MRC –  50% of pixels in view for one second for display and two seconds for video. Version 3 of MRAID, which is in the process of being finalized, will support those time and area components, said Cooper, a member of the IAB’s MRAID Working Group.

From GroupM’s perspective, though, MRAID as its stands isn’t “acceptable” for in-app viewability because it doesn’t use “active technology” to determine whether an ad is viewable in the browser window of a device, Barone said. Rather, it just relies on a signal being passed by the publisher.

It’s also true that not all mobile inventory is coded in MRAID, which “very possibly makes it blind in certain cases if there are certain apps or publishers not using it,” said Barone.

Another issue: SDKs, although they’re the preferred method, present their own complication.

“The code needs to be small and efficient so it doesn’t bloat the app and take up too much space on mobile,” said Tom Grey, a product manager at PubMatic. “Clearly, app developers are not keen to add more weight to their apps, so very few apps have the code embedded.”

And then there’s what Grey called “self-selection” going on.

“Some app developers are trying to game the system and will not install code to track viewability,” he said. “They have no incentive to police themselves since they are pursuing fraud.”

Integral’s offering also tries to determine whether a user is in-app or on the mobile web, Cooper said, but it remains an “inexact science.” Although the IAS technology does what it can to make that distinction, in the cases where it can’t – if the content is displayed in-app using a webview, for example – the measurement can default to the mobile web, which impacts reporting.

The playing field is also constantly shifting, sometimes subtly, as when an SDK like MoPub rolls out an update, which happens roughly every six to eight weeks, and sometimes tectonically, as when Apple makes one of its more significant annual updates.

“There’s no public conversation about this,” Cooper said. “They’re Apple so they do it and then the industry has to scramble for a few weeks to try and overcome it.”

All of that leaves publishers in a bit of a pickle. The industry is turning its attention to mobile, but properly forecasting and optimizing for viewability still feels “impossible,” said Brian Fitzgerald, CEO of Evolve Media, which happens to use Moat as its viewability vendor.

“Each client picks their own vendor [and] if it doesn’t match with yours, for example IAS vs. Moat, then there will certainly be discrepancies,” Fitzgerald said. “We can’t know what those discrepancies will be at the time of forecasting.”

To try and address that concern, GroupM, for one, has whittled its preferred verification vendor list down to three, one of which is Integral Ad Science, Barone said.

But the larger issue in the ecosystem is that clients, agencies and, in particular, vendors, “aren’t incented to try and fix the problems” or solve for viewability, Fitzgerald said.

“If there is one unified viewability vendor of record, like Nielsen for TV, then there is an even playing field and all players can understand the rules of the game; we are all judged equally,” he said. “Verification vendors have a business because there is a problem ‘they can help detect.’ Verification vendors stay in business, grow and become profitable by finding problems.”

PubMatic President Kirk McDonald has heard a similar sentiment expressed by other publishers. Advertisers and agencies make their demands and the burden is generally on the publisher to make good on any discrepancies.

“The one writing the check to cover that gap in one way, shape or form is the publisher,” McDonald said. “At this point, a lot of publishers do seem to feel that there is more promise than deliver when it comes to vendors.”

In essence, even some of the vocabulary is broken, said Barone.

“In my mind, we’ve gotten to the point where we need a different word other than ‘discrepancy’ because a discrepancy is defined as two people counting the same thing and coming up with different answers,” he said. “But in this business, when buyers and sellers talk about discrepancies, what they’re actually saying is, ‘My guy is counting one way and your guy is counting another way.’”

But all of the headaches are part of a necessary growing process, said Barone, even if that process is often painful or uncomfortable.

“We’re changing the way the industry has been working for the last 15 years, so it’s a big deal and hopefully it’s a one-time disruption,” he said. “Moving away from the ad server count is a major sea change. It’s worth the time and effort to get it right.”


  1. DetroitViewability

    This article is a great discussion starter but the dots need to be connected around Moat. While "...Moat is the only vendor accredited for mobile viewability", Moat requires an app to install an SDK. The article mentions the limitation but never connects this with the limitation of Moat. None of the current third party viewability vendor solutions are complete solutions.

  2. Lee Cassingham

    Until viewability providers sit around a table and agree to use one method for tracking ad viewability accreditation providers are going to have a hard time figuring out what’s what. At Viewex we built the first mobile and desktop compatible viewability product I know of and tested it with AdQuota -Mobile Ad network sites (recently acquired by Opera MediaWorks) so i'm familiar with the challenges faced in this area. Fortunately, our technology was able to see 99% of ads and developed a S2S hack for those which presented initial restrictions.

    Notably, viewability data collection/accuracy varies depending on where in stack you collect the data due to the number of technical restrictions in the stack, inherent to different implementation methods and technologies between the creative and the page.

    Brands for instance won’t get the same level of accuracy available to publishers. Accuracy is relative to where in the chain the data is collected so it makes sense for viewability verification to be reassesed and awarded on a product/implementation level not company wide.

    Another point to take into consideration is Publisher view time tracking discrepancies, when publisher loads their unit, a few seconds later the creative loads but with many iframes between the publisher and the creative it's impossible for a publisher side tracking solution to know how long it took for the creative actually loads and so without a publisher Server 2 Server/API integration to ALL buy side platforms it’s difficult for publishers to accurately track and report on view times. Similarly when brand and publisher tracking technologies start sharing data it will then enable the brand to see how much of their potential view time has been wasted due to Ad tech latency.

    A viewed ad is classified as 50% in view for one consecutive second and with 300ms of latency between the creative and the page that’s 300 seconds of lost view time per 1 thousand ads served. We track the click/time bell curve measuring how many seconds after an ad loads that it is clicked when plotted on a graph this forms a bell curve with the median click time per unit on a page. We know and can predict number of user clicks lost due to latency. Adtech latency affects brands by lowering the number of confirmed viewed ads and revenue for publishers.

    Latency mitigation within adtech is key to ensuring creatives load faster. One solution to this is for ad platforms to make more decisions clientside reducing the number of processes in the server. Naturally this opens up to sharing some IP but offers everyone faster advert load times and better level of customer service, performance and revenue for all.

    There are plenty of problems in adtech which need to be addressed. It would be interesting to recieve some non biased feedback in relation to what everyone else is seeing.


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>