This article originally appeared in AdAge on November 29, 2012.
One the hardest decisions to make as a product manager is whether a feature is really ready to launch. Which brings me to the ongoing drama around the IAB’s push to for display advertising to use “viewable impressions” to replace impressions as the currency for online display advertising. In short, viewable impressions are not ready to be launched and we need an alternative roadmap for the industry. The IAB, to its credit, recently admitted as much, after tests of viewability measurement resulted in widely disparate results. However, the intention remains to drive towards a currency, which is in my opinion is the wrong goal.
The problem that we’re trying to solve is one of microeconomics. The display advertising world has too much supply – one might say, infinite supply – as compared to large, but finite demand. Sellers of ad inventory (aka publishers) can artificially create as much supply as they choose by simply adding more ad slots to each page of content. These new ad slots may be of unquestionably poor value to the advertiser given their position “below the fold” or, in some cases, they may only be viewable with extensive scrolling on behalf of the user. This situation devalues the inventory of the whole market, and especially hurts premium, high-quality publishers whose inventory is being compared to the inventory of low-quaity sites and/or sites with outright fraud.
The solution to this problem is to enable buyers to better differentiate between and transact based on the quality of the ad slots, such that sites that serve non-visible ads are punished and those with visible ads are rewarded. No argument from anyone so far. The proponents of the new viewable impressions currency have made a mom-and-apple-pie argument: how can you charge for impressions that can’t be seen? We’re just bringing the displays business into conformity with other media like TV. Hogwash. TV commercials play whether you’re in front of the set or in the bathroom, or TiVoing through at 30 miles-per-hour. Most print ads are never seen. No one has any idea whether people look at billboards. (Note, I’m using a bit of hyperbole to make a point, measurement nerds please don’t blast me in the comments about all the obscure techniques used throughout the industry. Thanks.)
The problem comes in the practical applications of such a change. The IAB and Media Ratings Council (MRC) have advocated making viewable impressions a new currency for online advertising. But viewable impressions lack some of the key attributes of a currency – they are not standardized, and they cannot be measured consistently between campaigns.
Let’s discuss what it means to be a “currency”. Every media business has a source of measurement truth that both buyers and sellers agree is the billable record for the media. In the US television market, the currency is the Nielsen ratings. In radio, it is Arbitron. In search, it is Google’s record of clicks. In online display in the US, it is currently the advertiser-recorded impressions through whichever ad server they’re using e.g. DFA or Atlas. (Interestingly, in the rest of the world the online display currency is the publisher-recorded impressions.) Whatever the currency, the important points are that a) everyone in the market agrees on the currency; b) the numbers must be consistently right since they will be used for billions of dollars of transactions. It should also be noted that changing a currency is a very big deal.
So, what’s wrong with viewable impressions? Simply put, they can’t be measured consistently enough to serve as a currency. Tech talk ahead…
A standard impression can be measured by either the delivery of a “302 redirect” from an ad server or the delivery of a beacon (pixel) within the ad from the user’s browser. This standard, set by the IAB ten years ago, is remarkably easy to understand and implement. Much of the IAB’s industry efforts have been focused on reducing the potential discrepancies between impression counts on different systems and bringing them to within a 10% threshold. The reason it is possible to bring the counts to within such fine tolerances is that the causes for miscounting or divergent counting are limited and unlikely to vary much between systems. For example, if a publisher system records an impression when delivering the ad tag to the browser, then the advertiser system does so immediately thereafter, discrepancies are going to be limited to the falloff in traffic between two consecutive http requests with very light payload. Sure, discrepancies still happen, but they are mostly based on trafficking errors, browser latency, or differences in fraud detection.
To measure viewable impressions, on the other hand, you need to not only deliver the ad, but also run client-side javascript to evaluate the position of the browser on the page and make a determination of viewablity. There are three ways in which this is an order-of-magnitude different problem than standard impressions: 1) instead of counting on the call for a beacon, you’re dependent on code being delivered and executed on the browser, making both network latency and browser execution latency a much bigger problem; 2) the measurement is dependent on non-trivial logic, which is going to differ significantly between ad providers and browser environments making discrepancies a matter of debate about which system is “right”; 3) the viewable nature of the ad is a fluid construct since over the course of an impression the ad may move in to and out of viewability over the course of the page view, thus adding to both the latency and subjective measurement problems.
Also, none of this works inside Iframes and roughly 40% of all ad impressions are served into IFrames. The IAB is proposing SafeFrame, an alternative to IFrames, but this means that effective deployment of the new currency is dependent on everyone in the industry retagging. Vendors claim to be able to “break-out” of IFrames but these techniques are usually based on browser hacks that are not near 100% reliable.
Also, this standard doesn’t easily work for video ads, native ads, or most mobile ads. These are the three fastest growing sectors of “display”.
Also, viewabiltiy is not a feasible currency in programmatic/RTB markets since realistically the impression can only be evaluated after it is auctioned and served. RTB is the fastest growing method of purchasing display.
Also, the ability to measure viewabliity is threatened by patent lawsuits from ComScore. If those suits are successful then the industry will have handed the currency to one company without extracting any value from that company.
Also, viewability doesn’t take into account traffic generated by bots, which some estimates put at over 20% of online traffic and which could be a bigger source of waste and fraud than below-the-fold impressions.
Also, the international community has made no movement towards adopting this standard, making the currency and technology potentially quite different in the US and the rest of the world.
Also, because the viewability of a campaign is difficult to estimate in advance, publishers will be forced to over-allocate inventory to hit viewable impression budgets, resulting in waste and confusion.
Also, from tests thus far, different vendors are seeing radically different results for viewabliity on the same campaigns.
So here’s my modest proposal. Let’s all agree that viewability AND bot detection are vitally important to the success of the display ecosystem. Let’s establish IAB standards for measuring these, and MRC certification for doing so. But, let’s revise our goal from replacing the currency until we have a solid track record or consistent and agreed-upon measurement over a meaningful period of real-world activity. Moving ahead with a currency too quickly will be a huge setback to display and will reinforce the perception that it is hard to measure and subject to technology for its own sake.
Instead, let’s update the IAB standard Terms & Conditions (“Ts & Cs”) used throughout the industry to require all line items to preform with a minimum X% human-viewability and no more than X% in cross-domain IFrames as measured by any MRC-audited system and require publishers to offer make-goods on those line-items that fall below that threshold. Just as with impressions, in cases where publisher- and advertiser viewablity metrics have large discrepancies, both parties use reasonable efforts to uncover the sources of the discrepancies and come to reasonable solutions to reconcile the discrepancies. If we consistently have these large discrepancies then we know we have a deeper problem with the technology, if we don’t then we’ve created a de facto currency, but with enough fault tolerance to deal with the uncertainty of measurement.
This Ts&Cs approach will make viewability a vital part of the media negotiation but will acknowledge that the technology for measurement is far from exact and too unreliable to be used as a currency at this time. The stated goals of viewability will still be accomplished: publishers will redesign sites to reduce poor-viewable ad slots, advertisers will use view ability as a metric for negotiation, and bad actors will find their inventory increasingly de-valued by the market, resulting in reduced overall supply and increased quality.
I look forward to your feedback.