Are Viewable Impressions Cannibals At The Doorstep?

By
  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

stevegoldbergfixed"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 Steve Goldberg, senior adviser at EmpiricalMedia.

Earlier this month, panels at Advertising Week on viewable impressions revealed a subtle new point of resistance for adoption by sellers and aggregators.

Resistance is bad because viewable is good. But the rationale is worse because it contravenes business logic and facts at the same time.

In essence, the aggregators’ concern is that once only a portion of their inventory is tagged as viewable, the rest will be devalued and the net result will be lower overall revenues. In other words, viewable cannibalizes their overall business.

That's wrong on sentiment and wrong on facts.

First, let’s give a giant raspberry to the sentiment. Steve Jobs was renowned for ignoring “cannibalization naysayers.” He often said, "If you don't cannibalize yourself, someone else will,” which worked out pretty well for the iPod-iPhone decision. Nokia took the opposite approach with feature and smartphones, and now they are part of Microsoft. If Steve Jobs were alive he’d probably get an extra special kick out of that ending.

But aside from that, the assumption that viewable could cannibalize business is just shockingly ill-informed. Time and time again it feels like people in ad tech skipped ninth-grade algebra and went straight to getting MBAs or Ph.D.s in quantitative theory.

So, let’s look at facts vs. fancy. Using a cross sample of several SSPs, I’ll offer this hypothetical profile of a typical aggregator.

  • SSPs “see” 1 billion impressions a day from a mix of premium and long-tail websites. They sell roughly 200 million impressions a day to DSPs. They have historical viewable rates of 25 to 30%.
  • Since the sell-through rate is less than the viewable rate, we could stop the debate right here. Swapping the 200 million of mixed (viewable to nonviewable) impressions with only viewable ones would not cannibalize in the short term.

But let’s add some icing to the cupcake. Typically, from experience with vendors like RealVu, when identical Deal IDs were used against a pool of viewable impressions and a pool of mixed impressions, DSPs end up paying more (between 25% and 35% more) for the viewable-only pool. Simply put, DSPs “pay up” for superior inventory, and viewable impressions are,  for the most part, found on superior sites. So really our "typical" aggregator could sell 140 million viewable impressions and still have no revenue cannibalization.

But some might argue that in the longer term this still might cannibalize revenue. Yes, it could if the growth rate of display advertising was based on completely stupid decisions.

But it isn’t. Many agree our industry’s growth will come as dollars shift from other media. But that will not happen if buyers are purchasing unseeable ads. Make a better mousetrap, however, and you will grow rather than cannibalize.

Additionally, as shown above, our typical aggregator would have to double their sell-through before they maxed out the viewable pool. At 20% year-over-year growth, that still means six years until cannibalization is even probable.

Six years. Probably best to worry about something else for a while.

So now that aggregators' general cannibalization concerns have been debunked, what’s left? Well, several aggregators have shared the opinion that selling only viewable impressions might create a cannibalizing impact on some of their publishing partners.

But for good publishers, the logic above still applies. The number of unsold viewable impressions is high enough – relative to overall demand – where their yield will increase. And the nonviewable ad slots will still get the crappiest of crappy ads or a pass back.

Which just leaves long-tail publishers for whom viewable impressions could definitely be cannibalizing. Absolutely. And aggregators are concerned this would be bad for their own business.

But wait, seriously? A handful of fewer sites contributing to your billion impressions a day would not matter at all if the quality of your 900 million impressions went up. Remember quality scores? That actually helped your business.

Put another way, let’s say there are two aggregators: Bert and Ernie. Bert says he only sells views. He loses 100 sites that are totally marginalized by this development and, with them, 50 million daily impressions that were creating low yield. Instead, those sites go to Ernie who puts the inventory out to bid for DSPs.

This is where the marketplace and those great algorithms everybody refers to kick in. They create the logical network effect. Bert’s yield goes up (roughly 30%, according to tests), which leads happier publishers on to even better yield and so on. Meanwhile Ernie is dropping lower in yield, and soon, even his good pubs are calling Bert.

Who’s cannibalizing whom? Bert eats Ernie’s lunch, or maybe Ernie, too. Game over.

So, as we can see, worrying about viewable cannibalizing business is simply wrong-headed and wrong to boot. It’s nothing more than a proxy remark for, “I make money selling something of no value so I’ll stop when I have to vs. when I want to.”

And that’s an opportunity for a forward-thinking aggregator.

You can follow Steve Goldberg (@stevegol) and AdExchanger (@adexchanger) on Twitter.

  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

Email This Post Email This Post

By on at

8 Responses to “Are Viewable Impressions Cannibals At The Doorstep?”


  1. Olivia says:

    Hi Steve, great article. One point I'd like you to clarify: In your bullet point rationale you make an assumption that the 200m viewable impressions that replace the 200m mixed would be still be purchased by the demand side? Id have thought it quite likely the viewable 200m imps wouldnt meet the same targeting requirements (e.g. Audience, bid etc.) and so wouldn't be bought. Isnt the aggregators problem that there really is no easy way to predict what the sell through rate would change to when you factor in all the other variables as to why impressions are picked and bought by the DSP?

    On your last two sentences, I couldn't agree more. Simple economics explains why Viewability is a huge headache. Too many individuals make too much money to be incentivised to change for the good of the ecosystem. The good are too afraid to only sell viewable for fear the demand side won't respond, and therefore don't want to jump. The bad will keep up their side of the arms race and work out other means to keep selling crap. It's like prohibition all over again.

    • Hi Olivia -- thanks for the feedback. Agree with you, change is hard when the economic incentives are murky. No doubt. On your specific question, the answer is in the paragraph that follows...when you put in the same parameters (e.g. audience, geo, weighting, etc) you still can find enough viewable to "deliver" on the impression goal. Interestingly, since bids "have wiggle room" you get the delivery and a yield bump. But you certainly get the delivery. The liquidity needed to satisfy a VIEWABLE ONLY buy request is out there...we just have to want to find it.

  2. Hi Steve,

    Nice article.

    You are right, when all is said and done, in a free enterprise environment, delivering value wins.

    Publishers who deliver ads that are seen are bringing value to their advertisers.

    To use a sports analogy: Bert is batting 355 and Ernie is batting .055. Who is going to be playing in the majors?

    Your Bert and Ernie story is great.

    KISS, keep is short and sweet.

    • Thanks for the feedback, Ephraim. Bert is going to "The Show" but free enterprise takes a while to sort itself out. An interesting point is that two things (fear of the unknown and unnaturally injected capital) both slow up the development of free enterprise. Between the seller fear (mentioned above by Olivia) and the VC capital (Unnatural because of over-investment in the sector) we have some serious "drag" on Free Enterprise.

  3. Drew says:

    On of the biggest problems I have with viewable impressions is how it relates to last touch attribution. You can serve above the fold, which severely hinders your ability to perform as your inventory is shrunk to a very small level. Also a lot of the exchanges don't state whether ads are ATF at all, and the one's that do self select and incorrectly advertise themselves as ATF. That is why you can still only get ~70% viewability when serving purely ATF.

    The worst part is that you can pay higher CPMs to the exchanges to get the ATF ads, drive value, but the person who gets the credit... the last ad loaded on the page which is likely to have been the least viewable.

    Advertisers need to move away from last touch attribution, but also stop tactics like first touch spray and pray, cookie bombing for LTA and stopping companies buying every impression available on a page (another form of cookie bombing), it is going to be a long road. Targeting companies need to show value at all levels of the funnel - Reaching NEW prospects, seeing how many are actually visiting the site, and then seeing if those customers are actually converting (hopefully at a rate higher than the site conversion average). The worst thing is to be driving significant numbers of qualified prospects into the funnel, only to see a dogfight between 10 retargeters to get that last touch. You tell me which is providing true value?

    • Drew: These are good points. The basic problem is that last touch attribution can and does lead to inadvertent and intentional cookie shenanigans. Many attribution models still will show that an all viewable campaign has a higher ROI than a mixed campaign and that (clearly) means the attribution model is off. Clearly, they are flawed and the flaw is cookies -- many of which are dropped for non-viewed ads.

      The one thing that we've seen one company do is use viewable technology to guarantee the validity of the ATF designation. This designation -- which is done by pubs themselves - is extremely flawed (which is why some exchanges stay away from it). This is a benefit of "Viewable" versus ATF.

  4. Chris says:

    Really provocative piece. But don't even premium publishers have non-viewable/below-the-fold inventory to sell? Your Bert and Ernie example made it sound like each individual publisher is either all viewable or all non-viewable, but not mixed. This would make it an easy choice if I wanted to become a premium aggregator/SSP, but there aren't really sites that are all one all the other in real life. It's gotta be much easier to increase CPMs/demand for objectively higher quality inventory, so what marginal value is a viewable-only aggregator providing to the publisher? If I were a publisher I'd be looking for a partner to increase the yield on all my inventory, not just the low hanging fruit. Seems like it would create a temporary buzz at best while real value is eventually found elsewhere by smarter buyers, until market forces finally win out...if that ever happens :)

    • Hi Chris. Thanks for the comment. Great question. Two part answer...
      1- The Bert and Ernie are examples of aggregators (SSP) versus individual sites so it is more an example about sites in their aggregation that are either very viewable or very not viewable. And -- not surprisingly -- the gulf between those two designations is pretty broad. A lot of premium pubs are 70%+ viewable and an uncomfortably large portion of the rest are 20% viewable. So unlike what you wrote above, turns out many many sites are one or the other (I was surprised by this as well.)

      2- Having said that, I do not agree publishers should want to "increase yield on all inventory". That's exactly the problem. Increasing yield on all inventory (when you know 30% of it is unseen) is flawed and not good long term thinking. At best you are wasting time and money for both you and your advertisers. At worst you are liable to backslide into other similar practices like click fraud. You should not increase the yield on something that should have no yield. So, YieldEx (for example) is terrific...but using it on a smaller pool of viewable inventory is more terrific.

      Put another way, anything and everything that you do to be a better buyer or a better seller or a better vendor (everything) is better if you eliminate non-views from the equation.

Leave a Reply