The PostView: Why The vGRP Might Not Be The Silver Bullet For Display's Metric Problem

pv1

The PostView is a new column written by senior execs working in the European online advertising industry.

It is fair to say there has been some fanfare recently around comScore’s attempts to bring digital metrics into the 21st Century. Or is it, as vocal detractors have been arguing, actually regressing the industry back to 19th century metrics and dumbing down the channel? It has been heralded by some as the key to unlocking valuable brand budget – while others remain sceptical. Regardless of what side of the fence you sit on, you can’t escape the fact that this new measurement standard has the potential to disrupt the industry.

There are two key questions around this debate. First, how does the industry change the way it packages impressions in view and monetise them differently? And secondly, how does a different vGRP change the way we trade?

Monetising Viewable Impressions

It is going to be hard to be able to price and sell only viewable impressions. The current definition of an in-view ad from comScore is 1 second in view (in the UK). So whilst retrospectively, you could report on it, how would publishers be able to forecast and sell it? It could be argued that publishers are already in effect selling unknown quantities of impressions, but modelling equations have become so precise, it has become easy to predict traffic flows and consumer consumption, ie. how many impressions can I sell this month? Will it be easy to predict volumes of viewed impressions? This could be trickier to execute. It seems more susceptible to erratic consumer behaviour and how sticky any given piece of content is.

With regards to RTB, it simply would not be possible to help publishers monetise only viewable impressions. It would not be efficient to teach technology to second guess an impression that could or could not be viewed. A major problem could arise when publishers did try to sell viewable impressions (via the predictive modelling approach outlined above). In this situation, they would only be able to sell into direct channels. This would ultimately mean inventory they know to be out-of-view would inevitably be dumped into the exchange (what buyers would knowingly upfront budget into inventory more likely to be out-of-view?). This could be a major setback for the industry given the advancements made over the last 12 months in building more premium liquidity.

The Publisher Opportunity

There is an opportunity here though. The smarter publishers could start using data “clean up”. Clean up in the sense of de-cluttering – removing non-valuable inventory. They could start leveraging the insights a comScore (or technology offered by the likes of Burt Corp) can provide by learning what ads are never seen. Coupled with mounting pressure from agency side and the procurement divisions of major advertisers, publishers will be forced to remove impressions adding no monetary value. Everyone knows bottom-of-the-page leaderboards are placed there only for cookie-bombing gamification. Hopefully this initiative will start putting the spotlight more on this type of erroneous tactic.

How can a publisher make more money from cleaning up? Simply by creating more scarcity for valuable inventory. Develop a holistic yield management strategy and let RTB create upward pricing pressure for valuable inview impressions. The smarter publishers will see this as a massive opportunity – one where value will not be created with out-of-view, “last ad call” inventory.

It will be interesting to see if comScore open up this data via an API. Smart apps could be developed around this, creating a possible Viewable Impression Index (VII). This type of index could then be used for bid decisioning. A centralised repository of viewable impressions across all campaigns by site would enable DSPs to potentially factor in VII scores into any bid decision. It is hard to imagine this would ever come to fruition.

Implications Of The vGRP

The second part of this debate is centred around the vGRP being adopted as a measurement standard. Looking beyond the obvious of how a digitally native GRP metric might make it easier for marketers to interpret digital performance, there are a numbers of implications to consider:

Audience Validation

Those active within the data space will appreciate the need for a separate validation service. 3rd party data is incredibly inconsistent. Different vendors have different taxonomies and methodologies, which ultimately creates a lot of fragmented data. comScore should be seen as a useful validation point. Granted, it’s still grounded in panel based data, but it would provide a form of consistency.

Creating Scarcity

Although covered in some detail above, a vGRP could help promote the need to create scarcity. To date there has been no financial incentive to do so. Publishers have been rewarded for trying to play a game that ultimately no one wins. Mike Nolet referenced this as the “Prisoner’s Dilemma” at the recent Appnexus Summit. However, if advertisers throw their weight behind a vGRP metric, there is no hiding place for publishers who stuff ads below the fold and out-of-view. They would be simply be poor performing and inefficient from a vGRP measure. There could now be an incentive to build better user engagement and deliver more value to advertisers.

Algorithmic Decisioning

This obviously becomes dependant on the speed of the feedback loop from a comScore data source. However, for a given campaign, after an initial data collection phase, why wouldn’t a bidder be able to ingest a data feed from comScore with vGRP figures? A bidder currently uses CPC/CPA as a measure in order to systematically decide how much it should bid. Could it soon use a vGRP and CPP (cost per point) metric instead? The data flow would be different (external data feed versus native pixel data) but it should be possible. If the RTB space wants to evolve and continue scaling and the brand spend still follows outdated measurement models, why not follow the money?

comScore’s initiative is certainly not the Silver Bullet the industry had hoped for. But it’s certainly a step in the right direction.

Tags

Comments


  • alenty

    “With regards to RTB, it simply would not be possible to help publishers monetise only viewable impressions”.
    Yes, it is possible. Alenty’s App in Appnexus allows to buy and optimise “viewable impressions”, and even “efficient impressions” (seen more than 10 seconds for instance). So, high quality publishers should bring their premium inventories to these platforms, where they can be bought at their real value.
    “Could it soon use a vGRP and CPP (cost per point) metric instead?”. Yes Efficient CPM can already be the optimizing goal for DSPs. And it works! Some of our clients have automatically divided their EffCPM (cost per 1000 impressions seen more than 10 seconds) by a factor of 2.7!
    So, yes, new ways to buy and sell brand advertising are ready.
    Laurent NICOLAS, founder and CEO of Alenty.

  • Dark Knight

    How can you buy a viewable impression in real time if you do not know whether it is going to be viewed?

  • http://twitter.com/ProfessorHulk Professor Hulk

    Agreed. Laurent does the app work as you say? How can you know its going
    to be seen for 10 seconds? Doesn’t seem very “RTB” in my opinion. Could you elaborate more on how it actually works.

  • alenty

     Right, you can never know in advance that the ad will be seen 10 seconds. But you can also never know in advance that the ad will be clicked. But by measuring the past performance (clicks or efficient impressions), RTB sytems can optimize and bid accordingly.
    Our system measures the visibility of each impression, when the rtb system can use the results as a goal and use their fantastic optimization algorithms to make the right decisions.

  • http://twitter.com/pauldahill Paul Dahill

    Re your section on ‘Audience Validation’. At the beginning you make the point that the vGRP might be “actually regressing the industry back to 19th century metrics and dumbing down” and then recommend a way to validate sophisticated online 3rd party data sets would be using a panel (which you do recognise). Whilst any additional validation options are useful – there are more interesting ways to undertake this process. I’d argue the best way to validate a new data set, is to run analysis against it (using a DMP) – comparing it to long-established and trusted 3rd party data sets – such as Experian. Or even better, against your own 1st party data.

  • http://twitter.com/ProfessorHulk Professor Hulk

    OK. That makes more sense. So its taking aggregated data on past bids and and impressions to make its decisions going forward. Thats impressive but doesn’t answer the point above that vGRP would be able to answer the RTB “viewable” problem. Although appreciate you have a solution.

    We are all looking for that RTB “viewable” impression which in my opinion comes down to publishers cleaning up inventory through use of vGRP.

  • http://twitter.com/ProfessorHulk Professor Hulk

    1st party/DMP all the way.

  • Martin Kelly

    I buy viewable impression metrics as I think they address some of the flaws in measurement of display but I’m not so sure about the GRP element of it, seems like it’s taking the output of this innovation in the wrong direction.  Are we trying to create a benchmark to TV, print etc where the GRP relates to ads that could have been seen but a lot of the time weren’t, and why does and ad being on the page mean it’s been viewed anyway?  Viewable ad metrics seem to have a more logical application in screening publishers to blacklist them for poor ad placements than anything else.

  • alenty

    Copy/paste from my answer to Professor Hulk:
    Right, you can never know in advance that the ad will be seen 10
    seconds. But you can also never know in advance that the ad will be
    clicked. But by measuring the past performance (clicks or efficient
    impressions), RTB sytems can optimize and bid accordingly.
    Our
    system measures the visibility of each impression, when the rtb system
    can use the results as a goal and use their fantastic optimization
    algorithms to make the right decisions.

  • alenty

     Even if publishers clean up their inventory, not all impressions will not be seen. Top of the page impressions are not always seen (10% on average) because people scroll down before the ad actually loads.
    And don’t forget the price in the equation: a site with a low visibility rate but a lower price may be worth buying…
    So, I agree, there is still some work to do on the business side, even if the technologies are ready…

  • http://twitter.com/ProfessorHulk Professor Hulk

    “And don’t forget the price in the equation: a site with a low visibility rate but a lower price may be worth buying…”

    If your Criteo maybe…

  • Stephan_Pretorius

    From the publisher perspective, would it not make more sense to move comprehensively to viewable impressions, i.e. only serving ads into viewable ad slots?  

    At Acceleration we’ve developed some custom ad templates to test this and the user experience impact is minimal. Taking this route would eliminate the need to try and predict whether an ad is going to be viewable or not.  It would also destroy the market for non-viewable ads, thus decreasing overall inventory in the marketplace.

    Is this too radical an approach?  If so, why?

  • ciaran_mcconaghy

    Tend to agree with Paul Dahill here: vGRP is helpful to a point in the audience validation – it tells you the size of the audience rather than the efficiency of hitting the target audience (vTRP) you set out to reach.

    1st party DMPs, using 1st party data as a starting point, are already
    making big in-roads into being able to provide greater audience
    definition by having consistent audience definitions/taxonomies across brands and markets.

    At Fabric Worldwide we are already seeing agencies getting better briefs, at cookie level, to buy
    against as a result.

  • Spiderman

    If you’re anyone really

  • http://twitter.com/ProfessorHulk Professor Hulk

    Anyone on a CPC model looking at CPM’s that ensure a “nice big fat margin”.