Buy It Now: Changes Needed To Improve RTB Valuations

casale-ddt“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 Casale, VP Strategy, Casale Media.

What’s an ad impression worth? That question remains surprisingly difficult to answer when it comes to real-time bidding. On our exchange with connected demand-side platforms, we’ve seen bids within the same auction for the same impression range from one penny to $999.99. The perceived value of these impressions is all over the map. And would-be buyers who are outbid never find out which price ultimately won the impression they attempted to buy, thus raising a huge barrier before the alignment of bid prices with market demand.

Improving today’s system of price discovery is imperative if we want to continue enjoying torrid growth rates in the RTB marketplace.

To dive a little deeper, less than 5% of the auctions we run have a second bid within a penny of the winning bid price (e.g. $2.50 vs. $2.51). That would be a sign of a positive valuation curve in which those in the market mutually agree upon the approximate value of an impression. By contrast, in our auctions the winning bid price is quadruple that of the second highest bid (e.g. $1.00 vs $4.00) an eyebrow-raising 40% of the time, underlining the lack of agreement on appropriate valuation.

It doesn’t have to be this way. I’d like to propose that we borrow principles from both the “first look” private marketplace model as well as from eBay, to help narrow this gap. The idea is to set an upfront expectation about the value of an impression. This would enable exchanges to conduct real-time bidding and real-time buying in parallel, putting a greater emphasis on the value of any given impression.

Here’s how it would work: We’d add the RTB equivalent of eBay’s Buy It Now concept, which I’ll herein refer to as a “publisher’s ask price.” A publisher would continue to set a floor as they normally do today – say at $2.00 CPM – but could also set an ask price for a given impression – say at $5.00 CPM. As is the case today, unless the floor price is met, the impression will not be bought. But if a buyer is willing to go above that floor to the posted ask price faster than competing bids, they’re guaranteed to win the impression.

Introducing this additional element of speed would let buyers win impressions not only by submitting the highest bid but also by submitting the fastest bid that meets the ask price. The average auction takes place in 100 to 150 milliseconds, but with bidders incentivized to bid even faster, the exchange could conceivably close an auction early,  in one to five milliseconds (yes, we do get bids this fast, especially with peering). The overall ecosystem would benefit from this. Ads would be served faster; page load times would accelerate, benefiting both the publisher and the user experience; and brands could more predictably value the impressions they most covet. The exchange would then only ever “go to auction” if inventory fails to be bought outright for the ask price.

An ask price option benefits buyers in a number of ways. Potentially, they can get a discount on their clear prices by avoiding bidding wars that ratchet higher than the ask price. Ads would also serve on the page faster, which we know has some positive influence on ad performance. Finally, buyers would spend less time guessing what the current arbitrary high bid price is for an audience segment they deem to be of great importance. Is it $10, $100, or $999? With an ask price, buyers would finally know.

One might argue that the publisher loses here because buyers can win auctions for prices lower than their potential maximum bid. It’s true that this could happen, but looking at the raw transactional data, I contend that the downside is marginal at best. As with the data I presented earlier, the number of times a bid clears far above a floor is low. This won’t change until bid density dramatically rises, and – as it does – publishers can opt to increase their ask prices, opt out of the model, or perhaps only employ this on their long-tail impressions. Furthermore, publishers could set prices in closer alignment to how they value their media (e.g. based on audience segments stored in an integrated DMP), rather than letting the sparse bidding market arbitrarily decide, the way it does today.

Today's RTB market faces a huge disparity in impression valuation. The overall benefit of putting an “ask price” option in place would be to arm buyers with more insight into how sellers perceive the value of their media. Currently, we have a zero-to-infinity price delta, which is not an accurate representation of the ecosystem. By creating a tighter price curve, we’ll add more predictability to buyer expectations in this marketplace, which will benefit all players over the long haul.

Follow Casale Media (@casalemedia) and AdExchanger.com (@adexchanger) on Twitter.

 

9 Comments

  1. Ramsey McGrory

    Interesting idea Andrew. What about lost bid data? Should advertisers/bidders be able to retain lost bid data to build models for bidding/targeting? You could argue that the bid landscape would be more accurately valued. Are the downsides too great?

    Reply
    • Andrew Casale

      I like opening up lost bid data to solve the root cause of this issue. Providing more transparency into the auctions conducted is where I think things are going to go longer term.

      That said, I don't think lost bid data benefits the publisher unless the data is very transparent - like that which is available through the financial markets, so as to incite competition between buyers. And I get the sense that it's going to be tough to sell that in through a standard that all agree to... therefore my thought was to propose something in the mean time that merely treats the underlying issue. If nothing else, it makes for good discussion!

      Reply
  2. It would seem to me that one the causes of such disparate bids is information asymmetry. In your example - it's not that the $1.00 bidder didn't know he had to bid $4.01, it's that the $4.00 bidder knew something about the user behind the impression that the $1.00 bidder didn't. That's entirely likely in a scenario where the $4.00 bidder is a retargeter and the $1.00 bidder is just trying to be in front of a more generic audience target - the retargeter has more/different information about the impression.

    Assuming that's the case - why is this a problem?

    Reply
    • Andrew Casale

      There's nothing wrong with a retargeter having more information than say bids based on 3rd party audience data at all. The issue at play here is the a general lack of a consistent price band for media. To me, this undermines the auctions that are being conducted, where a bidder can submit a bid of say $99.00CPM with no intent of actually ever paying it. It would be in a publisher's best interest to cap the high bid and broadcast it so that prices align closer to their expectations. This will lead to more scenarios where all bidders have a more equal chance of getting high value impressions, leading to more density at price levels that are comfortably high but not unrealistically high. As it is today we'll never see high bid density at $99.00 and so I see it as being more of a detriment than a benefit to allow bidding to infinity.

      Think of it another way. Private Marketplaces manifested to created more interaction between buyers and publishers. The hallmark of the model is first look where an agreed upon price point between buyer and seller "trumps" the auction and is generally only 2-3x higher than a floor. This is done to allow a consistent buying experience where the buyer knows the price they have to pay to win and if it's warranted they can pay it and get a predictable result. My thought is to move this benefit into the open market as well whereas today this is already happening, but only manifested in private marketplaces.

      Reply
  3. Your point about a 'zero to infinity' price curve making price / volume / performance predictions challenging on the buy side makes a lot of sense. But I disagree with this approach. I think it hurts pubs.

    The whole premise of RTB is that the buyer has discrete information from the seller about the impression, and that this information enables them to valuate the impression at higher rates than they would have without it. This drives up yield, and is why the model works for pubs. By creating a time dimension in the model - and arbitrarily setting an ask on the pub side w/o knowledge of this information - publishers could diminish the value they could create in today's ecosystem. Maybe the slower bidder would have bought the impression at a higher price.

    Inherently, this 'valuation information' is going to vary bidder to bidder; therefore the vast variance in valuations makes sense.

    I get the counter argument to my position above that you state in the article, but my gut feel on this is that it's more a benefit to the buy than sell side. I agree with another commenter that loss notifications are a better means to address this issue.

    Reply
    • Andrew Casale

      I think the issue you present can be treated quite easily by monitoring bid landscape reports. If a publisher starts to see huge density at their ask price, it would stand to reason that they ought to consider raising it.

      Reply
  4. One important distinction here is that when someone bids $4, for example, it's usually a "most I'm willing to pay some of the time" bid by that buyer. Surely if they're eCPM came in at $3.99, they'd lower their bids. It requires more work for the buyer in this case, which is something worth considering.

    Reply
  5. Andrew, I like the idea in principle though I would say that if the buyer has a "buy it now" prive in mind, call the sales rep from the Publisher and cut a direct deal 🙂

    If programmatic mechanisms are required, then let's call it a preferred or reserved programmatic deal and set the price up front.

    Reply
    • Exactly. What this post proposes is already done today with programmatic direct. RTB is build for bidding.

      Reply

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>