The 'Temporary' Problem with Second Price Auctions

By
  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

The Sell Sider

Today's column is in response to "Second-Guessing the Second-Price Auction Model" and written by Andrew Casale, VP of Strategy, Casale Media.

I am not an advocate of dynamic floors. Nor do I think switching RTB to first price auctions is likely to happen. Every auction that publishers have run on our platform, to date, has been a true second price auction with no dynamic floors or other algorithmic market manipulations. I am merely commenting on this debate with a different perspective formed on the basis of actual bid data.

Everyone actively participating in RTB from the sell side has generally brought most, if not all, of their unsold supply to the party. Some are now even bringing their sold (read: premium) supply to the party by activating bid requests at the time of their raw ad calls and letting RTB demand compete with direct sales. Therefore, while the supply side of the scale is full, publishers participating are still waiting for demand to catch up. The estimates I see in practice and hear about is that somewhere between 25-30% of unsold display is cleared through RTB. While we have, for arguments sake, 100% of the supply available, we have nowhere near 100% of the demand available to fill it.

How this gap manifests in the second price auction model is that many impressions get one or no bids, and therefore effectively clear at the floor. However, a (very) small number of impressions do get multiple bids.

The thing is, when impressions get multiple bids, the second price model works really well. It should, in a true market, if price is dictated by demand. When demand is at a representative level, prices should be fair. This is happening when the perfect storm of bids allows it to:

Bids Per Impression

The data above is based on a representative sample of Q2 RTB activity across the supply we tested. In the case of a single impression getting five or more bids in the fraction of a second we collect those bids, the clearing price of that impression will, on the average, be five times higher than an impression that saw just two bids. It’s incredible.

I quote the word "temporary" in the title of this piece because it’s impossible to say just how long this will truly be a problem. Even when we get to the point where the majority, or in the blue-sky scenario 100%, of unsold impressions are cleared by RTB pipes, this problem might still exist. The model only works when there is so much demand for every single impression that bids stack and market prices are actually set by demand in a true auction.

Knowing that every day more demand moves into the RTB ecosystem, prices will become more in line with what publishers expect. Some publishers are patiently waiting for solutions, while others are testing models that effectively turn a second price model into a first price model by dynamic or soft flooring. A need for a solution in advance of balance coming together is critical. Based on the data above, the one suggestion I’ll throw into the air would be that if an impression is truly in a position to be auctioned off in a second price capacity because there exists multiple bids, a true second price auction should occur.

If an impression is not in a position to be auctioned off in a second price capacity because there is only one bid, that impression could be cleared through alternative means while publishers wait for demand to grow. What that alternative means still needs to be determined, but today, in my view, that’s all we should be looking to change.

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

  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

Email This Post Email This Post

6 Responses to “The 'Temporary' Problem with Second Price Auctions”


  1. Todd Curry says:

    Would be interesting to see you further dissect your data:
    - with intender tag / without
    - bid index for DR clients / brand clients
    - dayparting

    Also, what leads you to believe that demand will ever equal supply in aggregate in the digital ad marketplace? Hasn't been the case since publisher sales teams started courting agencies, got worse with the advent of publisher pagination and interstitials...

    Furthermore, the rate of publisher inventory growth may exceed the rate at which demand enters exchanges, which doesn't bode well for demand intersecting supply at at higher clearing price.

    • I will look to explore the additional data dimensions you mentioned in a future piece.

      As to what leads me to believe demand will ever equal supply, this may never happen and I certainly am not suggesting in the piece that we will hit parity at a specific point in time. I will point out that there was once a time in search when one could buy a high value keyword like "credit card" for pennies a click. While I don’t want to parallel display with search as the two are different beasts, there are similarities in the market building dynamics that have yet to materialize in the relatively nascent RTB model which is what I intended to point out.

      We may never have as much demand as we have supply for say unmoderated Q&A sites, but there may and hopefully will come a time when a higher value category such as automotive or financial sees sufficient demand for their impressions that drives their value where it ought to be.

  2. Nick says:

    Are we really supposed to believe "However, a (very) small number of impressions do get multiple bids."?

    Are we including non-biddable impressions in that statement? Have ad exchanges and bidding platforms so failed us that the technology can't respond in time to produce more than one bid? Or is it that the inventory is of such little quality and therefore interest to the buyer that nobody bids?

    I suspect the last statement is true and so price floors do nothing for us. Garbage inventory = garbage monetization.

    • Quality has a lot to do with it, as does over supply, but I don't think this is the case of strictly the availability of "garbage" leading to a lack of demand. Yes there's no shortage of unmoderated/unviewable/unbranded - fraudulent even - impressions crowding the ecosystem, but there are plenty of high value placements from premium publishers that we see go out to bid that clear one bid or less on the average. Typically they're clearing no bids not because they're bad but because the audience lacks a cookie valued by a marketer (yet) or that has sufficient budget available on its given campaign to produce the necessary bid.

  3. Benny Radjasa says:

    Andrew, thank you for the interesting finding. I want to add a different perspective to your comment on this debate, information you mentioned below only offer one plausible reason why this is happening, I think there are multiple reasons conspiring together.
    I think most of our colleagues do not understand why this is happening, and it is just more than a glut of inventory available in the RTB environment. In the past seven years examining second price action in banner display, I notice that there are quite a lot of not very effective impressions being injected in exchanges and RTB market. For example we are talking about those 5 cents CPM impressions one can purchase in the market, and I am not saying 10 cents CPM impressions are twice better, I am just saying...
    There are plenty of such low quality impressions. Those impressions of higher quality tends to come from professionally design content. Impressions that historically convert better, produces higher KPI and ROI, will naturally be price higher as campaigns that have been programmatically design to optimize will continue to bid on such impressions until it has reached it's bid ceiling, thus raising the prices of so called " more effective" placements/impressions and ignoring on bids to potential low quality placements/impressions.
    So the low CPM impressions are cleared at low prices because not much campaign demand for it. Even if these impressions/placements have a floor price, it does not mean that these impressions will be sold at the floor price. It might not get sold and in that case it will serve the publisher/seller designated default campaign. We should believe that the market will self correct the prices for any placements/impressions. It is not a good system if it does not, think of it like the stock exchange. There will be some special cases where market forces will not set and adjust the pricing, and I hope these are few and far in between.
    All sellers want to be able to convert lead into gold and sell it as such, they will fail and its quite easy to spot the difference even to the most casual observer. Now there will be some who has more devious intention of trying to pass lead as nickel, and some might succeed, and these are harder to spot.

  4. Josh Dreller says:

    Let's not forget the gist of Esco's post (which Andrew's post is a 'response' to...).

    Esco is saying that true price auctions may be fairer than second price because the second price auctions can be manipulated. Even if no one is actually doing anything nefarious here, the fact that it COULD happen is a negative check mark to the channel. Even a perceived problem could be detrimental to the channel, let alone if it's actually happening.

Leave a Reply