Advertising Technology

Are You Bidding Against Another Buyer, or Yourself?

At a recent industry conference, a controversial topic that crept up on panels and during networking portions was dynamic floors. For the uninitiated, dynamic floors is the practice of a supply side platform or ad exchange bumping up the clearing prices of auctioned-off RTB inventory to more closely match the highest placed bid. It is done arbitrarily and ad buyers should care because they may be spending more than they should be.

Throughout my day-to-day dialogue with buyers and through my discussions at the event, it’s evident that no advertiser believes in dynamic floor pricing, a position I maintain as well. Publishers who have employed it also have mixed reactions. So why are some selling platforms practicing it? It is done to fill a gap in liquidity that currently exists in the RTB marketplace. The trouble is this could widen the gap in the long run by diminishing buy side trust in the RTB model.

 Here’s why this happens. Real-time bidding is all auction-based. In fact, it’s predominantly a second price auction – highest bidder wins but pays the second highest price – usually a penny more. During bidding, one buyer may bid $2.00 CPM and another may bid $50.00 CPM for the same impression. That’s a huge gap, but the winning bidder will only pay $2.01. That buyer isn’t expecting to pay $50.00, but she bid that way to win. The buyer probably knew she’d pay around $2.00 or a little more, assuming competition for the impression was heavy. If she starts seeing her $2.01 CPM clear price bump up to $5, $10, $20, $30, she’ll get suspicious.

With dynamic floor pricing, the auction automatically raises the clearing price to the middle of the two highest bids, not based on actual bidding behavior, but instead on past bid activity from the buyer. It’s like robots automatically bidding up your price in the absence of true buying activity, based on what you bid. This is a manipulative variant of the second price auction. This practice discriminates against buyers by effectively using their bid against them, and violates buyers’ trust because of how they’ve been bidding. This is an unfair and opaque market force artificially creating a price because of how advertisers valued an impression in the past. In a true second price auction, buyers should pay the second price – not a manipulation of the price. Buyers who are aware of this are starting to move budgets away. Paraphrasing Meredith Goldman of Criteo, during a conference panel, when Criteo finds out that a marketplace isn’t operating a true second price auction, the company “will immediately cut bids.” Clearly this practice isn’t helping anyone.

Let’s compare this to bidding for baseball cards on eBay. If I’m a collector and I have enough expertise to know the value of a Babe Ruth Yankee card, I’m going to bid for the card at a calculated price, expecting to pay that price only if there is another collector that has my same taste and buying power. If eBay took my bid and forced me to pay something close to it, regardless of whether anyone else in the auction was bidding the price up, I’d quickly take my money elsewhere or I’d severely cut my original bid. If I think the auction is not operating honestly for one item, how am I to know what other item prices they may be manipulating?

For RTB to develop, demand needs to form naturally. There needs to be a fair market. Look at paid search advertising. Even early on, it priced bids by market value. With few advertisers, buyers got major key words for mere cents. Google waited for the market to take hold. That patience paid off as Google now operates one of the largest and most successful marketplaces. If advertisers didn’t trust the fairness of paid search, they would have been conservative with their budgets and growth wouldn’t have been what it was. This is what display advertising faces if dynamic pricing is forced on advertisers. Buyers should take a stand. They can and should vote with their budgets. If anything, the practice should be fully transparent, disclosed upfront and more discussions have to come out of this topic. If markets are fair and demand brings more budgets, prices go up. If we don’t get to that point and if we manipulate the market randomly, no one wins.

My ultimate issue – you can’t call it an auction if you’re manipulating the clear price, and we need to do something about that.

  • Oliver G

    I could not agree more with Andrew, and I find the eBay example spot-on. In the sell-side platforms’ ongoing war over publishers’ inventory, buyers have gotten squeezed by arbitrarily aggressive yield management techniques that distort RTB “auctions” for the sake of artificially elevating CPMs. Even in the case of a buyer with unlimited budgets, such aggressive and arbitrary yield management behavior forces the buy side to create artificial margins by bidding down – below the inventory’s full and fair value – which in turn drives down CPMs and forces even more aggressive yield management practices. This downward spiral is good for nobody and will ultimately stunt RTB’s burgeoning presence in the marketplace.

  • http://twitter.com/NathanThomas456 Nathan Thomas

    Good article and certainly agree. However, it is also written clearly from an advertisers perspective. It cannot be ignored that RTB is often used to get the cheapest possible access to a publishers inventory. I am not saying this is the right way to combat this problem but the publisher side should not be ignored.

    • Andrew Casale

      While I am acutely aware of what the advertiser side of RTB needs, I am speaking from the sell side. In head to heads we continue to find that the fair marketplace we operate consistently outperforms manipulated marketplaces in the long term. We see far greater bid densities than the average which helps set better prices. And while dynamic floors may spike rates in the short term in the long term their rates drop and dip below our averages as budget is pulled back and bidding strategies cut back.

      As RTB has not yet achieved majority fill status budget consideration is as important as rate in these nascent days.

  • http://www.facebook.com/profile.php?id=100002129940919 Yuli Sergeivich

    This article is dead on. I’ve experienced this buying RTB inventory in the last month or so. Cost has gone way up.

  • Willisible

    ‘With dynamic floor pricing, the auction automatically raises the clearing price to the middle of the two highest bids’
    Could you name your sources?

    A DFP is calculated while analyzing the bidding behaviour of every dsp, at every level of the user session on a certain site/zone, and in real time. Which is why it drives prices up for the first impressions, but also reduces the price of end of the session inventory. Calculating DSP specific floors is predatory to the marketplace, and to my knowledge does not exist.

    • Andrew Casale

      You would be hard pressed to speak to a trader who leverages multiple DSPs and has not encountered the practice I have described. DFPs are going beyond pricing the performance decay of the frequency curve. I should qualify that statement however – as there is no standard for what a DFP is, some platforms may strictly set them based on the curve, whereas others are going further, the reality is no one knows who is doing what which is ultimately the problem. My sources are traders at agency desks and a pretty unanimous set of DSPs, some of whom have commented here. Transparency will go a long way.

  • http://www.facebook.com/profile.php?id=569525037 Sunyu Chu

    Come on…just come up with a counter logic to beat dynamic floors and be successful at it. Then someone else on the pub side will come up with a better logic to beat that. Then you come up with an even better logic to beat that! That’s how the mankind evolve, at least in my world. Life is awesome if you live this way 🙂

  • http://twitter.com/AlexAndreyev Alex Andreyev

    Great article Andrew. Interesting, that very recently financial markets are seeing issues with algo, which creates an algo war-zone. Rather than putting one algo against another fighting for pennies, marketers and publishers should be using this great technology to drive greater user experience and beneficial solutions for both sides.

    http://www.technologyreview.com/view/428756/watch-high-speed-trading-bots-go-berserk/

  • Dave S

    surely in your first example, if the buyer is bidding $50 to ensure they win (knowing they will pay the second price, and assuming it will be much lower) they are already manipulating the market themselves?! The principle of ebay is you should bid what you are prepared to pay, and expect that you may pay it. Reagrdles of what sellers do (DFP etc.), if all buyers just bid the fair price they are prepared to pay, they cannot be manipulated. (also from a publishers viewpoint, if they pulled budgets because they felt price was too high, they may then spend more on guaranteed inventory, all the better!).

    • http://twitter.com/NathanThomas456 Nathan Thomas

      very good point. the second price model in itself is flawed to a certain degree… I’m sure the guys on “storage wars” would love this practice – bid $10,000 on the unit just to win it $1 higher than the previous bidder.

  • Adam Lynch

    Kind of agree with the article, but do you also think that maybe, just maybe if everyone is bidding at $50 to win that bid, then it’s actually this practice that’s driving prices up? This would especially be the case if you’re bidding through multiple DSP’s. Maybe advertisers should only bid the actual price that they’re willing to pay for an impression and not inflate these prices to such an obscene levels just to win that bid? I’m not disagreeing that you could be bidding against yourself, but just think there needs to be some clarity as to why the prices are being driven so high and so fast.

  • http://twitter.com/alexmerwin13 Alex Merwin

    What about algorithmic adjustment to the floor price (pre bid request) instead of the clearance price? Then the buy side gets full visibility. My beef is the buy side utilizes algorithmic analysis to determine the future value of a placement based on historical performance. I think it’s unfair that pubs can’t utilize the same technology to determine the future value of a placement based on it’s historical demand. However, if this is done in a manner that is opaque to the manner, it screws us all for reasons detailed here. However, I think by applying the adjustment pre-bid request, it levels the playing field and allows the buy side to make a go/no go decision based on the true value of the impression.

    • Andrew Casale

      We’ve toyed with this idea internally… referring to it as the RTB equivalent of the financial “ask” price… it’s a topic worthy of more discussion… it meets a lot of the needs in the middle by disclosing upfront to avoid a blackbox driving prices while potentially compensating for bid density gaps… but ultimately I can’t help but think we’d need to get the buy and sell side in a room to hash this out together.

Sponsor

Sponsor