Jul 122010

Auction Stuffing?

Is Google stuffing the auction?

Sid Shah of EF wrote a great piece for SEL talking about the mechanics of the auction, and some interesting phenomena observed near the top of the listings. He pointed out that oftentimes pushing bids to get from, in his example position 1.09 on average to position 1.01, involves paying much higher cpcs.

How could that be?

AdRank is determined by the following formula for each ad in a given auction:

Google AdRank Formula

These rankings then determine which ad earns which position on the page.

Once that's done, the actual CPC paid is calculated based on the AdRank of the ad below, plus a penny.

Since the cases we're studying involve ads almost always in position 1, I've framed the formula in terms of the number 1 ranked ad vs the number 2 ranked ad:

Actual CPC calculation based on AdRank

What is clear from this is that any increase in my position 1 bid here should have no impact on my actual CPC against this particular second place finisher. That makes sense. It doesn't matter how badly I beat the number 2 ad in AdRank, I still just pay a penny more than what is required to match number 2's AdRank.

There are then only a few ways to explain why the average CPC would go through the roof for a brand keyword moving from position 1.09 to position 1.01 in Sid's example:

  1. The CPCs required to win the auctions you weren't previously winning are huge. If we assume for the moment that the bidding landscape hasn't changed, then we can say that the first 1701 clicks still cost ~$0.50 each, meaning the incremental 167 clicks gained from moving from 1.09 to 1.01 came at a cost of ~$10.01 each.Is that plausible? Well, maybe. It would imply that you're being outgunned in certain auctions, likely certain geographies, by someone with a pretty compelling QS. Their QS would likely have to be quite a bit stronger than the folks you normally compete with to explain the huge CPC jump needed to beat them. That could happen if local franchises are competing against a national website for that regional traffic. It could also happen if an affiliate is geo-targeting around certain regions, like your headquarters.
  2. The bid landscape has changed dramatically. It's possible that the existing participants in the auction have raised their bids and therefore their AdRank, forcing you to pay more for the same traffic. This kind of pre-Panama era bid jamming could be happening, I suppose, but given the opacity of the bid landscape it seems unlikely, and particularly unlikely to happen every time and almost instantaneously.
  3. You are competing in different auctions. My first reaction was that this phenomena simply shows broad match getting broader as the bids get more aggressive, meaning your brand terms start qualifying for auctions on "rhubarb pie recipes", or whatever. However, we're talking about keywords on exact match, so it can't be this. It could be that we're competing in a wider array of syndication partner's auctions, but that seems unlikely unless they're ignoring the matchtype. We think we've seen this on Google.com only campaigns as well.
  4. You are competing against different ads. This one is odd and a bit creepy, frankly. The notion would be that Google brings other ads into the auction that have higher AdRanks than the old second place ad in order to increase your CPC!But why wouldn't those higher AdRank ads have been included in the auction in the first place? They would have, if they actually had a higher AdRank, but they don't. They DO have a higher product of MaxBid x CTR, but the QS isn't high enough to get them into the auction ordinarily because the relevancy scores are poor.

    The hypothesis is that as gaps appear in the AdRank landscape, created by one advertiser bidding more or another reducing bids, Google sacrifices relevancy for the sake of higher CPCs and inserts less relevant ads into those gaps. They don't do this all the time, because less relevant ads sour users on the sponsored listings. It's a classic short term vs long term payoff when done opportunistically.

Now, at this point, RKG hasn't researched the phenomena enough to reach a definitive conclusion about which of the 4 explanations above dominates. We doubt that 2 or 3 play a significant role. It could be that we're entirely wrong about #4, and that #1 is the whole explanation. But we don't think so, and apparently neither does Sid.

So what?

What if Google is stuffing the auction? There is certainly nothing illegal about it -- it's their game they get to make the rules. While it strikes me as unseemly, I suppose there isn't really anything unethical about it either. Heck, in the old days of negotiated advertising prices there was no rational basis for the price advertiser's paid and no assumption that you paid about what others were paying for similar placement.

Indeed, you're still assured that you won't pay more than your bid for the clicks, so the fact that through a bit of gamesmanship you end up paying a something closer to your bid than you might have otherwise, it's hard to fault Google for that.

It does speak to the importance of bidding what you can afford to spend for the quality of traffic you receive. If you're willing to pay any price to be on top there may simply be an extra tax applied for that decision.


11 Responses to "Auction Stuffing?"
Nice blog post as always George. Needless to say I am on the same page with you here. I think its a combination of #1 and #4 going on with more of #1. My data shows that when you bid real high, your CTRs go down too. I am inclined to conclude that this is because you are participating in auctions that you really don't want to participate in. Is Google stuffing the auction (#4) ? Hard to prove until they give us more transparency. And like you said, its not really wrong. The take away for me is that many advertisers who believe that bidding high on certain keywords (like branded ones) and then forgetting about it, are exposing themselves to big risks. ALL keywords need to be actively managed. Your article is a great segue to my next SEL post which talks about the difference between position 1.05 and 1.0. Although the position seems very similar, 1.05 means you are at position 2, 5% of the time. If you want to be at position 1 all day, Google charges you a massive premium. I have tried to explain why in my upcoming article. Would love to hear your comments when it gets published. Regards Sid
Sid, thanks for 1) raising the issue in the first place; and 2) dropping by our humble blog. I eagerly await your follow up post! Mark Ballard on our team has seen instances where cranking the bids on certain non-brand keywords (as part of a brand-building initiative) resulted in much higher CPCs with almost no increase in impressions or change in CTR. In fact, I'm curious that you're seeing the CTR drop on an exact matched ad. That suggests to me that perhaps it is either wider syndication, or perhaps the syndication partners not respecting the matchtype, or engaging in some auction stuffing themselves. I can't see any other reason why the ad would be less attractive to the incremental users. Fascinating stuff, and we agree wholeheartedly that paying attention to brand bids is important. We think that since most of the sales through brand traffic are not incremental, and simply cannibalize sales from natural listings and affiliate listings, it may make sense to set the efficiency targets for brand terms to whatever the rev share percentage is for one's affiliates. Bid just enough to keep from having to pay their commission plus the additional discounts from the coupons. Thanks again for starting the thread!
Billy Wolt says:
I would be more inclined to believe #1 if the cost of the incremental clicks came in at a cpc above $17.80, but since they are $10, the middle bid should have captured those impressions and clicks. the quality score is the missing variable, but somehow I think it would not have changed much.
Billy Wolt says:
Since we are on the topic of not making sense, here's one for you. A few months ago we a bunch of customer generated keywords that were searched on our site. the bids for all were the same. I noticed there was a range of quality scores from 3-10, before anything even received an impression. I was surprised to see a few totally unrelated keywords get quality scores of 10, and keywords that relevant to the landing page get a 4. Needless to say our account reps answer were unsatisfactory. Another fun tidbit: Keywords were launched. One keyword received 3 impressions, 2 clicks, $.27 cpc (and a sale on our end). There were only 3 other advertisers. A few hours later: keyword is inactive Below first page of $1.00. I did a search, still only 3 advertisers. Google gave me some story about the competition probably pushing bids up giving me a lower adrank, but I don't buy it. I doubt the other 3 advertisers started a bidding a war on this relatively obscure keyword.
Thanks Billy. Falls into the category of "Things that make you go 'hunh?'" The auctions where less than a full compliment of ads are shown is an interesting topic in and of itself, and shows the flip side of Google's behavior. Where the user's search shows little commercial intent, Google really doesn't want to serve ads, and those ads that are served are served at a premium.
Billy Wolt says:
well, considering it was a product sku keyword, i'm not so sure what isn't commercial about it. Obscure meaning there was only 4 advertisers.
Tim Ogilvie says:
Another potential explanation without the nefariousness: Google estimates the probability that you've got the highest performing ad, sampling other ads in position 1 with a frequency that matches their statistical confidence that you're #1. If this is true, increasing your bid will decline in effectiveness as you approach position 1.0 (where increasingly high levels of confidence are required to eliminate your competitors altogether). Purely speculation, and virtually impossible to get a clean test, but we've seen data that fits.
Tim, you make an excellent point. Google normalizes QS in part by testing ads in different positions to determine the CTR in each. Could be that if you pay a high enough premium they don't test as much. Thanks for stopping by! George
Johnny Firic says:
Greetings from Split, Croatia! I'm a relative newcomer to the magical world of PPC, although I do have some search engine experience. (It was a local, topical SE.) I cannot tell you how excited I am to see such a wealth of expert advice to help me learn the ropes. Coming from a pure computer engineering background, everything that has to do with bids and budgets makes me very nervous! Thankfully, You Guys exist. I have already read "The Subtle Science pt.1" and pt.2 is open in my next tab. PPC Hero and the Red Fly Marketing blog are also my new best buddies :) Based on my newbie status, I realize that offering any opinions of my own must come off as downright arrogance. I don't suppose that anyone will be offended by the following, though. My company (not through my decision-making) bids primarily on broad matches, and the behavior of our top-ranked KWs and their various metrics (CPCs, CTRs et al.) congrues with what is said in this article and the others which are mentioned. We are ranked highly on several small European markets, where there are rarely more than 3-4 competitors. I'm talking about countries such as Slovenia, Slovakia, Ukraine... important markets for us. It makes no sense to rank first and pay more, as opposed to third (out of three, or four) and pay MUCH less. One way I use to estimate the effectiveness of the many broad KWs we bid on is to actually do the searches myself, using (for example) the Slovenian version of Google to search for whatever related KWs I can think of. I'm pretty happy with the results this method produces. In a few months, as I gain more skills and experience, I'm hoping to contribute more in terms of actual insights to this great community, but so far let me just repeat a big THANK YOU for everyone making my life easier with their conclusions and advice. I know that I'm speaking for thousands of people not confident enough to post comments on their own. Cheers!
Johnny, welcome to the club! Thanks for the kind words. Your processes probably make great sense in your markets, and it's all about doing what works. As competition levels increase, the SERPS become saturated with ads and the long tail becomes more important you'll need to think about scalable processes that can work effectively on huge numbers of keywords constantly. Best of luck, we look forward to your commentary.


Check out what others are saying...
[...] This post was mentioned on Twitter by Alan Rimm-Kaufman and others. Alan Rimm-Kaufman said: rkgblog: Auction Stuffing?: Is Google stuffing the auction? Sid Shah of EF wrote a great piece for SEL talking abo... http://bit.ly/aBHuVI [...]

Leave A Comment