Sep 302008

Why Position Bidding Wastes Money

I write regularly for the Paid Search column at Search Engine Land. This was posted yesterday in case you missed it.

It's remarkable that in 2008 there are still many bidding systems in use by SEMs and in-house PPC managers dedicated to "finding the right position" for each keyword. These position crawling systems guarantee inefficiency and lost opportunity; to put it concisely: they're playing the wrong game. Here's why:

  1. The value of traffic doesn't vary by position. Careful study on our part, confirmed by University Statistics researchers, has proved that the conversion rates (orders per click), average order sizes, and margin percentages do not vary by position on the page. In other words, the people who click on ads at the top of the page behave the same way on their visit as the folks who click on the same ad in the middle of the page or at the bottom of the page. The quantity of traffic is much greater at the top, but the quality is almost exactly the same. In fact, the quality in position 1 tends to be slightly lower than position 2, and the quality improves slightly as the ads get lower on the page -- these are small effects that can be ignored for practical purposes.
  2. Value of traffic times the percentage of value the advertiser can afford to spend on marketing = the bid. Maximizing the top-line within some efficiency constraint -- what we're typically asked to do -- is "simply" a matter of measuring the value of traffic on each ad and bidding according to the above formula. That will place the advertiser's ad as high on the page as they can afford to be, capturing the most traffic for each ad within their efficiency needs. If that bid places an ad in position 1 -- that's great, position 6? Okay, position 15? Oh well. The position is what it is, and is determined by what your competitors choose to do at any moment.

How do position crawlers work?

Largely, trial and error through the following steps:

  • Test ads in different positions on the page.
  • Measure the efficiency (cost to sales ratio or whatever) at each position.
  • Set crawler to maintain the highest position where the efficiency worked.

At first glance, this may look like the same process. It isn't. The critical mistake is the flawed assumption that the position produces the efficiency, when in fact the position is a coincidence of the value of traffic and cost of a particular position lining up at a given time. The position crawler determines that a position, say position 6, is "magic" for this term, when in fact the ever changing bid landscape will mean that position 6 either costs too much or is too far down the page much of the time.

Let's look at some graphs.

Average Bid Landscape

On this ad, let's say it's "Foo Bar" on Adwords, the exact match version, we know that the sales dollars per click is basically $3. Let's say the advertiser can afford to spend 33% of revenue on marketing, so by economic rationale, we'd say bid $1 on this ad. On average, that puts us in position 6 as it turns out.

The position crawler will get to this same place eventually, trying different positions until it learns that "position 6" is its happy place. But that's the problem, it learned the wrong thing. Position 6 is irrelevant.

At any given moment, the bidding landscape will not look like this average. Instead, it might look like this:

Actual Bid Landscape Version A

Perhaps several competitors got directives from their corner offices to "Be more aggressive". We'd say, well, we can still only afford to spend $1 for traffic, so we're going to get less traffic, but we're not going to overspend.

Position crawler will say: "Gadzooks, I've fallen out of position!" and will start merrily climbing his way back to position 6, even though it means wasting money in the current environment.

On the other hand, maybe the actual landscape looks like this:

Actual Bid Landscape Version B

In this case, the wise system would say: traffic is worth $3, I can afford $1 -- lookie there! We're on top of the page, reaping the HUGE benefits of the higher CTR and Impression counts, and it's cost effective -- Yipee!

The position crawler will instead say: "Egads, I'm in position 1, I need to crawl back to my happy place of position 6!" And so, even though they can get the tremendous extra traffic associated with the top spot cost effectively, he'll waste this opportunity by crawling back 'home'.

The position crawlers are built with all kinds of cool features to "jam" competitors and "take advantage of holes in the landscape", but all these complexities don't change the fact that they're playing the wrong game.

Bidding based on the value of traffic is simple conceptually, but complex in practice. The value of traffic is difficult to measure for low traffic "tail" terms, requiring smart stats and tiered clustering mechanisms. Moreover, while the value of traffic doesn't change based on the position of the ad it does change based on the time of day and day of week, the season, the match-type, the syndication network, special promotions, etc, so the calculations must factor in all those effects to do this well. This is a difficult game, but it's the right game, and the results speak for themselves.

The next time someone tries to engage you in a discussion about "finding the right position" for a particular term, remind them that the value of the traffic is measurable; but the cost of a position is unknown, and unknowable, changing based on the whims of your competition. Bottom line: don't let your competitors run your search program.

George Michie is Principal, Search Marketing for the Rimm-Kaufman Group, a direct marketing services and consulting firm founded in 2003. He regularly writes for the Paid Search column here on Search Engine Land.


12 Responses to "Why Position Bidding Wastes Money"
Hey George, Regarding the university study, I read it over and I don't understand all the math, but I don't think it could be that accurate in their conclusions. I don't understand all the math they used in the report, so this might now sound right when reading this comment, but... At 1 click/day/ad, I don't think that's a fair evaluation of the campaign, even if it is over a 3 month period. According to their stats, they averaged about a 0.9% conversion rate, given 1 click/day/ad, that means each ad they ran generated a single conversion during the testing period. Now I get that they ran 1200 ads which you could justify would make up for this (damn I wish I had that kind of budget!), but it doesn't seem like the scope is set right to draw the conclusions they made. Aside from that, I agree completely with your post. Whenever I setup a PPC campaign I don't just max out the CPC budget and try to get the #1 spot, I work with the client to figure out the math of how much it should cost to get a sale and work out a budget from there.
Thanks Christian, I think they simply present those aggregate figures to highlight the difficulty: the data is incredibly sparse, not evenly divided, and for that reason they need to apply some pretty hairy techniques to get at this. They ran this study on just one of our client's data. We've taken a different approach to the problem and studied the numbers across all of our clients and reached this conclusion. It a difficult problem. Done wrong you reach the conclusion that position 1 converts the best. However, it converts the best because our bidding system pushes the best converting keywords towards the top of the page and pushes the poor converting terms down the page. Doing the study correctly requires looking at this through much more sophisticated means. Best of luck with your campaigns as we roll into Q4! George
Ophir Cohen says:
The Problem with Positions in SEM The whole concept of positions (1 vs 3 vs 15) has a lot of meaning when the goal of a campaign is branding and awareness. When looking for ROI and CPA calculations the results are bound to be as you so nicely describe. One of the main reasons and pitfalls for such position crawlers - is the actual position of an ad. Position #1 can be on the top, and can be on the top right end of a page. Position #4 can be the top right end of the page or in the middle of the right column Position #10 can be on the 1st page and sometimes the second When going to the content network and search partners such as AOL etc. it gets even worse as they display the ads differently. Same goes for languages and Geo Targeting - as SEM (AdWords for this matter) are text ads, adaptive to their position and changing by different browsers and browsing behavior - We are bound to receive such results. That being said, I think this all changes when we discuss fix spots on pages with display advertising and branding and awareness as goals.
SEO Services says:
Nice Post. Thanks for sharing this information with us.
Thanks for your comments Ophir, you raise excellent points. Particularly as Geo-targeting competition in different areas moves positions around and browsing history influence the SERPs, the whole question of "what position am I in?" becomes confused, even for branding purposes.


Check out what others are saying...
[...] Why Position Bidding Wastes Money [...]
[...] (Aside: Google loves these “stomp-on-the-gas” tests. Not only does one advertiser pay more, often additional advertisers who rely on position crawler bidders scramble up the deformed bid landscape as well. CPCs soar, sales stay flat. Bad for the advertisers, great for Google shareholders. See George Michie’s post, Why Position Bidding Wastes Money.) [...]
[...] two different grounds: they were generally based on hunches rather than data, and they were often position-based. Rules often looked [...]
[...] Let the value of traffic drive your bids, not the position. Position bidding both wastes money AND misses opportunities. [...]
[...] you are bidding based on efficiency and not position, as RKG recommends, you are left to accept the lower position or work to improve your quality score [...]
[...] you are bidding based on efficiency and not position, as RKG recommends, you are left to accept the lower position or work to improve your quality score [...]
[...] jockeying for a particular position on the page between different advertisers.  While we’ve long argued against position bidding largely because, as Google has confirmed, traffic value doesn’t vary by [...]

Leave A Comment