Via John Battelle, a very interesting paper (.pdf) on pay-per-click search from one of TheLadders‘ founders, Alexandre Douzet. His discussion of multivariate (aka Taguchi) testing – with it, his landing page conversion went from 3.47% to 6.52% – really couldn’t be a better commercial for the likes of Optimost and Offermatica. The effect of a single tweak to his AdWords copy – switching the position of the verbs ‘search’ and ‘find’ – resulted in a near-doubling of click-through rate.
Neat stuff. However, the real eye-opener was his examination of AdWords for Search vs. AdWords for Content. According to Douzet,
Content listings [...] behave very differently. [...] it appears as if there is a volume-maximizing position. This position, occuring between positions two and three, generates the most volume of actions, although not necessarily the most volume of clicks.
He’s got a nice graph in there to back this up. If generally true, this has huge potential consequences for Google, since the first-place bidder on AdWords for Content is throwing money away – more clicks for more cost but less quality actions, and actions are where the money is. It’s therefore in the first-place bidder’s interest to lower his bid. Of course, once he does this, he’ll no longer be the first-place bidder – and now it’s in the new first-place bidder’s interest to lower their bid. Rational behavior on the part of all parties would result in a gradual decrease in bids across the board, sinking both Google’s revenue and the revenue of every publisher in the AdSense network…
Why would favorable actions peak between the second and third position? I don’t know, but I suspect it’s due to all of the designed-for-AdSense sites with little-to-no useful content. People land on them inadvertently, see a big batch of irrelevancy, want to get off the site as quickly as possible, and click on the very first link they see… combine that with publisher-initiated click fraud, which will disproportionately fall on the highest-paying clicks, and Douzet’s findings start to sound both plausible and general.
If I were in the AdWords for Content game, I’d be doing some experimentation.