Bob Cringely on Google's AdWords Algorithm

Bob Cringely does a very interesting piece this week on Google's AdWords algorithm.

I have no idea what the heck is happening here, but my friendly reader, who makes his living from this stuff, has a theory. He believes the Google AdWords algorithm tries to do many things and one of those is to encourage advertisers to pay more for words. By modifying something that in turn modifies the results, Google is effectively encouraging advertisers to change their behavior.

So increasing the amount per word DID increase sales, though not enough to justify the additional cost. Google's revenue per word, of course, went up by 10X. But dropping the price by more than half was greeted by a huge decrease in clicks-through that could only have resulted from some unknown resultant change in GOOGLE's behavior, given that all other variables were constant.

Here's what he says happened. Some guy put up a copy of his already-successful Adwords-driven site. Instead of paying the $.10 for an adword that he was normally pays, he paid $1.00, and saw better placement. Expected. He was trying to figure out how much more he would get for the extra $.90. Then he dropped to $.40 and saw his traffic drop well below what he had been getting at $.10.

What??!? $.40 is still above $.10 and should have been bringing in some bacon.

This seems to indicate that Google's algorithm manhandles placement based on the sign of the first derivative of your pricing. In other words, not just how much you pay this week but how much you pay this week as compared to last week. Pay less this week than last week and you'll get slapped a little for going the wrong direction. Add a little this week and you'll be given something of a premium for your good behavior. (Of course, this is all speculation... Google protects its algorithms.)

I wonder if they take the second derivative into account. What would happen, for example, if you paid:
Week 1: $0.10
Week 2: $0.20 (.10 increase)
Week 3: $0.25 (.15 increase)
Week 4: $0.45 (.20 increase)
Week 5: $0.70 (.25 increase)
Week 6: $1.00 (.30 increase)

Would that get you better placement on week 6 than, say:
Week 1: $0.10
Week 2: $0.10 (0 increase)
Week 3: $0.10 (0 increase)
Week 4: $0.10 (0 increase)
Week 5: $0.10 (0 increase)
Week 6: $1.00 (1.00 increase)

It'd be interesting to see the results. In order to do such studies one variable needs to be removed... the original site. Google incorporates (or, I should say, appears to incorporate) a number of credibility and history metrics into its algorithm. The fact that the original site had a number of inbound links, search result history and adwords history means that it wasn't an apples-to-apples comparison with the copied site, even though the copied site had the same content.

I'd like to see the same and more done with two completely new sites. Surely some enterprising SEO company would do/has done this. What about Dave Williams at 360i?

I'm not an SEO guy so I'm sure I'm missing a few subtle points here. Overall I was very interested in Bob's article and would like to see more of what's happening. This sort of thing is why Doug Cutter et. al. are working hard on Nutch... an open source search engine. The value is transparency... people get to see the algorithms. Some argue that this makes the ever-escalating war between spammers and search engines tougher, but others believe that access to the algorithms themselves won't aid index spammers all that much.

Bob concludes with:

If that's indeed what's happening, it isn't illegal and to some might not even be unethical (I guess) but it feels just a little bit EVIL.

It's funny... this morning I was just thinking that Google has ended its honeymoon. After a few weeks (months?) of hard press it's clear that the halo effect is gone and people are back to evaluating Google with a logical eye.

Thanks Bob for the great article!