Reprinted with permission from the Lyris.com blog archives.
By: Dane Christensen

A few months ago I implemented Google’s Conversion Optimizer on several of our campaigns. I did so with apprehension after reading about others’ experiences with it. I have a healthy fear of Google’s power over the Web, and hesitated to take myself “out of the loop” when it came to optimizing campaigns. It just seemed a little too much like the fox guarding the hen house. But based on the advice of our Google account representative, whom I trust, I gave it a shot.

I am now happy to report that my Google Conversion Optimizer experiment was a complete success, and I strongly recommend it without reservation.

Before I describe my experience, here’s a brief overview of Conversion Optimizer for those who aren’t familiar with it. This Google AdWords feature allows you to specify the desired cost-per-acquisition (CPA) that you’re willing to pay as opposed to a maximum cost-per-click (MaxCPC). Based on past conversion data, Google AdWords dynamically displays your search and content network ads in the most efficient places at the best times and to just the right audience. Naturally, this algorithm is a mystery to everyone except the engineers who work on it. And it does almost fall into the category of “too good to be true”. Name my own CPA? Yeah, right. So skepticism is definitely warranted. But for me, the proof was in the results.

Like others who have posted about their trials, I got off to a shaky start. A common experience seems to be that Conversion Optimizer will recommend a CPA that is well above the CPA you are already achieving. That seems counterintuitive. You’d think you could start with your established CPA as a baseline and work down from there. I had the benefit of a Google account rep who warned me of this ahead of time and tried to explain the reason why it does that. As expected, the explanation was fairly esoteric and I won’t even attempt to explain it here. Suffice it to say that it’s necessary for Google to recommend a higher CPA in order to gather enough data to begin forming and testing hypotheses.

For several weeks I observed our CPAs under Conversion Optimizer. I’d see some promising signs that were quickly eclipsed by an ad group showing a CPA three times our requested level. I examined our results in an attempt to understand the mechanics under the hood to no avail. Although the results were not at all clear or consistent, we weren’t getting killed by it, so I let it ride.

Over time, the positive results became more and more obvious. I’d find ad groups here and there that were outperforming our requested CPA so I’d ratchet those down a bit. It was a gradual process that played out over a few months, but at this point it’s clear that the Conversion Optimizer is reliably driving down our CPAs.

I think the key point to understand about Conversion Optimizer (and Google AdWords as a whole, really), is that it’s much more like a giant cruise ship than a little speed boat. I read a post by someone who killed Conversion Optimizer after three days. Three days!? This stuff requires WAY more data than that. We have to understand that it is a self-teaching system. It doesn’t just learn how to adjust ad displays based on what visitors do. It learns based on how visitors respond to the changes it makes.

For example, the Conversion Optimizer algorithm may guess that it’s best not to show your content network ad on Saturdays on a particular site. So it tries that, checks the results, and adjusts accordingly. Again, I’m only speculating about what goes on under the hood, but it’s easy to understand why Conversion Optimizer would require lots of data over a period of time to render the desired results.

And speaking of results, our CPAs are currently 25% of what they were just six months ago. Not 25% lower–we pay ¼ what we used to for a lead. That’s dramatic! Of course, Conversion Optimizer doesn’t get all the credit, but it was a big factor.

With dozens of pay-per-click campaigns, hundreds of ad groups and thousands of keywords in which I’m making various other optimizations on top of Google’s Conversion Optimizer, it is difficult to identify crystal clear evidence of it at work. But like oxygen, I don’t need to see it to know it’s there. And I can’t live without it any more.