Is click-through-rate normalized by position? Of course it is

I have heard this debate numerous times, so I felt it was time to step in and spread some truth.

Your ad’s rank in Google is based on three factors:

1. Maximum cost-per-click
2. Click-through rate (CTR)
3. Quality Score (QS)

Given the fact that ads in higher positions tend to get a higher CTR as a result of user behavior (people click the first thing they see), many SEMs have concluded that an easy way to reduce your CPC is to simply bid at a high position when you launch your ads, get a great CTR, and then surreptitiously lower your bid and position over time. The idea here is that Google’s system will have already given you a lot of credit for your CTR, and you can then pay a lower CPC to maintain a good position moving forward.

This plan would make perfect sense if only Google were run by idiots with no knowledge of math or programming. Google’s objective is to make money. They make money on AdWords by maximizing the revenue per thousand impressions (RPI), which CTR X CPC calculates. If people could game the system by getting a quick CTR boost and then quickly reducing bids to a lower position, Google’s RPI would suffer.  So how do you stop the gamers? Well, there are two ways to do it: normalization and sandboxing. Since I’m trying to write a blog post almost daily, I will save sandboxing until tomorrow. Let’s focus for now on normalization.

Normalization means that the CTR at a given position for a given query is compared to other advertisers’ CTR at that position. So if I have a 20% CTR in position #1 and a 22% CTR in the same position, assuming max CPC and QS are the same, you will show up ahead of me. And if I have a 20% CTR in position #1 and you have a 2% CTR in position #2, this does not necessarily mean that you will show up ahead of me because Google might extrapolate that my projected CTR at position #1 might in fact be 22%, so given the same bid and QS, I could still out-place you.