Dan Friedman at Google recently posted that conversion rates don’t vary much by position.  This is one of the few times where Google has posted an analysis of this type with which we disagree.  We have conducted a study that looked at conversion based on individual keywords in varying positions and analyzed the conversion trends.   We came back with several conclusions:

1. First position will generally have significantly worse conversion rates than lower positions.  We believe that this is because of what we call “reflex clicking”, which simply means that a certain segment of the population will always click on the first sponsored link even if the search result is clearly un-targeted to their need.  Whether first position is a good decision for any keyword is determined by whether the substantial increase in volume at this position improves aggregate profitability given the higher cost of acquisition created from the combined increase in click cost and lower conversion rate.

2. For most campaigns, conversion rates improve as position declines, down to about 4th-6th position, where it levels out.  However, this improved profitability on a per-unit basis is offset by a significant loss in volume as position drops. Campaigns managed solely on a per-transaction CPA basis will likely undervalue keywords because the volume component will be ignored. Again, aggregate net profit is the right metric to look at.

3. For some markets, a high position sensitivity means a drop in both volume and conversion rate below a certain position. These position-sensitive campaigns can be dangerous, as your cost-per-action will rise considerably as you drop below the optimum converting position.  For this type of campaign (and many financial services or lead generation campaigns fall into this category), it may be necessary to pause keywords that are unprofitable at the optimum position while conversion improvement is addressed through landing page/site testing.

So why are our findings different than Google’s?  Our guess is that this is one of those cases where evaluating performance across all advertisers can be misleading.  The trick to evaluating position influence on performance is to isolate the analysis to the same keyword/ad and vary only position.