Split Testing Ads

[Andy]

I want to host an affiliate link ad on one of my high traffic pages where I have no idea what kind of ad should convert best. It is monetized by adsense so maybe I can get a clue from the adsense ads that are served? Since google should be fine tuning their ads to deliver most revenue.

I can't easily tell what kind of people visit this page. So I am thinking about ads for things of general interest for computer users like spyware removal tools.

What I decided to do was rotate text ads around several products and monitor each one. I have written some php code that displays a different text ad each time the page is displayed. The hits on the page are counted together with ip address, time, user agent and referring page (if available). Also, clicks on each ad are recorded and the sales data for each ad is tracked.

The first stage of this operation will be to create a set of product ads that actually get clicks. Then I will adjust the wording of each ad and look for the ones that  result in most sales revenue. This isn't ideal since it relies on me making a good choice of words for each ad. To improve my chances I am getting a clue from the headlines used on the product sales pages since I am assuming they have been tested and tuned for maximum sales.

In summary:

1. find products that are of interest to my visitors.
2. settle on a set of popular products.
3. fine tune the ad wording for each ad.
4. focus on the highest revenue earning ads and further refine the delivery of them based on factors such as time of day or day of week.

If you have experience of tuning ads for maximum performance, please post.

I am thinking I may have missed out on some big money for not doing this kind of thing before.

If it is a successful system I will probably market my code and tutorials later for people to replicate my system.

[Andy]

Well, nothing is ever as easy at it first appears. First of all I had to adjust the code to filter out a lot of spam hits (and I mean lots!). Now this is done, I can see the real visitors to the page and where they came from.

Also, the click tracking is working nicely.

My code saves 2 sets of data with a daily date appended to the filename: hits and clicks

The files look like server logs but without the spam hits from robots and only specific to the one page I am monitoring.

I am storing the ip address, referer url and user agent.

Also, I can rotate any number of ads but I just have 2 rotating at the moment.

I hope checking my stats is not going to become an addiction like crackstats.