The Adsense Help Center provided some Optimization Tips for Adsense. First of the tips is Test your success with channels.
How do you do that?
The tip mention using channels and experimenting but isn’t very clear on how to design a test whether a certain ad placement, format, or color will be effective.
Here is what I am doing on one of my sites. I hope it provide you some insight into designing your own test.
Before we begin, you might want to refer to this very enlightening article on Experimentation and Testing.
There are 2 things I like to test,
- Effectiveness of different position. (See also Where should I place Google ads on my pages?)
- Effectiveness of different colors. (See What color palettes are the most successful?)
You should test each of them separately so that the effect of what you want to test is isolated and not jumbled up with the effects from the other case.
The testing technique we are using is what is known as A/B Testing.
The measure we are using here is click through. Since the value of each click is different, using absolute dollar value to compare the cases is not feasible.
Testing Effect of Different Position
There are 2 methods to do test positioning.
- Have the ads in the respective position and use channel to track which position is effective.
- Put the ads in position 1 for a period of time (1 to 2 weeks). Change to position 2 for the same period of time. Compare the 2 results.
Method 1 is preferred since the positions are tested at the same time under similar condition.
Method 2 could be influenced by external factors such as your promotion activities which may be done during one period but not the other.
However layout constraints may restrict you choice to method 2.
Testing Effect of Different Colors
Again there are 2 methods,
- Use color 1 for a period of time (1 to 2 weeks). Change to color 2 for the same period of time. Compare the 2 results.
- Use a tool or server-side scripting to distribute ads of the two colors randomly but evenly. Collect the result over a 1 to 2 weeks and compare.
As in the case of testing position, separating the test over 2 periods can subject the test cases to different factors.
Method 2 is an improvement over Method 1 but would need the use of tools or programming. I use phpAdsNew for this.
I am still in the midst of my own testing and I will share some results when they are ready.
If you have done or are doing similar test, I like to hear about them.