AdSense Split Testing with PHP

[vertex42]

Perhaps this is common knowledge, but if not, hopefully this will help people do some split-testing to increase AdSense revenue.

1. Set up 2 or more channels for your split tests. For example, create channels called "test1a" and "test1b".

2. Create your two different ad units (one for each test case) and copy the AdSense code into a text document for convenience.

3. At the top of your php-enabled web page, enter the following code:

<?php
$testcase = rand(0,1);
?>

4. Place the following php code where you want your ad unit for test1a to show up:

<?php if ($testcase == 0) {?>
-- Copy and paste the AdSense code for test1a here --
<?php };?>

5. Place the following php code where you want your ad unit for test1b to show up:

<?php if ($testcase == 1) {?>
-- Copy and paste the AdSense code for test1b here --
<?php };?>

6. Sit back and wait until you've received enough pageviews so that the average CTR for these channels stops jumping around.

Using this approach, you can actually be running as many splits as you want. For example, if you want to test four different ad units, you would use rand(0,3) to choose a random testcase between 0,1,2,4.

Enjoy.

[AlexKing]

Not a lot of helpful topics here lately.
Thank you vertex42 for the simple and, most important, helpful solution.

[vertex42]

Thank you, AlexKing.

I should also mention that this technique can be used with a single channel to randomly change the look of the ad. Although I don't have a lot of data on this, but I have read that on sites with return visitors it helps to have the ads change now and then.

I did quite a bit of testing with inline ads (ads appearing within the main body of text). Here is what I found, although the results will certainly not apply in every case (but they are worth experimenting with).

1. Ad units without borders were slightly better than the same ad unit with a border.

2. A block of two 468 x 60 ad units usually outperformed a single medium or large rectangle.

3. A block of two 468 x 60 ad units resulted in a much better CTR than a single 468 x 60 unit. However, sometimes the same ads appeared in both units (I thought that wasn't supposed to happen).

4. I haven't found any consistent answer about whether to make the URL text color the same as the other URLs on the page. Sometimes, it helps to have the URL standout, and sometimes it helps to have it look like the other links on the page.

I didn't experiment with image ads because I don't like them.

[netmeg]

I've found this very useful when testing colors for ads - for example, I wanted to know which colors worked best for the links on my ads. So I displayed randomly between red, blue and green, and found that red vastly outperformed blue, and nobody seemed to like green (that was on a site that largely red/white/blue themed) I also intend to use it to randomly display ads from different sources.

I also used it on referrals when they first came out, but nobody converted on those no matter what I did.