2

My AdSense Split Test: Results

Remember I have recently switched to responsive ads on my Elephant 🐘 Website and the ads had a bug? I found a few ways to get rid of the bug and decided to run a split test to see which implementation would be the most revenue effective.

I promised to share with you the result – so here I go! But first, let’s do a quick recap of what happened and what I did.

Issues With Adsense

As you already know, two Adsense ad units were not showing ads after started using the responsive format.

In total, I have 4 ad units. I called them according to their placement:

• Top Ad;
• Sidebar Ad;
• Middle Ad;
• Special Ad.

Both Middle Ad and Special Ad were not displaying ads while the rest of the units were working fine. The whole website was affected and the revenue dropped by 50%.

I had a few ideas about how to solve the problem.

Since all the ads had become responsive, it really did not matter what exactly unit to serve at a particular location. Top Ad could easily go the middle or onto the sidebar. So I could basically throw away Middle Ad and use Top Ad instead.

Another solution would be to try to set up the ad size manually to see if it was the responsiveness that affected the units. I did it with one ad and the change actually fixed the unit and it started serving ads.

So the problem got solved while I was poking around it… But I still decided to run split tests to see which improvement would be most effective.

Adsense Split Test

For the experiments, I chose two pages:

• the main page due to its high traffic
• a page hosting Special Ad

For each experiment, I had three variants of each page:

1. original “as is” with an underperforming ad unit;
2. Top Ad instead of the underperforming ad;
3. the underperforming ad unit with a manually changed size through CSS modifications.

My guess was that #1 (the original) would perform well and win the test. Why wouldn’t it?

For the split test implementation, I used Google Content experiments. (You do not know what it is? Read this excellent post by AuthorityHackers.)

Results Of The Tests

While the test for Special Ad page is already finished, the main page one is still running. I decided to share the results anyway because it is clear already which variation will win.

So, the first is Special Ad page. Here is the experiment chart:

As you can see, the variant #2 (Top Ad instead of the underperforming ad unit) won. The original responsive unit got the second place while I thought it would take the 1st place. The modified original ad unit performed the worst – this is good since the responsive format makes the ads be size agnostic. Less headache for me!

Here’s Main page chart:

This experiment is not finished yet but we can already see the trend. It is similar to Special Ad page’s one. The variant #2 is leading while the original is on the place #2 and the CSS modified one is the last.

Final Words

Do not let yourself assume when it comes to choosing what format/modification/etc to choose. It is always better to test the assumptions and make the decision based on hard cold numbers.

Google Content experiments are quite great for running split testing. Of course, you need to make sure that what you want to test is actually covered by the experiment metrics. The Adsense metrics are and that’s why I chose it for my Adsense split test.

What do you guys think? Let me know in the comments!