How AdSense Experiments Increased My Ad Revenue by 49% for Some Ad Units

Aerial view of the huge Halloween Corn Maze in Pennsylvania, Poconos Region.
Aerial view of the huge Halloween Corn Maze in Pennsylvania, Poconos Region.

Now that’s a corn maze.

Every year in October we go to a pumpkin patch with a corn maze.  The corn maze is a huge hit with our kids.  They’re pretty cool.  Fortunately, the one we go to has one entry/exit point so we can sit there and wait for the kids to pop out.  It’s also small enough that kids can go in on their own without fear of ending up lost for hours.

Mazes, in general, are like website ad monetization.

You can choose color A or color B.  Ad type A or type B. Placement A and/or B.  The options are endless.

For me, investing time in testing ad types, designs, placements and ad networks pays off.  Modest increases in an RPM can result in a nice monthly revenue bump which adds up over time.

Until recently testing AdSense ads against one another either required a third party platform like Ezoic or manually testing and tracking every little change.

But lo and behold, now you can test AdSense ad types (text vs. image) and styles directly in AdSense.  It’s called Experiments.

I have to admit I’m a bit late to the party with this feature.  I think Experiments have been around for almost a year.  I didn’t start using them until September 2018.  Better late than never.

It didn’t take long for me to really appreciate the results.  within one month, I was able to increase RPM for some units 40% to 50%, which is huge.

Below I set out some results I’ve had followed by how to set up your own Experiment.  I wrap this up discussing two suggestions that would make AdSense Experiments much better.

5 AdSense Experiment Results

Bellow I set out screenshots of 5 concluded AdSense Experiments.  4 results illustrate excellent improvements and one is an example where the original ad was the best (by far).

I’ve run quite a few more than these 5… this is just a collection for illustration purposes.  Results are all over the map.  Some tests result in the original being better and others result in the new design performing better.  I had one Experiment result in a tie.

I’d be happy if just one ad resulted in an increase in revenue, but I’ve had several result in huge revenue increases, which has made AdSense Experiments very much worth doing.

Either way, it’s a win/win.  Check them out:

AdSense Experiment example with 49% RPM increase AdSense Experiment example with 22% RPM increase AdSense Experiment example with 50% RPM increase AdSense Experiment example with 44% RPM increase

AdSense Experiment example where the original ad outperformed the test version by 51%

I’m still in disbelief that some Experiments resulted in 40% to 50% increases in revenue per one thousand visitors (RPM).  That’s huge.  Of course, not all of my Experiments had such dramatic results, but the fact 3 of them did is amazing.

While a decrease seems bad on the surface, it really isn’t.  It just means the original unit performs well and that there’s no need to change it.

How to set up your own AdSense Experiment

I have to hand it to AdSense.  The process couldn’t be easier.  The user interface is intuitive and it takes less than one minute to set up an experiment.

Here are the steps inside your AdSense account:

Step 1: Access the Experiments section inside AdSense

Step 1 in setting up an AdSense Experiment

Step 2: Name your Experiment, choose your ad unit and experiment type

Step 2 in setting up an AdSense Experiment

Step 3: Test Type – Image vs. Text

Step 3 in AdSense Experiment - choose to test ad types or styles

If you choose to test “type”, make your selections, then click “Create”.  You’re done.

Step 4: Test Style – Select or create a style that will be tested against the original

Step 4 in setting up AdSense Experiments - test styles

Another testing option is testing ad styles which boils down to colors and font.  Select your new style options then click “Create”.  You’re done.

Choosing a winner

You can opt to have AdSense automatically choose a winner at the end of the Experiment.  I keep that set, but I have yet to wait until the end of the experiment.

After about 10,000 to 12,000 impressions, AdSense will tell you which version is performing best.  Being an impatient person, I make my selection at that point instead of waiting for the entire Experiment to complete.

When you choose a winner, AdSense automatically assigns the changes to that particular ad unit so you don’t have to do anything else.  The winning type/style immediately begins displaying on your site.

How AdSense Experiments could be improved

I love how simple and effective this is, but I have 2 suggestions for improvement.

Suggestion #1: Ability to test different ad types against one another

The first big problem is you cannot test different ad types against one another.

Yes, you can test a text ad unit against an image ad unit.

However, you cannot test the following (as examples):

  • Link ad vs. in-article
  • Regular text ad vs. link ad
  • Matched content vs. link ad


In a nutshell, you can only test the same type of ad unit against one another, which leaves testing colors and fonts.  In the case of the regular ad unit, you can test text vs. image.

Suggestion #2: Ability to test multiple ad types and designs at the same time

It would be much more efficient if we could test 5 versions against the original (or more).  I appreciate it would need to run longer than it does, but it would be a big time saver.

What I do now is I run one test, complete it, then set up a new test… working to get the highest earning design possible for each ad placement.  While it’s not all that much work, it sure would be nice to test multiple ad types and designs against one another at the same time.


Can you test in-article ads or matched units?

No.  At this time you can’t test those units, which is too bad.  I like those units a lot so it would be good to be able to see how they measure up with different designs and against other types of ads.

How long does an AdSense Experiment take?

You need to generate a sufficient number of ad impressions in order to get a result.  If you get 100,000 page views per day, an Experiment won’t take long.  You could run multiple Experiments per day.  If you get 1,000 page views per day, it will take a number of days.

Can you test placements against one another?

No.  An Experiment will only test ad variations within the same placement.  If you want to test ad placement configurations for maximum RPM, you need to use a more robust testing platform like Ezoic.

Can you run more than one Experiment at a time?

Yes, but each Experiment must be for a different ad unit.  If you have an Experiment running for ad unit A, that ad unit A is no longer eligible for another Experiment while it’s in an active Experiment.

However, you can run an experiment on ad unit A, ad unit B and so on all at the same time.

Do the results tell you the RPM for the original and the test ad?

Yes.  I blocked the RPM data from my results screenshots above because AdSense doesn’t permit Publishers to share that information.  However, once enough impressions accrue you will see the ad unit RPM information for the original and the test version ad.

Can you test AdSense ads against ads from another ad network?

No.  AdSense Experiments only tests AdSense ads against one another.  Again, if you want to test ads from a variety of ad networks against one another, check out Ezoic.

Leave a Comment

Your email address will not be published.

Scroll to Top