The Optimization tab

Experiments FAQ

Expand all  Collapse all
What are experiments?
Experiments allow you to compare the performance of different ad settings. They work by splitting your site’s traffic between the two ad settings, so that their performance can be measured side by side. Experiments help you to make informed decisions about how to configure your ad settings, and can help you to increase your earnings.
Which ad settings can I compare in experiments?

The following ad settings are available for use in experiments:

  • Ad unit settings: All ad type and text ad style settings are available for experiments.
  • Allow & block ads: The following ad blocking settings are available for experiments:
    • General ad categories (including first level sub-categories)
    • Sensitive ad categories
    • Ad serving opt-outs (e.g., enhanced text ads, expandable ads, etc.)
  • Ad balance: You can compare your current ad fill rate against a different ad fill rate. Learn more about ad balance.
Can I run more than one experiment at the same time?
It depends on the type of experiment you want to run. For instance, you may only run one property-level experiment (e.g., "Allow & block ads") at a time, but you may run both property-level experiments and ad unit settings experiments simultaneously.
How do I run an experiment?

There are two ways to create experiments. You can either:

Will I have to change my ad code?
No. You don’t need to make any changes to your ad code to run an experiment.
How long will my experiment take to finish?
You choose when to end your experiment. However, we recommend that you wait until your experiment is marked "Ready to complete" before you choose a setting as the winner. Depending on the volume of traffic to the settings that you're running the experiment on, this might take from several days to one or two months.
How do I know if my experiment is complete?
Your experiment is complete once it has gathered enough data—we'll let you know when that's the case. It's important to consider both revenue (RPM) and ad session length when deciding whether to go ahead with your change.
Do I have to choose the winner of an experiment?
When you create an experiment, you can opt to let Google choose the winner for you. Google will automatically apply the winning setting after your experiment has finished. If you prefer not to select this option, then you'll need to choose the winner of the experiment yourself.
When should I choose a winner?
To help you decide the right time to choose your winner, we show you the progress of your experiment. Progress lets you know how many days to go your experiment has before we're confident that either your original ad setting or the variation is the best performing of the two settings. We recommend that you wait until your experiment is marked "Ready to complete" before you choose a setting as the winner. Note that because we calculate progress based on the data that we’ve collected up until now, the number of days to go can go up, down, or stay the same over time.
How do I choose a winner?

To choose the winner of your experiment, you should:

  1. Review the performance of your original versus the variation you created for the experiment.
  2. If one branch has higher RPM and longer ad session lengths, we recommend you choose this one.
  3. If not, you will need to make a trade-off between the two, based on your judgment.

    In general, we don't recommend you apply changes that significantly reduce ad session length and only slightly increase RPM. Likewise, you might consider applying changes that slightly reduce RPM if it results in substantially increased session length.

Can I stop an experiment part way through?
Yes. You can end an experiment at any time by clicking Stop experiment on the experiment’s page. We’ll revert your settings back to the way they were before you started the experiment.
What if I don’t want to run an experiment?

If you see an opportunity on your Optimization page and you don’t want to run an experiment, you can choose to either click:

  • Apply now to implement our recommended settings
  • X to dismiss the opportunity.
Was this article helpful?
How can we improve it?