Conducting experiments

  1. Experimentation guidelines

The Yandex Advertising Network gives you tools to analyze statistics that help increase the efficiency of your ad units and revenue from your website. You can also experiment with ad units yourself.

Experimentation is a controllable method for researching some kind of phenomenon.

Results of conducted experiments serve as a confirmation or disproof of your hypothesis.

You can use experiments to find ways to improve your ad platform. For example, you can find out the following:

  • The optimal number of ads in a single ad unit.
  • How to place ad units on the page so that ads are noticeable but don't get in users' way.
  • What designs and formats blend seamlessly into the content of your platform.

Experimentation guidelines

One experiment — one hypothesis
When conducting one experiment, test no more than one hypothesis at a time. For example, test only a new block configuration or a different block position if you want to see how your revenue is affected.
Avoid times when significant changes are made to your website
Choose a period of time when there are no other changes being made to your website (such as changes to the design or layout).
Choose similar periods to compare metrics
Compare impression dynamics for the same periods. For example, compare viewable impression dynamics for January of the current year with the impression dynamics for January of the previous year. To learn more, see Seasonal statistical fluctuations.
Run the experiment for at least two weeks
Two weeks is the recommended experiment duration. Don't make changes to the experiment during this period of time.

To learn more about reports that help analyze the results of experiments, see Reports for evaluating the efficiency of monetization.