Microsoft Ads allows experiments based on cookies

Unlike experiments based solely on search patterns, tests based on the use of cookies allow the system to remember which version of the ad was shown to users.

This process achieves that a person is exposed to a single version of an ad, since the platform prevents the user from being identified as a new entry, each time they make a query in the search engine. This is how the audience is grouped based on their cookies and not based on a user's session.

Microsoft points out that this form of cookie-based segmentation is preferable when measuring the creativity of a piece, since when the user chooses an option, the system immediately prevents further impact.

How did it work before?

Previously, the iteration of experiments randomly chose which users to show the original campaign and test ads. This process was generated every time a user performed a search.

There was no record of how users had previously interacted, as is the case in the system based on the use of cookies to divide an audience. This meant that a user was hit multiple times by both the original and test campaign ads, each time they entered a search query.

How to allow splitting based on cookies?

From the Microsoft ad platform, you can enter the experiment icon.

Entering the module to "Create an experiment" select the campaign in which you want to run the test. Give it a name, start and end dates and what percentage of your audience you want to be part of this measurement.

The new cookie-based audience division settings can be found in the Advanced Options section:

The difference in the rate of experimentation

To obtain the level of accuracy offered by the measurement through the use of cookies, it is necessary to wait longer to obtain significant results in the tests.

Let's see an example of what happens in an experiment based solely on user searches:

If you have an audience of 10,000 users to do an experiment and each of them performs an average of 3 searches, you will get approximately 30,000 impressions to carry out your experiment. In this way, the system would give each ad 15,000 searches if the division is executed in half.

However, in the event that you perform the same test using cookies, the system only counts unique sessions, so to obtain the same 30,000 impressions that you would achieve with the previous system, you will have to wait three times as long. This will allow the smaller groups to take longer for the results to be statistically significant.

This cookie-based feature is available in all accounts on the Microsoft ad platform, so I recommend that you get to know it and do some tests to compare the results of the experiments you carry out on other platforms. This will not only give you more conclusive answers, but it will also allow you to reach some groups of users who do not browse traditional search platforms.