CODEWITHSUNDEEP
statisticsab-testingconfidence-interval
ielashi
ielashi

Reputation: 245

Calculating AB Test Results

I have a web page where users are shown a set of products and they express interest in the products that they like by selecting them.

I am running an AB experiment to test two versions of this page, each version showing a different set of products. The metric that I am testing in this experiment is the number of products that are selected by users.

All the resources I found on AB testing assumed a "trials and conversions" model, where there are X number of trials and, from those trials, Y number of conversions occur. In this model, Y <= X.

In my case, I have X = # impressions of the page and Y = # products selected on that page where Y > X in some cases. How can I determine which version performs better in this case?

Upvotes: 1

Views: 456

Answers (1)

barracel
barracel

Reputation: 1829

For every variant you can divide the total number of objects by the number of impressions. That way you will select the variant that gives you more objects per user.

Upvotes: 1

Related Questions