Value Performance

What is it?

Value Performance analyses the individual changes within an experiment to let you know if they contributed towards an uplift, or negatively impacted your site.

Why would I use it?

  1. Gain deeper insight from your tests
    1. Understanding the ideas driving change on your page
    2. More information to tell a fuller story of how the experiment went
  2. Build up transferable learnings to take into future testing

Where can I find it?

On the main report page, you will find the value performance data underneath the candidate graph.

We have 2 approaches for calculating a value's performance - our team will work with you to identify which is most appropriate

Value Modeling

This approach uses linear modeling to calculate the expected performance of an individual change.

Pros:

  • It provides an expected relative uplift value for the change

Cons:

  • It doesn't (yet) take into account interaction effects between values

Note: As this approach uses a different methodology to evolution, and doesn't take interactions into account, it can sometimes give results that don't match the values promoted by evolution

Frequency Analysis

This approach reports on the interaction of these changes during evolution - showing which changes evolution promoted. It does this by analyzing how commonly changes appear in the best candidates.

Pros:

  • As it reports on the behavior of evolution, the effect of interactions between the values

Cons

  • It ranks the values but does not provide a prediction on whether or not the change is positive or negative

image__16_.png

image__15_.png

Have more questions? Submit a request

0 Comments

Please sign in to leave a comment.
Powered by Zendesk