Multi-armed bandit experiments

[annikaschmid]

Feature request

Is your feature request related to a problem?

Experiments are often designed to explore which variant leads to a better outcome (e.g. conversion). In a multi-armed bandit experiment, the winning variant is dynamically rolled out to more people, to maximise for "success" as defined by your experiment.

Describe the solution you'd like

1. Set an experiment goal
2. Launch the experiment
3. As data comes in and it becomes clearer which variant is more successful, PostHog shifts more traffic to the winning variant

If you like this idea, please leave a 👍 or ❤️ reaction on this post to vote for it -- your votes and feedback help us prioritize what to work on next!

Please also let us know if there are any important functionalities that are not yet described in the initial list.