The reassignment maybe shouldn't be random but it should look at the current "score" and decide which verison (a or b) should be shown so that the significance is increased. For super large experiments this will probably not matter, but I guess since you are building for indie founders, this would help to avoid "bad luck" scores

That would be cool, a quick summary by marketing channel to help you identify which are doing best

Right now you can only do it from the front end package

E.g. should be able to make an experiment with variant A, B, C

I'd like the ability to test the code locally in a mode where the cookie doesn't save (to make it easier to see the variants) and the visits/conversions aren't saved to avoid skewing the real test

Integrations with analytics platforms might be interesting. But more than that, I think some kind of a React context wrapper component in the NPM package might help front-end implementations.