Skip to content
2 minute readAPIAgent-ready

Test price experiments for a winner and split their traffic

Two pricing functions that were listed but could not be called now work: a significance test that says whether a price variant has won, and a traffic allocator that decides how to split the next visitors.

The API reference in dark mode listing POST /api/compute/significance and POST /api/compute/allocation

The Elo significance and bandit allocator functions were in the catalog but could not be called. Both now have their own endpoint.

  • POST /api/compute/significance takes each variant's visitors and conversions and returns each one's chance of being best, its expected loss and a 95% interval, with a verdict: winner, equivalent, or keep collecting data. With two variants it also runs a sequential test against the control once the control has at least 30 conversions.
  • POST /api/compute/allocation returns how to split the next traffic, giving more to the variants most likely to be best while every variant keeps a minimum share. Answers are exact by default, or sampled with a seed you choose.
  • The Experiments page shows which variant is winning under each experiment's results, with each variant's chance of being best. Looking at it is free.
  • Each call is billed with your other usage at the function's listed rate: 1x for the significance test, 2x for the allocator. It is not drawn from a prepaid balance. Both need a key that can compute prices.