The product on staged data, with no signup.
There is no trial. These are the three main screens, filled with invented numbers.
Every figure below is made up. No customer’s traffic, revenue or experiment is shown on this page. Your workspace would show your own.
Every experiment, and whether it is settled
The list is the registry: what is running, the split, how long it has run, the lift and the p-value on one row. It shows which runs are ready to ship, which are losing and which are not significant yet.
Active experiments
The verdict
Each variant is joined back to the payment data it produced: authorisation rate, rebill rate, revenue per acquired customer, and tested against control. The winner is picked per metric, in that metric's own direction, and the recommendation is one sentence.
“direct” earns €2.15 more per acquired customer than control. The gap holds at p=0.02.
auth rate up 3.3pt, not significant · recommendation: ship it
The flags behind the variants
Every variant is a flag, and every flag is visible: which environment it is in, what fraction of traffic it holds, and which ones are stale because the experiment behind them ended and the flag was not removed.
Next step
Tell us what you run and what you want to test. We open a workspace and set up the first experiment with you.