Test one change at a time against the same revenue and retention numbers as the rest of Analyse. A result is only called once it is past the confidence bar and the minimum run is over.
Store prices, bundle contents, sales and discounts, when you point players at the store, the first session, and rewards and economy.
Revenue per player, store conversion, credits spent per player, or day 1 and day 7 return. Everything else is still tracked.
By default a winner is called at 95% confidence, once every arm has 100 players and the minimum run is over. A minimum run can grow but never shrink.
Each player is assigned once and stays in the same variant for the whole test. Up to 4 variants per test.
The metric over time per arm, confidence over time, and a breakdown by new or returning players and by traffic source.
End the test and ship the winning variant to every player, or end it and send everyone back to the control.
Your VIP rank sells well at $9.99 and you wonder whether players would still buy it at $14.99.
You run a store price test judged on revenue per player. The report tells you when the difference is real, and if the higher price wins you ship it to everyone from the end-test dialog.
Your team thinks the Roblox tutorial is too long, but cutting it might leave new players confused.
You test the short version against the current one on day 1 return. The segment breakdown shows how new players reacted, and the test is only called once the minimum run has covered both weekends.
You want to make crates feel more generous without hurting what players spend.
You test the new odds as a rewards and economy change, judged on credits spent per player, and see whether players spend more or simply get more for the same.
Name it, pick what kind of change it is, and add the control and up to 3 variants with a traffic split.
Choose the metric the call is made on and how many days it must run at least. The default is 14 days.
Ask Analyse for a player's variant from the Java plugin, the Roblox module or the JavaScript SDK for a web store. It is safe to ship this code before the test starts.
Servers pick the test up within a few minutes. The report shows confidence as players arrive and tells you when it is safe to call.
Create a store price experiment, then branch on the player's variant in your plugin. For a web store, the JavaScript SDK buckets players the same way as the game, so a player sees the same price in both.
Yes. The Roblox module returns a player's variant and its config, so your server script can decide what each player gets.
Analyse waits for enough players in every arm and for the minimum run to finish, then calls a winner only past the confidence threshold, 95% by default. Until then the report says how many more days it likely needs.
Experiments need the Pro plan or higher. Every new account gets 14 days of Pro, so you can set one up during the trial.
Free to start, 14 days of Pro, and about four minutes of setup.