Skip to main content
xee.ai/case-studies/gamma

Gamma tests which plan to show first

One test, from the agent’s idea to a result you can ship.

This is an example. Gamma did not run this test, and the numbers are made up to show how a story reads. A real story uses the app’s own numbers, with its permission.

The agent saw it first

Yearly subscribers stay far longer, yet the paywall led with the monthly plan. The agent suggested a test, and said why.

Subscribers still paying after a year, by plan
Yearly44.1%
Monthly17.0%
Weekly3.4%
RevenueCat, 2025

The test

You approve it. The agent writes B and runs it for 14 days. The goal is revenue per paywall view, so a cheaper-looking plan cannot win by accident.

  1. Day 1You start it

    9,800 paywall views on each side. A shows monthly first, B shows annual first.

  2. Day 7B pulls ahead

    A 71% chance is not enough to ship, so the agent keeps it running.

  3. Day 14The test ends

    93% chance B is better. The agent asks whether to ship it.

app.xee.ai/apps/gamma/tests
Show the annual plan first Ended · 14 days
B · Annual first
9,800 paywall views
$0.37Ahead
A · Monthly first, control
9,800 paywall views
$0.31
Chance B is better
93%
Ship BKeep A

The result

B earns $0.06 more per paywall view, 19% more. Trial starts hardly moved, so the gain comes from more people choosing the annual plan.

Revenue per paywall view
B · Annual first$0.37
A · Monthly first$0.31
Paywall views that end on the annual plan
B · Annual first47%
A · Monthly first28%
Trial starts per paywall view
B · Annual first4.3%
A · Monthly first4.2%

An example story. Gamma did not run this test and the numbers are made up. The yearly, monthly and weekly figures in the first section are medians from RevenueCat’s State of Subscription Apps 2025.

What happens next

You decide. Shipping B changes the value in Xee, not the app’s code, so there is no release to wait for.

You decide

  • Ship B, and the paywall changes for everyone
  • Undo it with one click if renewals drop later
  • Pass on the next idea, and nothing changes
You approveevery change in the dashboard before it goes live

The agent keeps going

  • Keeps measuring B after you ship it
  • Suggests the next test, with its reason
  • Waits for your yes before it starts