Skip to main content

​
Monetization Experiments (RevenueCat)

The agent already reads pricing intelligence (your entry-price percentile vs competitors) and cross-platform revenue. This capability lets it act on monetization instead of only observing it — via RevenueCat.

​
What the agent reads

Each run the collector reads synced RevenueCat overview + chart KPIs from revenuecat_overview_snapshots / revenuecat_chart_daily_metrics (nightly Trigger.dev sync — same pattern as ASC). Offerings and experiments are still enriched live when credentials are connected (best-effort, so a permission gap never blocks the revenue overview):
  • Offerings — every offering, which one is current, and package counts. This is the monetization write surface.
  • Experiments — the price/paywall A/B tests you set up in the RevenueCat dashboard, their status, and (for running/stopped tests) a summary of the leading variant.
  • Trial→paid conversion — the latest conversion rate, feeding the auto-hypothesis logic.

​
What the agent can write

One reversible operation: promote an offering to “current.” When a RevenueCat experiment is stopped with a clear, adequately-powered winner, the agent proposes a revenuecat_experiment action (operation promote_offering) that sets the winning offering as the project’s current offering — rolling the winner out to everyone. It’s reversible: promoting the previous offering rolls it back.

​
Always gated

A monetization write moves real money and is hard to undo cleanly, so revenuecat_experiment always requires your explicit approval — even at the highest autopilot level, and even if you’ve added it to auto-approve types. It never runs unattended.

​
What it deliberately does NOT do

  • It can’t create experiments. RevenueCat’s API has no experiment-create endpoint — you set experiments up in the RC dashboard. The agent reads and promotes winners; it doesn’t launch tests.
  • It can’t change App Store prices. App Store IAP price tiers live in App Store Connect, not RevenueCat. A raw price change is an ASC concern. The RC lever here is which offering serves, not the store price itself.

​
Auto-hypothesis engine

When the pricing percentile + trial→paid conversion suggest a monetization gap (e.g. you’re far below the competitor median with weak conversion, or far above with high churn), the agent opens a hypothesis (metricPath: rc.trial_conversion / rc.mrr) with a concrete test strategy rather than guessing a price. You then run the experiment in the RevenueCat dashboard, and the agent promotes the winner once the test concludes.

​
The loop

  1. Read — offerings, experiments, trial conversion, pricing percentile.
  2. Hypothesize — open a pricing/paywall hypothesis when the numbers warrant.
  3. Promote (gated) — when an experiment produces a winner, propose promoting the winning offering; you approve.
  4. Measure — the promotion is recorded as a pricing experiment in memory (expected metrics: MRR, revenue, active subs, trial conversion) and its outcome is reconciled into the decision log after the measurement window.

​
Requirements

  • RevenueCat connected (Settings → Integrations). Reads use your existing secret key; the write reuses the same credentials.
  • Experiments must be created in the RevenueCat dashboard for the agent to read results and promote winners.