Every pricing experiment we run has three parts: the hypothesis, the measurement plan, and the rollback path. UsageBox keeps the mechanics manageable so we can focus on whether the idea actually helped customers.
Start With a Single Metric
We limit each test to one lever, maybe a higher included quota or a new overage price. Touching multiple levers at once makes it impossible to understand what worked. The ProductLed team has a good primer on this discipline.
Version the Catalog
UsageBox keeps every product and plan version in Firestore. We clone the current plans, edit the variant, and assign only a pilot cohort so existing customers are unaffected.
Collect Behavioral Signals
- Acceptance rate: how many eligible customers opt in or upgrade.
- Quota usage: are people hitting ceilings faster or slower than before.
- Support load: do new terms create extra questions for the team.
Automate the Rollback
Before we launch, we schedule the revert job. If the experiment underperforms we run a script that reassigns the prior plan and credits any differences automatically.
Share the Results
We publish a short Loom and Confluence note summarizing the outcome. Sales, success, and finance all need the context, and future experiments are faster when we have an accessible archive.
None of this requires new deployments, UsageBox handles the calculation work while we stay focused on learning which pricing model actually fits customer behavior.