AI unit economics and quota simulation
Compare AI cost per paying user with what they pay, see what free users cost, compare models and cache savings, and simulate a daily AI limit.
AI unit economics answers “what does AI cost me per customer, and is it worth it?”. It reads the AI calls recorded by @cockpitify/node (see AI cost per user) and, for the revenue side, RevenueCat.
Open it
- Phone: Margin tab → AI unit economics card.
- On the web: Money → Margin → AI tab. The period picker at the top applies to it too.
What you see
All amounts are turned into a monthly figure so they sit next to monthly prices.
- Per plan: for each product, MRR per paying user next to the AI cost per paying user (median and 90%), and the AI share of what they pay. A plan where AI takes more than a third is highlighted.
- Free users’ AI cost, and what share of premium net revenue it is.
- Models: calls, cost, cost per call, latency (median and 95%), milliseconds per output token, and how many calls were aborted.
- Cache: how much input was read from the prompt cache and what it saved, net of what writing to the cache cost. It can be negative: then the cache loses money for that model.
Quota simulation
Below the numbers, enter a daily limit per user: a number of calls, a dollar amount, or both, and tap Work out. Cockpitify replays the last 30 days and shows what the limit would have saved (premium and free apart) and how many people and days would have hit it. The line underneath shows typical daily use per person (median, 90%, 99% and the most), to help you pick a sensible limit.
To enforce a limit, use the daily AI limit.