Gross revenue (Adapty) less total cost (fixed subscriptions + revenue-share + variable usage and overages), projected to month-end. Figures come from the maintained registry and live signals; verify any flagged 'assumed'.

Per-vendor usage projected to month-end against each tier limit. For vendors whose billable usage we reconstruct from our own telemetry (Convex, Sentry, PostHog, Axiom) a crossover date means the run-rate breaches the tier this month. GCP shows actual cloud spend; Vercel and Expo forecast from a usage figure entered in the sheet.

Convex database I/O is the dominant usage line on the invoice, and the gauge above only shows the total. This attributes it to the function that caused it. Production traffic only — dev and prod share one telemetry dataset and dev is the louder of the two, so an unfiltered leaderboard would aim optimisation at throwaway backfills. Sort by read/call to find read amplification: a function costing far more than its call count suggests is usually an unbounded query or a missing index.

Every tracked vendor, its plan, and known monthly cost. 'Assumed' rows need a billing-page check. Edit the figures in the vendor-costs Google Sheet (or vendor-costs.ts as the fallback).