Guide
Estimating the operating cost of an embedded analysis feature
Monthly cost is roughly runs per month times charge per released run. Under self-serve billing the charge is twice measured model cost; recorded development runs measured about $0.35 to $0.53 in model cost. Measure on your own material before committing.
For product managers and finance partners. Written by the Encevra engine team. Reviewed 2026-10-12.
The formula
Estimated monthly charge = released runs × (2 × model cost per run). Incomplete and failed runs are not charged under self-serve billing but still take time and user attention.
released_runs = monthly_runs * release_rate monthly_charge = released_runs * 2 * model_cost_per_run
Where the inputs come from
Model cost per run is reported in usage.cost_usd on every finished run. Release rate is something you measure: every recorded development check on the catalog so far ended incomplete, so do not assume a high rate.
Contracted plans
Contracted Build and Scale plans have per-run prices and monthly minimums published on the engine page. Those are agreed terms, not self-serve billing.
A worked estimate
Suppose 1,000 runs a month, a release rate you measured on your own material, and model cost per run read from usage.cost_usd. Multiply released runs by twice that cost to estimate the self-serve charge. Replace each number with your own measurements.
- monthly_runs = 1000
- release_rate = your measured value
- model_cost_per_run = from usage.cost_usd
Frequently asked questions
- Where do I see the cost of a run?
- Every finished run reports usage.cost_usd.
- Are there spending limits?
- Yes. Per run and daily provider spending caps are enforced before model calls.
- Is there a contracted option?
- Yes. Contracted plans have agreed per-run prices and monthly minimums.
