How Org Scan estimates its figures
An Org Scan report is built from the history of your merged pull requests. This page explains each figure in the report email: what is counted, where the AI spend number comes from, and what the report leaves out.
The passing share
The headline is the share of your merged pull requests that passed every check. It counts pull requests merged in the report period, across all the repositories the scan covered, pooled together.
A pull request passed every check when each automated check on its final commit finished and none failed. Checks that finish as success, neutral or skipped count as passing. A pull request with no checks, or with checks still running, does not count as passing. The scan reads the first 100 checks on each pull request.
The share is out of the pull requests the scan examined, and the report says how many that was. If a scan covers more pull requests than it reads, it reads an even sample and says so.
AI spend
The report shows AI spend from the best source available, in this order, and labels which one it used:
- From your usage key. Measured spend read from your Anthropic usage and cost reports, for the period.
- From the figure you gave us. The monthly spend your CTO or admin entered when running the scan. It is not verified.
- Our estimate. A range worked out from the pull requests that carry an AI tool mark, using public list prices and a typical amount of AI work per line changed. It is not billing data. When too few pull requests carry a mark, the report shows no spend line at all.
Pull request size
Size is lines changed in one merged pull request: lines added plus lines removed. These are the size bands we define. Reports do not yet show anything by size.
| Size | Lines changed |
|---|---|
| Small | Under 50 |
| Medium | 50 to 500 |
| Large | Over 500 |
Cost per merged pull request and cost by pull request size are not yet available.
What is left out
- Pull requests that were closed without merging, and pull requests merged outside the period.
- Reverts. They are counted as merged but not read or checked.
- Pull requests over 20,000 changed lines, and pull requests the scan could not read. The report counts these and says how many.
- Lines in lockfiles and generated files, which are not counted toward size, so a dependency update does not look like a large pull request.
- Anything inside the pull requests: the scan does not read code, titles, descriptions, comments, branch names or author names.
Limits
This is an estimate from pull request history and, where you provide it, your usage data. It is not a measurement. Lines changed is a rough stand-in for effort, and a share of passing checks is not a judgement of code quality.
What Mnemom keeps
Mnemom keeps no copy of these numbers. Our email provider keeps a delivery log. See our Privacy Policy and sub-processors list.
