> For the complete documentation index, see [llms.txt](https://kerno.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://kerno.gitbook.io/docs/kerno-audit/read-the-results.md). # Read the results What the audit's numbers mean, how to tell a real change from noise, and what to look at next. Open **Audit** in the portal sidebar, or the link `kerno audit` prints when it finishes. The page shows the newest audit of each repository.
The Audit page for a repository, showing the period pickers, the Quality score and the headline metrics
### Choose what to compare Every number compares **Period A** with **Period B**. By default, Period A ends the month before agents arrived and Period B starts with it. When the audit couldn't date adoption, the history is split in half. Pick any two periods with the date pickers, such as the last quarter against the one before, to see whether a change is still moving. The reset button returns to the defaults. Switching repository resets them too. ### Read a number Each value reads **A → B**, then the change. Shares change in points (`−11 pts`), everything else in percent (`+479%`). **Green** moved the better way, **amber** the worse way. Some metrics have no good direction on their own, such as pull request size. These are **context** metrics, and the pages below mark them. The portal colours some of them anyway, so on those, read the numbers and leave the colour aside. Hover the info icon on any card for its definition. Many values open the commits behind them. ### When a value is missing or uncertain * **Not measured** or **Not measurable.** No value for the periods shown. Hover for the reason. * **too few to test.** Too few observations for a statistical test. Treat the movement as a hint. * **unjudged.** The run had no judge, so judged counts are what counting alone could settle. Read them as a floor. Read the caveats, marked with an amber warning icon, before you quote a number. They say when a month was left out, a sample was cut, or an input was missing. ### Worked example From an audit of one of Kerno's own repositories, with agents arriving in February 2026. Period A runs from February 2025 to January 2026, and Period B from February to early November 2026. **A metric that moved the better way.** **Rework** shows `39% → 28%`, in green. Fewer commits rewrote code written in the previous 30 days. Before calling it a win, open **Rework & Waste**. * **By month**, the rate dropped right after adoption, then climbed back to between 30% and 40% by the autumn. The Period B figure averages a low start and a high finish. * **Why later changes touched earlier ones** shows repairs falling from 23% to 7% of judged reworked commits. Period B rests on 14 judged commits, so treat that as a hint. Next step. On **Commits**, choose Period B and check the **Caused** column to see which commits rewrote recent work. If the last few months look like Period A, the improvement isn't holding. **A context metric.** **PR size** shows `34 → 197`. The median pull request grew almost sixfold. Read it beside **Throughput**, `82.7 → 28.6` merged pull requests per contributor, and **Production lines**, `8,884 → 25,473`. The team ships fewer, larger pull requests into a codebase that nearly tripled. Throughput covers the whole period, and Period B is three months shorter, so part of its fall is the shorter period. Larger pull requests are harder to review, so look at **Delivery** next. ### What each view shows * [Overview](/docs/kerno-audit/read-the-results/overview.md). The Quality score and the headline numbers. * [Quality](/docs/kerno-audit/read-the-results/quality.md). Complexity, deep nesting and duplicated functions. * [Delivery](/docs/kerno-audit/read-the-results/delivery.md). Review, time to merge, pull request size and throughput. * [Rework & Waste](/docs/kerno-audit/read-the-results/rework-and-waste.md). How often the team rewrites recent work, and why. * [Test coverage](/docs/kerno-audit/read-the-results/test-coverage.md). Which API routes have a test, and whether tests arrive with changes. * [Commits](/docs/kerno-audit/read-the-results/commits.md). Every commit behind the numbers. * [Repositories](/docs/kerno-audit/read-the-results/repositories.md). Every audited repository side by side. ### Share the results Every chart has a share menu to download a PNG or copy a link. The image carries the repository, both periods and the adoption month. `kerno audit` also writes `brief.md` on your machine, a short summary for a deck or a chat. The full `report.md` adds two readings the portal doesn't show yet. One compares AI-attributed and unattributed commits in the same months. The other checks whether each metric was already moving before agents arrived.