> ## Documentation Index > Fetch the complete documentation index at: https://docs.codewarden.co/llms.txt > Use this file to discover all available pages before exploring further. # Measurements > How often CodeWarden's comments led to a change in the code, on your own repositories. The measurements page, `/orgs//measurements`, answers one question for your organization: **of the comments CodeWarden published, how many led to a change in the code?** If most comments are ignored, the reviews are not worth reading, and you should know. Every number on the page is **suggested from what GitHub shows**, never judged by the bot. No model is involved in counting. ## How a comment is counted Once a day, CodeWarden reads from GitHub what happened to each comment it published on your recent pull requests. Each comment lands in one of three groups: | Group | Meaning | | - | - | | **Changed the code** | The file changed after the comment, or GitHub moved the lines the comment pointed at. | | **A person replied instead** | Someone replied in the thread and the code did not change. Shown apart and **not counted as success**: a reply is as often "this is a false positive" as it is "fixed". Read the thread. | | **Ignored** | Nobody replied and nothing changed. | The headline is the share that **changed the code**. A comment published more than once on the same line counts once, since only the first could be acted on. Pull requests an owner [excluded](/configure/repositories#excluding-a-pull-request-from-measurements) are left out. The page says when it last read from GitHub. A comment on a pull request that is still open can move from "ignored" to "changed" later. ## What the page shows * **Window**: the last 30, 90 or 365 days. * **The headline** and the split between changed, replied and ignored, with the number of 👎 reactions and of repeated comments left out. * **Serious, and merged anyway**: `high` severity comments on pull requests that merged with nothing answered and nothing changed. Each is listed with a link. If they were wrong, the filter needs work; if they were right, they are worth a second look. * **Week by week**: the changed share for each week. * **By category** and **By severity**: where comments are acted on and where they are not. * **AI turns per week**. ## What it is not It is a suggestion from evidence, not a verdict. A file can change for reasons unrelated to the comment, and a comment can be right and still be answered with a reply. For a careful judgment, read the threads behind the numbers. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.