What is the Kodus Review tab
The Kodus Review tab answers one question: is your team acting on what Kodus says?
Unlike the Productivity tab (which measures general delivery metrics like deploy frequency and PR cycle time), Kodus Review is about Kodus itself — how many suggestions get implemented, which categories and rules land or get ignored, and where the team pushes back.
Every chart and table respects the global repository and date range filters at the top of the page. Click a repository row, a category bar, or a rule to drill into the underlying suggestions.
Summary cards
| Card | What it means |
|---|
| Implementation rate | % of sent suggestions the team implemented in the period |
| Suggestions sent | How many suggestions Kodus delivered (as PR comments) in the period |
| Negative vote rate | Share of reactions that were 👎, with the trend vs. the previous period |
| Criticals ignored in merged PRs | Critical suggestions left unimplemented on PRs that were already merged — the actionable risk list |
Implementation rate
The core metric. A suggestion counts as implemented when its final status is implemented or partially_implemented at the time the PR closes.
Implementation rate = implemented suggestions ÷ sent suggestions
- Only delivered suggestions count (the ones actually posted as PR comments). Drafts Kodus filtered out before commenting are never counted.
- Suggestions are attributed to the week the PR closed, because implementation status is only final once the PR merges.
Week over week
The weekly chart shows the trend, with a toggle:
- Overall — a single implementation-rate line.
- By severity — one line per severity (critical / high / medium / low), so you can see whether higher-severity suggestions get implemented more.
By category and by severity
- By category — sent vs. implemented per suggestion category. Click a bar to open the suggestions explorer filtered to that category.
- By severity — implementation rate per severity level. The expectation is a descending gradient (critical implemented more than low). If it looks flat or inverted, severity isn’t guiding the team.
The “All / Kodus only” toggle on the severity chart matters. A Kody Rule carries the severity you set on the rule, not a risk assessment by Kodus. Mixing the two distorts the calibration read — for example, a pile of medium Kody Rules with high adoption can make Kodus look like it under-rates medium. Switch to Kodus only to see Kodus’s own severity calibration. Bars built on very few suggestions are faded and marked with * — a 0% or 100% from a handful of suggestions is not a real signal.
Negative feedback
Feedback comes from 👍 / 👎 reactions on Kodus’s suggestion comments.
- Negative vote rate (summary card) —
👎 ÷ (👍 + 👎), with the trend vs. the previous period. Lower is better.
- By category — where the team disagrees most. A category with many 👎 is a candidate to retune or disable.
- Trend — negative votes week over week.
Repositories — health
A per-repository table: PRs reviewed, suggestions sent, implementation rate, 👍/👎, and the weakest category (the category with the lowest implementation rate in that repo, given a minimum sample). It shows where Kodus is landing versus being ignored.
Clicking a row focuses the whole cockpit on that repository (same as picking it in the repository filter).
Kody Rules — health
How each Kody Rule is performing in the period: triggers, implementation rate, 👍/👎, and a status. Only active rules appear — deleted or inactive ones are excluded since you can’t act on them.
The status is computed per rule, in this priority order:
| Status | Meaning | Suggested action |
|---|
| Stale | No triggers in the period | Review whether the rule is still needed |
| Low data | Triggered, but too few times (fewer than 5) to judge | Wait for more data |
| Noisy | The team actively downvotes it (≥ 3 👎 and more 👎 than 👍) | The rule is miscalibrated — rewrite or scope it (e.g. exclude tests) |
| Ignored | Triggers a lot but almost nothing is implemented (≤ 20%) | Question its relevance — does the team care about it? |
| Healthy | Everything else | No action |
Suggestions explorer
Every drill-down — a category bar, a rule row, the “criticals ignored” card — opens the suggestions explorer: a filterable, paginated list of the actual suggestions behind the number.
Filters: repository, category, severity, implementation status, Kody Rule, and free-text search. Each row expands to show the existing vs. suggested code and a link to the PR comment.
