Code review

Dashboard

Review outcomes and metrics

The dashboard measures outcomes, not comment volume: whether feedback was applied, what Ellipsis catches, and where the gate filters noise. The charts below use example data.

Did the review land

The share of suggestions applied, regions later edited, and fixes delegated back with @ellipsis. The north-star: did posted feedback change the merged code?
68%code suggestions applied after merge
54%comment regions later edited
12%@ellipsis asked to fix

Volume and severity

Posted comments over time, stacked by severity. A flat total with more blocking findings means the risk profile changed even if volume did not.
May 20Jun 3Jun 18
BlockingHighMediumLowTrivial

Calibration

The share of comments acted on, by severity. Blocking findings should land more often than nits; if they do not, tighten finding quality or instructions.
Blocking88% addressed
High74% addressed
Medium59% addressed
Low41% addressed
Trivial22% addressed

What it catches

Comments classified by type: logical bugs, security, performance, maintainability, testing, style, and more. Shows whether attention is on substance or nits.
Logical bug24%
Style19%
Security14%
Performance12%
Maintainability11%
Testing9%
Docs7%
API design4%

Where issues concentrate

A heatmap crossing comment type with severity. Bugs skewing high is healthy; style skewing blocking means severity labels need tightening.
BlockingHighMediumLowTrivial
Logical bug82241183
Security6141971
Performance2924162
Maintainability0318339
Testing0211216
Style0042824

From found to fixed

The funnel from findings surfaced, through the gate, to posted on GitHub, to addressed after merge.
Surfaced1,840 · 100%
Passed the gate74%1,360 · 74%
Posted to GitHub73%990 · 54%
Addressed62%612 · 33%