Occasional-contributor wait

What it measures

Two things about occasional contributors, compared against regular ones: how long they wait for a first review, and how often their PRs get no review at all. The cohort is volume in this window — up to 2 merged PRs counts as occasional — which groups newcomers, drive-by contributors, and returning regulars who just had a quiet stretch alike. The clock starts at reviewableAt: time spent as a draft never counts, and it also restarts whenever a reviewer is first requested — so a PR that sat unnoticed before anyone was assigned reads as answered quickly, and the panel understates the gap rather than overstating it. Only a review from someone other than the author counts as an answer. The wait bars are merged PRs only, because only an answered PR has a wait to measure; the unanswered share also counts PRs closed without merging, because a PR nobody replied to has an absence to measure instead.

Why it matters

A regular contributor who waits an hour barely notices. Someone making their first or second contribution can wait days for a first review — and the worst version of that story is a contributor who quietly concludes the project doesn't want their help and never comes back. The wait bars cannot show that person: their PR was never answered, so it has no wait to record. The unanswered share is where they appear — the percentage of each cohort's PRs, merged or closed without merging, that drew no review from anyone but their author. A share that is far higher for occasional contributors than for regulars is the sharpest thing this panel says, because it counts the person who gave up rather than only the person who eventually got through. One loss stays invisible even so: a contributor who reads the room and never opens a PR at all leaves no trace in any number here.

How to read it

• The two bars are medians — regular first, then occasional — in hours to first non-author review.
• The p90 tail usually shows the gap more starkly than the median, but is withheld for small cohorts.
• The unanswered share is a footnote, one percentage per cohort, over every PR that ended in the window.
• A high unanswered share in both cohorts usually means the team merges routine work without review — read the gap between cohorts, not the level.
• A ratio below 1.0 is a real, good finding — the team answers occasional contributors faster, which is worth protecting, not a rounding error.
• The chart disappears — but the card and this explanation stay — when either cohort has fewer than the k reviewed contributors this window needs to stay anonymous, by design. The unanswered share can still appear underneath it, because it needs no one to have been answered.
• Thresholds are fixed: 2 merged PRs is the occasional ceiling, a 2× wait ratio or a 15-point unanswered-share gap flags attention, and a 0.5h baseline floor keeps a fast team from producing a meaningless ratio.

Anti-blame

Common misreads

• "Occasional contributors are slower" — the panel measures the team's response time, not their work.
• A small median gap means we're fine — a 1.2× median can hide a much wider p90 tail, or a much wider unanswered share.
• Reading the unanswered share as abandonment — it also counts merged PRs that shipped without review, which is a different, often deliberate, habit.
• Reading one window as a trend — cohort membership is window-relative, so compare like windows.
• Assuming every low-volume contributor is new — the cohort also holds returning regulars who had a quiet stretch.