Visible method

Scoring config version 1. Size curve {"baseHours":8,"fileStart":20,"lineStart":1000,"steepness":1}.

• sizeCurve — expectedHours = baseHours * (1 + steepness * max(max(0, files - fileStart)/fileStart, max(0, lines - lineStart)/lineStart)). Leniency starts at 20 files or 1,000 lines. Continuous — no hard cutoff tiers.
• reviewResponsiveness — For each PR, score = 100 - 10*(actual/expected) when first non-author response is within the size-adjusted window, otherwise 100 * expected/actual. Category average is then normalized to the rolling org distribution (mean/stddev). Formal review completion and merge time are tracked separately; only first response drives this category.
• standupConsistency — Weekday completion rate from Geekbot timestamps only (no response content). A day counts complete if a report exists for that person on that UTC date. Time of day is ignored so time zones are not penalized.
• releaseQuality — Average of: merged-PR checklist completion, required-check pass rate, Dev-Tested confirmation rate (PR checkbox/comment or Linear comment/label, first event wins), and QA-Verified rate before Done.
• bugTrend — Inverted slackbug + internal bug open counts relative to org mean. Numeric/display/export issues are shown as a zero-tolerance line and included in the inverted count. Lower volume scores higher.
• collaborationBreadth — PROXY: unique authors reviewed + share of reviews that are cross-team, using comment-length/code-reference heuristics only as a depth caveat — not a precision metric. Normalized to org.
• prTurnaround — Share of PRs whose unmerged (or until-merge) business-day age is on track (≤1), with partial credit for slow (2–5) and none for extremely slow (6+).
• sprintPredictability — Average of planned-vs-completed story point rate and (1 - carryover rate) across cycles intersecting the selected window.
• unscored — Not scored by design: PR size trend, review load balance, production error rate, E2E flaky-test rate, AI self-review diligence, instruction-file upkeep.

Not scored by design: PR size trend, review load balance, production error rate, E2E flaky-test rate, AI self-review diligence, instruction-file upkeep.