Per-engineer Performance is moving
The stacked area is the 90-day trailing rolling average of per-engineer Engineering Throughput, broken into Features (net-new functionality), Maintenance (keeping the system honest), Tests (test suites and harnesses), Docs (documentation and guides), and Fixes(defects and reverts). The total height is the team's overall trend; the bands show how the mix shifts over time.
PER-ENGINEER PERFORMANCE
90-day trailing rolling average of per-engineer ETV, stacked by Features, Maintenance, Tests, Docs, and Fixes. Total stack height is the underlying performance trend.
Baseline
0.915 ETV/dev/mo
90-day rolling avg ending Apr 2025
Latest
3.22 ETV/dev/mo
90-day rolling avg ending Sep 2026
Change
+252.3%
from 0.915 to 3.22 ETV/dev/mo across the 90-day rolling series
How to read it
Each band is the 90-day trailing rolling average of that category's per-engineer ETV. Stacked, the bands sum to the total per-engineer Performance — so the top edge of the orange band is the team's overall trend line.
The change in the stat strip is the relative percent change in per-engineer ETV between the leftmost point of the chart (the first 90-day window we can compute) and the rightmost point (today's 90-day window). It is not a calendar-quarter average — both endpoints are single rolling-window readings, the same ones the homepage tile shows.
Why it matters
Per-engineer Performance is the cleanest signal that an AI-assisted, agent-augmented team is actually shipping more — independent of headcount. Hiring doubles the wage bill; per-engineer ETVdoesn't.
Teams pulling ahead on this metric are reinvesting their throughput into Features, not paying it back as accumulating Fixes. A rising stack with stable Fixes share is the pattern to look for.
How we measure 500 oss performance index, and how to read the trend.
Show methodologyHide methodology
How we measure 500 oss performance index, and how to read the trend.
Show methodologyHide methodologyWhat we count
Every merged commit on the benchmarked repositories is scored as a depth-of-work value rather than line count. Each commit splits into five buckets: Features (net-new capability), Maintenance (refactors, deps, infrastructure), Tests (test files and harnesses), Docs (documentation files) and Fixes (bug fixes, reverts, hotfixes). Classification is per file, so a test file counts as Tests even when it ships alongside a feature. Per-engineer means divided by the count of distinct authors who shipped that day.
How the rolling average is computed
The overlay line is a 90-day trailing mean of total per-engineer ETV. The chart starts where the rolling line warms up so every point has a real 90-day window behind it.
How the change is computed
We compare the per-engineer ETV at the first and the most recent point of the 90-day rolling benchmark series — i.e., the first and last values that drive the chart — and report the relative percent change between them. That matches the homepage tile, so the figures shown here line up with what you saw before clicking through.