navigaraResearch
IssuesAI ROILive dataMethodologyContact
Navigara Research·Published quarterly·2 editions to date

Engineering performance,
measured quarterly

A standing commit‑level study of open source engineering output.
One shared axis, one edition per quarter.

Read the Q2 2026 editionRead the Methodology

Cloudflare · Vercel · OpenAI · Google · Meta · Microsoft

Current edition

No. 02 · published 7 September 2026.

Two comparisons, two answers

Commit-level analysis across Cloudflare, Vercel, OpenAI, Google, Meta, and Microsoft. 699 qualifying engineers, six quarters. Performance per engineer is up 130% year over year, while the quarter-over-quarter change cannot be distinguished from zero — and, tested against a re-scoring of the same quarters, neither can the quarterly ranking.

+130%
performance per engineer, YoY
Q1 2025 – Q2 2026 · N = 699

Performance per engineer · % Δ vs. baseline quarter

Q1 '25Q2 '26
Read the full editionWhite paper · 13 pages · PDF

All editions

Every edition keeps the numbers it was published with. Nothing is restated after the fact.

No. 02
Q2 2026
Two comparisons, two answers
Q1 2025 – Q2 2026 · N = 699 · 7 September 2026
+130%Read
No. 01
Q1 2026
Engineering performance measured at scale
Q1 2025 – Q1 2026 · N = 676 · 30 April 2026
+116%Read

The next edition revisits the same cohort one quarter on, on the same axis. Ask to be told when it publishes.

Method

One measure, applied the same way every quarter.

Every edition scores merged commits with the same two‑layer engine: classification is file‑scoped across Features, Maintenance, Tests, Docs and Fixes, and deterministic algorithms weight the change. Their sum is Engineering Throughput Value. Holding the measure fixed is what makes one edition comparable to the next.

Read the methodology in full