Engineering Yield Dashboard — developer output per $1,000 of AI spend. Higher is better.
Subject-bound credential required. Views are logged.
▲
Show the unjoined developers
▲
▲
▲
Show the flagged outcomes
Org TIER—
Time to merge · PR created → merged (includes draft time)smaller or later-created PRs merge faster; not a quality measure
AI Spend—
Spend Leverage—
Capture Fidelity—
Request-level share of captured spend — not capture completeness, not attribution; cannot see spend in uncaptured tools
Where the unattributed spend goes
Cost composition
Talk about it in your next retro.
What’s driving our yield? These levers are team-wide and name no one —
use them to improve together, not to grade anyone.
By model
Where the points and spend went, by team
These rows split the org total by team. Before rounding, they add up to it. TIER is shown
for context, not as a ranking: task mix and how work gets sized differ between
teams, so read each team against its own history, not against the others.
Team
Pointsshare
List-price costshare
TIERcontext
Nominal 95% bounds. Coverage is unvalidated: in constructed tests where several outcomes share one issue, they covered the true value less often than 95%. Bounds exclude uncertainty from unattributed spend, incorrect attribution and uncaptured usage. Inference not established.
Per work type — a retro lens.
Which kinds of work yield the most per token? Discuss the patterns, not the people.
Panels are scoped per work type and each has its own scale — comparing TIER across types (e.g. security vs. feature) is unsupported by design.