⚠ CHURN DEATH-SPIRAL 2 functions took churn in two straight sprints while running nothing — core.fn_•••a4, ops.fn_•••f7. Stop the bleed before you fund anything new.

[ LIVE DEMO / ENGINEERING ROI ]

Engineering ROI — what the spend buys, and what to do next

Commit churn joined with runtime value on the CodeNSM codebase itself: which code was expensive to build, which code the product actually runs on, and where the two disagree. Function names are aliased (svc.fn_•••2c) and authors are relabelled Engineer A/B; the churn, calls, value and quadrant math are the real distribution.

This is CodeNSM watching its own codebase — live, anonymized · read-only preview See it on your repo → How CodeNSM works

This page answers: what did the engineering spend actually buy? Build cost joined to runtime value — the efficiency grade, where the money leaks, and a ledger of every git-mapped function.

? How to read this office roles & states, in plain English

Roles — what kind of worker each function is

Project ManagerRoutes work to the right people — controllers, views, dispatchers.

Specialist EngineerDoes the actual skilled work — computation, scoring, business rules.

Records Clerk (reads)Fetches files from the cabinets — reads from the database.

Records Clerk (writes)Files paperwork into the cabinets — writes to the database.

Security & ComplianceChecks credentials and paperwork — auth, validation, permissions.

External LiaisonTalks to the outside world — third-party APIs, webhooks, HTTP.

Comms RunnerCarries messages out — email, Slack, notifications, publishing.

TranslatorTurns one language into another — serialization, parsing, formatting.

Floor SupervisorWatches the floor — health checks, status, metrics, monitoring.

FacilitiesKeeps the office clean — cleanup, expiry, retries, flushing, logging.

AssociateVersatile hands — no dominant role signal yet.

States — how each worker is doing right now

Fithealthy Snoozingunused 14+ days Growing fatlow value, barely used Stressedslow under load Injury-proneerror rate above 5% Quiet expertrare but critical Top performercarries outsized value

efficiency grade · 4/100

42% of build churn landed in code the product actually runs; 38% leaked into expensive dead weight. (518 churn-lines in framework entry points held out of the grade — invoked by the framework, not runtime-called.)

Score = share of build churn landing in code the product runs, minus the expensive-dead leak. Deterministic from the quadrant mix.

Executive summary

the 30-second read — computed live from this codebase

direct waste 5377 churn-lines

in 45 expensive dead-weight functions — high build cost, no runtime value.

capacity unlock 298 eng-hours

trapped in dead weight (71% of scored churn) — re-queue onto value without adding headcount.

core risk 5 load-bearing

expensive-and-valuable functions with the fewest commits — protect with tests first.

26 framework entry points held out of the grade — invoked by the framework, not runtime-called, so not counted as waste.

The reads

deterministic inferences — the joins say this, no guessing

[ WASTE ]

5377 lines of churn went into code with no runtime value (92 functions). Worst offender: core.fn_•••a4 — 173 lines over 3 commits, 0 runtime calls. Stop funding it, or find out why nobody calls it.

[ CONCENTRATION ]

115 functions are expensive AND valuable — the iterate-hard core. core.fn_•••a4 leads: value 8.9, 3 commits, 12d to stabilise. These deserve tests and review depth first.

[ ALLOCATION ]

Bandwidth split is uneven: Engineer A lands 50% of commits in high-value code; Engineer B lands 0%. That gap is a queueing decision, not a talent verdict — re-queue before you re-hire.

[ TREND ]

Share of sprint churn landing in high-value functions: 48% → 58% (rising — the roadmap is converging on value).

[ ACTION ]

Cheapest capacity is reallocation: move the dead-weight churn onto the ranked fix queue before adding headcount.

The money chart

99 commits analysed 537 functions mapped to git history 5377 lines of churn in expensive dead weight per-developer view →

Build cost × runtime value

efficient value

expensive value

cheap dead weight

expensive dead weight

cost → (churn × iterations, log scale) · up = more runtime value · hover a dot for the function. Trajectory (▲/▼) lives in the ledger below.

efficient value expensive value cheap dead weight expensive dead weight

Efficient value: 141Expensive value: 115Cheap dead weight: 163Expensive dead weight: 92

Dev bandwidth → business value

Engineer A 50%

Engineer B 0%

grey = all commits · gradient = commits landing in top-quartile-value functions · % = share of bandwidth on business value · authors relabelled Engineer A/B

Sprint time contributing to business

May 04May 18Jun 01Jun 15Jun 29Jul 13

Function ledger

every git-mapped function, costliest first — click a row for the business logic

Function	Quadrant		Commits	Churn	Runtime calls	Value	Days	Triage
core.fn_•••a4	Expensive value	▲	3	173	9	8.9	12	read-only
core.fn_•••a4	Expensive value	▲	3	173	3	3.0	12	read-only
core.fn_•••a4	Expensive dead weight	▲	3	173	0	0.0	12	read-only
core.fn_•••a4	Expensive dead weight	▲	3	173	0	0.0	12	read-only
core.fn_•••a4	Expensive dead weight	▲	3	173	0	0.0	12	read-only
core.fn_•••a4	Expensive dead weight	▲	3	173	0	0.0	12	read-only
core.fn_•••a4	Expensive dead weight	▲	3	173	0	0.0	12	read-only
svc.fn_•••6b	Expensive value	▲	4	83	548	548.0	2	read-only
web.fn_•••fe	Expensive value	▲	6	55	43	29.6	2	read-only
api.fn_•••57	Expensive value	▲	5	65	919	798.8	2	read-only
core.fn_•••57	Expensive value	▲	5	56	2	1.7	10	read-only
core.fn_•••c9	Expensive value	▲	4	50	8.5K	8476.7	2	read-only
api.fn_•••42	Expensive value	▲	5	36	93	74.7	13	read-only
data.fn_•••6c	Expensive dead weight	▲	1	171	0	0.0	0	read-only
svc.fn_•••97	Expensive dead weight	▲	1	171	0	0.0	0	read-only
data.fn_•••70	Expensive dead weight	▲	1	171	0	0.0	0	read-only
data.fn_•••08	Expensive value	▲	3	57	34	33.7	0	read-only
api.fn_•••9f	Expensive value	▲	4	42	74	73.9	2	read-only
api.fn_•••71	Expensive dead weight	▲	1	165	0	0.0	0	read-only
core.fn_•••ec	Expensive value	▲	1	165	235	235.0	0	read-only
ops.fn_•••01	Expensive value	▲	4	39	985	985.0	4	read-only
data.fn_•••ff	Expensive value	▲	3	52	83	50.6	2	read-only
core.fn_•••8d	Expensive value	▲	4	36	29	20.6	3	read-only
web.fn_•••b6	Expensive value	▲	4	35	8	7.9	0	read-only
svc.fn_•••9a	Expensive dead weight	▲	2	70	0	0.0	0	read-only
web.fn_•••20	Expensive value	▲	4	32	246	246.0	4	read-only
svc.fn_•••e2	Expensive value	▲	2	61	67.8K	67770.0	2	read-only
api.fn_•••75	Expensive value	▲	2	59	149	122.3	12	read-only
web.fn_•••1c	Expensive dead weight	▲	1	112	0	0.0	0	read-only
web.fn_•••1c	Expensive dead weight	▲	1	112	0	0.0	0	read-only

// Method: patch hunks are parsed for the functions they touch and matched to the functions this project tracks in production. Cost = churn × iterations per function; value = runtime contribution. Quadrants split at the medians. Dollar figures are estimates from an adjustable engineer-cost assumption, never presented as fact. In the public demo, function names are aliased and authors relabelled — the distribution is real; the identities are not.

// Triage decisions (delete / keep / test) persist in your own workspace, alongside the executive summary and per-developer coaching view.