A change that isn’t propagatedisn’t a change. It’s a memo.
The revision exists in engineering while the floor cuts the old routing from the old drawing, purchasing confirms POs at the old rev, obsolete stock waits to be kitted, and the customer who mandated the change was never told it landed. This tripwire grades every open ECN on five propagation points and answers the question your change log won’t: are you building to the old print right now?
The change log says released. The floor says otherwise.
How long ECN-1210 has been past its effectivity date with old-rev floor stock still on the shelf — a customer-mandated change, seven days old, comfortably inside every SLA, and already the most dangerous line on the log.
Where a change actually lands: drawings, routers, open POs, floor stock, customer notice. A revision that landed in four of five didn’t land — the fifth is where the old print keeps getting built.
The SLA runs from release and catches the slow bleed. The gate runs from effectivity and catches the acute one. A change desk that only watches the first clock discovers the second one from a customer.
Seven ECNs, none stranded, verdict forced — live.
| ECN | Released | Effective | Mandated | Propagation points | Age | Verdict |
|---|---|---|---|---|---|---|
| ECN-1189 | 2026-03-30 | 2026-04-15 | 93d | PROPAGATED | ||
| ECN-1195 | 2026-06-05 | 2026-07-20 | 26d | IN FLIGHT | ||
| ECN-1201 | 2026-05-02 | 2026-06-01 | 60d | PROPAGATED | ||
| ECN-1204 | 2026-05-25 | 2026-06-10 | 37d | PROPAGATED | ||
| ECN-1207 | 2026-06-15 | 2026-07-15 | 16d | IN FLIGHT | ||
| ECN-1210 | 2026-06-24 | 2026-06-28 | 7d | IN FLIGHT · PAST EFFECTIVITY | ||
| ECN-1213 | 2026-06-29 | 2026-08-01 | 2d | IN FLIGHT |
Try it: disposition ECN-1210’s floor stock (tap it to 1) and the gate releases — the verdict falls to CHANGES IN FLIGHT and fix-first shows a dash. ECN-1213 is also mandated and also has an open point, but its effectivity is a month out: mandated alone doesn’t gate. Two clocks.
Same math as the engine and the workbook — byte-for-byte, evaluation date pinned identically in all three. Nothing here is scored by AI, nothing leaves this page, and the verdict will not soften because you wanted it to.
A change-desk read you can run in twenty minutes.
The engine is zero-dependency Python — nine columns per open ECN, a pinned --as-of date, and it prints every ECN’s state, the headline, the gate, and the one change to land first. The workbook reproduces the identical math with the same pinned date, and the demo above runs the same logic. This is the engine’s verbatim output on the shipped sample:
ECN PROPAGATION TRIPWIRE - ECN-079 ================================================================== Evaluation date: 2026-07-01 Open ECNs read: 7 Open propagation points: 5 ECN RELEASED EFFECTIVE MAND AGE OPEN VERDICT FLAG ECN-1189 2026-03-30 2026-04-15 no 93 0 PROPAGATED ECN-1195 2026-06-05 2026-07-20 no 26 1 IN FLIGHT ECN-1201 2026-05-02 2026-06-01 no 60 0 PROPAGATED ECN-1204 2026-05-25 2026-06-10 no 37 0 PROPAGATED ECN-1207 2026-06-15 2026-07-15 no 16 2 IN FLIGHT ECN-1210 2026-06-24 2026-06-28 yes 7 1 IN FLIGHT PAST EFFECTIVITY ECN-1213 2026-06-29 2026-08-01 yes 2 1 IN FLIGHT Stranded: 0 In flight: 4 Headline: CHANGES IN FLIGHT GATE: ECN-1210 is customer- or safety-mandated, past its effectivity date, and not fully propagated - parts can be made to the old print right now. VERDICT: BUILDING TO THE OLD PRINT Fix first: ECN-1210
Built so the verdict can’t be negotiated.
Landed or it isn’t
No completion percentage, on purpose — “87% propagated” is exactly the lie. Each point is 1, 0, or NA, and NA never counts as open: a change with no customer-notice requirement isn’t penalized for not sending one. The honesty lives in marking NA at release, not at the review.
Two clocks, deliberately
The SLA runs from release: open points past 30 days are STRANDED. The gate runs from effectivity: a mandated ECN incomplete on or after that date forces the verdict however young it is — the effectivity date is the day the old print became the wrong print. Worsen-only, and it releases when the points close.
Pinned, never drifting
The evaluation date is set explicitly in all three layers — engine flag, workbook cell, demo constant. Nothing reads today’s date, so the month-end read is a reproducible fact and re-running is a deliberate act, not a sheet quietly changing its mind.
The cross-system read your PLM can’t give you.
- A deterministic, offline read of nine columns per open ECN against a pinned date — engine, workbook, and demo produce the identical verdict from the identical log.
- The cross-system question answered in one number: your PLM owns the revision and your ERP owns the POs, but neither will say whether the change landed everywhere at once.
- A weekly discipline: the runbook installs the point-marking at release, the effectivity watch, and the stranded-queue burn-down.
- A PLM, an ERP module, or anything that connects to one. It reads a log you fill from both in twenty minutes; your systems execute the fixes.
- An engineer review. It grades change propagation — never the engineer who wrote the change or the buyer working the PO queue.
- A compliance determination. Change control can be contractual or regulatory; your documented procedures and customer requirements override this tool — full stop.
Scope note. This tool grades engineering-change propagation from a log you enter — it scores changes, never people, and it modifies nothing in your PLM or ERP. Change control can be contractual or regulatory; your documented procedures and customer requirements override this tool. Not a compliance determination, not certification of any standard, and not legal advice.
Anyone who’s been surprised by an old rev.
Engineering managers whose change log shows released while the floor shows otherwise
Job-shop owners who found old-rev stock in a kit and now wonder how many ECNs are half-landed
Ops leads who need the customer-mandated changes watched on the effectivity clock, not the SLA clock
Quality engineers tracing an escape back to a drawing that was updated everywhere except the router
Change boards that want the weekly review to open with one verdict and one named ECN, not a status tour
Teams inheriting an ECN backlog and needing an honest day-one snapshot at a pinned date
The configuration leg of the quality desk.
Cost-of-Poor-Quality Ledger
$99What building to the old print costs when it's caught: scrap, rework, and the customer chargeback, priced from your numbers.
PM Deferral & Pencil-Whip Tripwire
$89The same integrity audit pointed at maintenance — the PMs marked done that never happened.
OEE Honesty Grader
$99Whether the metrics feeding your capacity story are even real — before you plan around them.
Straight answers, before you pay.
Every open ECN against five propagation points — drawings, routers, open POs, floor stock, customer notice — each marked 1 (done), 0 (open), or NA (not applicable). Per ECN: PROPAGATED, IN FLIGHT (open points inside the SLA, default 30 days from release), or STRANDED (past it). The shop headline is the worst state on the log, and the total open-point count tells you the size of the backlog in one number.
The SLA measures from release; the gate measures from effectivity — two clocks, deliberately. A customer- or safety-mandated ECN that is incomplete on or after its effectivity date forces the verdict to BUILDING TO THE OLD PRINT however young it is, because the effectivity date is the day the old print became the wrong print. It's worsen-only — it never lifts a verdict — and it releases the moment the points close or the effectivity moves.
So the verdict on a dated snapshot is the same verdict forever. You set the date explicitly — --as-of in the engine, a yellow cell in the workbook, a constant in this page's demo — and re-run with a new date each review. A sheet that drifts with TODAY() changes its answer every time it's opened, which makes 'we were clean at month-end' unprovable.
Exactly — NA never counts as an open point, so a change with no customer-notice requirement isn't penalized for not sending one, and an internal-only ECN isn't dinged on points that don't apply. The discipline is in marking NA honestly at release, not discovering it at the review.
No. Your PLM owns the revision; your ERP owns the POs and stock. What neither will tell you is whether the change actually landed everywhere at once — that's a cross-system question, and this tripwire answers it from a nine-column log you can fill from both in twenty minutes. Run the read, execute the fixes in the systems you already have.
No and no. Deterministic and offline: the same log and the same evaluation date produce the same verdict in the engine, the workbook, and this page's demo, byte for byte. It grades change propagation, never an engineer. Your documented change-control procedures and any customer or regulatory requirements override this tool. Not certification and not legal advice.
The rev is released.
Prove that it landed.
One purchase, lifetime access, 12 months of updates. $79, once.
Grades change propagation, never an engineer. Change-control procedures and customer requirements override this tool. Not certification or legal advice.
Sold by RedHub AI LLC · Secured by Stripe · redhub.ai