The Minimum Viable Work Package

At 4:12 p.m. the quote desk gets the usual bundle: a drawing PDF, a STEP, and one line that says "need it fast." The email subject says "Rev C," the title block on the PDF says "Rev B," and the STEP filename is "bracket\_final.step," which tells you nothing at all. The estimator opens the tolerance block, skims the notes for finish, and then stops at the same place every experienced estimator stops. You cannot price risk until you know which artifact governs, what evidence closes acceptance, and which requirements are real versus implied.

This series is a systems blueprint for coordinating American manufacturing without central command, by making work routable and verifiable across company boundaries using artifacts like drawings, travelers, and closure evidence. Paper 11 gave the network a canonical Request object. Paper 12 showed how free form RFQs become structured requests without forcing shops into a portal. This paper draws a line I have watched every good quote desk draw in practice, then makes it explicit so the system can enforce it and learn from it.

Paper No. 13's claim: when inbound requests are promoted to a Minimum Viable Work Package before quoting — meaning drawing and model precedence plus revision lock, material and certification requirements, tolerancing scheme, finish and special process requirements, inspection requirements, and shipping constraints are explicitly recorded against the drawing PDF and STEP fingerprints — the median clarification cycles per quote and the NCR rate attributable to spec ambiguity both fall compared to quoting from document bundles.

Definition: a Minimum Viable Work Package (MVWP) is the smallest version controlled packet of artifacts and declared constraints that lets a shop issue a quote and draft a traveler without hidden assumptions, while recording any remaining uncertainty as explicit questions or buyer approved defaults tied to the same revision locked packet.

A quote is a commitment gate, and the gate needs a packet-of-record

Every serious quality system treats commitment as a controlled moment, because after commitment you are on the hook for meeting the requirements whether you liked the inputs or not. ISO 9001's contract review concept is blunt about the direction of travel: review requirements before committing, retain documented information about what you reviewed, and control changes when requirements change. An estimator does all of this anyway, with a drawing PDF, a traveler skeleton in his head, and a mental list of the two missing facts that will bite him if he guesses.

The MVWP is that contract review gate made concrete for quoting, with one addition I think does most of the work: a packet-of-record. The packet-of-record is the controlled drawing PDF and STEP, plus a cover sheet that records the decisions the estimator had to make in order to quote honestly, and it carries a packet ID derived from fingerprints or a revision lock token. That packet ID shows up on the quote PDF, the traveler header, and the inspection report request. So when somebody in receiving asks "what did you build to," it is not a debate that starts over from zero.

The estimator's three decisions, and why they never change

In my experience, before an estimator can send a quote he has to make three decisions, and each of them maps directly onto a downstream artifact. First, what governs — drawing versus model precedence when the PDF and the STEP disagree, with a revision lock tied to the title block and file fingerprints. Second, what closes acceptance — whether the job requires an MTR in the cert packet, a CMM report, an FAI packet, or a specific report format that receiving inspection is going to enforce. Third, what can be assumed safely — which silent details the shop can fill in without creating a dispute later, written down on the cover sheet so the quote is not secretly priced on an invisible spec.

Those decisions are defensive work, not paperwork. They exist to keep a late surprise from turning into an NCR. If they are not recorded against the drawing PDF and the STEP, they get re-litigated when the PO arrives, again when the traveler is written, and again when a receiving inspector holds parts against an expectation that only ever existed in somebody's head. MVWP forces the decisions forward, binds them to the packet, and turns whatever uncertainty is left into a short, prioritized clarification list.

Machining: precedence and tolerance scheme decide whether the traveler is even coherent

In machining, the classic failure starts with a mismatch between the STEP geometry and the drawing's title block revision, then compounds when the tolerance scheme is either incomplete or read two different ways by two competent people. If the model has a fillet the drawing omits, or a hole pattern that does not match the PDF, the shop has to choose what to cut before a fixture plan even exists — and the wrong choice becomes scrap rather than a question. The MVWP cover sheet forces an explicit "spec of record" line plus the revision lock token binding the drawing PDF and STEP to the quote and the traveler, so a later "quick update" gets treated as a new packet version instead of a casual email attachment.

This is also where model based definition realities collide with small shop life. Standards like ASME Y14.41 exist specifically because digital product definition data sets can be the defining authority, and that authority has to be stated and controlled, especially when you are working from an annotated model with or without a drawing sheet. Meanwhile most custom job work still arrives as PDF plus STEP. I am not trying to win a standards argument here. The MVWP's job is to force the buyer to declare precedence and lock revision, because without that declaration the quote is a guess and the traveler is a liability.

Sheet metal: bend assumptions and grain direction create defects nobody sees until assembly

Sheet metal quoting looks fast right up until it hits the forming and finish assumptions the drawing packet never stated. Grain direction relative to the bend line affects cracking risk and bend integrity, and the difference between bending with the grain versus across it is not theoretical — it is a real driver of whether a flange survives. If a drawing PDF has tight bends and cosmetic requirements but never constrains grain direction, bend radius, or which face is cosmetic, the estimator can only price by guessing, and those guesses show up as assembly fit problems long after the parts have left the brake.

The MVWP makes those assumptions explicit in the packet-of-record and ties them to the traveler and inspection plan. The cover sheet records whether grain direction is constrained, whether bend deduction is fixed or shop standard, and whether the flat pattern comes from the STEP, from the PDF, or from the shop's own CAM export. Leave it out and the clarification loop still happens — just at the wrong time, on the floor, with parts half-formed, instead of at the quote desk with a clean drawing PDF and a stable packet ID.

Welding: symbols and examination expectations are what define acceptance

Welding is where acceptance criteria can be completely undefined even when the drawing PDF looks perfectly complete. A weld symbol can specify type, size, and other requirements, and the symbol system exists to communicate process and examination information concisely right there in the drawing block. But if the packet never states the governing code, the required documentation, and whether NDE is expected, the shop is pricing a job without knowing what "pass" means. Receiving inspection will define pass later, and it will define it in a way the shop cannot satisfy without rework.

The MVWP forces those symbol driven requirements into the packet-of-record and into the traveler before the quote goes out. The cover sheet records required process documentation, any WPS or welder qualification constraints if applicable, and the acceptance evidence required at close — even if that evidence is simply "visual inspection only," documented on an inspection plan. Fewer late fights, because the cert packet and any NDE report expectations were declared before anyone laid a bead.

Finish and special processes: the gap between "anodize" and a real spec is where rework hides

Finish is the quiet multiplier that turns a clean machining quote into a schedule blowup. "Anodize" is a category, not a requirement, and specs like MIL-A-8625 exist precisely because type and class matter, including distinctions like Type II versus Type III and whether dyeing or hardcoat behavior is intended. If the drawing note says "anodize black" but never states type, class, masking, and cosmetic expectations, the estimator will pad margin, ask questions, or lose money later when parts come back from finishing and fail an expectation nobody wrote down.

The way I would handle this is one hard rule and one relief valve. The hard rule: finish and special process requirements that affect fit, wear, corrosion, masking, or acceptance evidence must be declared on the cover sheet and bound to the drawing PDF. The relief valve: the system can propose a candidate finish spec extracted from the drawing notes, then ask only the single question that actually changes routing and inspection — "Type II dyed black versus Type III hardcoat" — and record the buyer's answer as a packet update rather than tribal memory.

What the system infers, what it flags, and what it hands to a human

A Minimum Viable Work Package does not get created by making buyers type more. It gets created by extracting what is already sitting in the drawing PDF, the title block, the tolerance block, and the notes, then enforcing a short decision list where the missing facts are made explicit. The system should auto-read revision, units, material callouts, tolerance blocks, and common process keywords, compare them across artifacts, and flag mismatches as blockers — above all the revision conflicts between subject line, title block, and file fingerprints.

When extraction cannot resolve a decision that changes liability, the request moves into a clarification state with a question that names the exact artifact conflict. "PDF Rev B, email says Rev C, STEP untagged, which governs" is a real question a real estimator asks, because asking it prevents a wrong traveler, a wrong fixture plan, and a downstream NCR. When extraction yields a safe default instead, the system records an explicit buyer approved assumption on the cover sheet. Assumptions that go unrecorded are the seed crystals for every dispute that ever started in receiving inspection.

Implications

A network that quotes from loose RFQ bundles will keep selecting for incumbents, because incumbents have shared context that substitutes for a packet-of-record, and that context does not travel. Quote desks will keep defending themselves with padding, silence, and phone calls, which are rational responses to an unstable drawing PDF, an untagged STEP, and unclear acceptance evidence. Buyers read that defense as "domestic is slow," even when the machine time is trivial, because the real bottleneck is contract review work happening in private instead of getting captured once on a cover sheet tied to a packet ID.

A network that enforces MVWP promotion is going to discover that a lot of its lead time problems were spec stability problems. Clarification cycles drop because the three estimator decisions get forced early and bound to the traveler and cert packet. NCRs tied to ambiguity drop because acceptance evidence is declared before the first chip is cut or the first weld is laid. And the whole thing produces a measurement loop that a generalist can follow: count revision churn events, clarification cycles, and ambiguity driven NCRs per packet-of-record, then improve extraction and question routing wherever the data points.

Outro

Sovereignty depends on being able to commit domestic capacity under stress, with enforceable requirements and verifiable closure, using drawings, travelers, and cert packets that do not drift as they cross firms. The practical failure mode is spec drift — the "real requirements" changing between quote, traveler, and receiving inspection because precedence, revision lock, and acceptance evidence were never bound to a packet-of-record.

In Paper 14 I take the next step and tighten closure, the cert packet and inspection evidence that turns "we shipped" into "we closed," so reputation can compound across transactions instead of resetting every time.

Questions to Ask

1. What is your packet-of-record today, and does your quote PDF and traveler header reference a specific drawing PDF revision and a STEP fingerprint that can be audited after an NCR?

1. In your last ten quotes, how many clarification cycles were caused by revision ambiguity, tolerance scheme ambiguity, finish ambiguity, or missing acceptance evidence like an MTR or CMM report, and do you record the reason code against the RFQ record?

1. Where do buyer approved assumptions live, such as grain direction, masking, sampling plan, or inspection report format, and can you point to the cover sheet or traveler note where that assumption is recorded?

1. When requirements change after quoting, do you treat it as a new packet version with a new validity window, or do you let changes smear across email attachments and side notes on the drawing PDF?

1. For welding and special processes, what is your minimum acceptance definition, meaning the weld symbol interpretation, any examination expectation, and the evidence that must ship in the cert packet for receiving inspection to accept the work?