Constraint Beats Capacity

Every conversation about American manufacturing eventually arrives at the same prescription: more. More shops in the directory, more machine hours on the floor, more people on the payroll. But I have watched a drawing packet sit untouched while the machine that could have run it stood open, because the RFQ record was incomplete, the vendor master was stale, and the request had gone out to suppliers who were never a real fit in the first place. The capacity was available. The work still didn't move.

This series is an attempt to turn soft failure modes into hard control points that a generalist and a builder can both execute. Here's this paper's version, stated plainly enough that you can prove me wrong: when quoting, qualification, and routing are the real bottlenecks, improving RFQ completeness, triage speed, and supplier verification will raise quote response rate and quote-to-award rate more than adding more suppliers will, because the system is constrained before production ever starts. By constraint I mean the step that most severely limits output. In a great many networks that step has nothing to do with the spindle, the press brake, or the welding cell. It's the front-end work of understanding the job, judging the risk, and routing it cleanly. Sometimes it is even simpler than that. Sometimes it is the will to do the work.

That's the plain lesson of Theory of Constraints. Identify the bottleneck, get more out of it, make the rest of the system serve it, and only then go buy more capacity. The American Society for Quality makes the same point without hedging: gross capacity analysis misleads you when the real queue is somewhere else. A full schedule board and an open machining cell can absolutely coexist, and when they do, a broken process is usually the actual bottleneck.

A manufacturing network makes this mistake constantly. Buyers assume a bigger supplier list means a stronger system, when a bigger list also means more screening work, more false positives, and more places to send a bad drawing packet. Choi and Krause, in their 2006 Journal of Operations Management paper, describe supply base complexity in terms of supplier count, supplier differentiation, and the interrelationships among suppliers, and they tie that complexity directly to transaction costs, risk, responsiveness, and innovation. Drop that into a live RFQ workflow and the practical consequence is obvious. If the supplier profile, the quote history, and the routing record are weak, scale produces chatter long before it produces throughput.

The quoting data points the same way. The Center for Automotive Research, in a 2025 update, found that automotive suppliers now process about 800 RFQs a year on average, up from 495 in 2002, while labor per RFQ climbed from roughly 134 hours to 157. The same report says cost data, BOMs, pre-RFQ tracking, resource availability, and supporting documentation are still fragmented across systems, which forces manual reconciliation inside the quote package itself. That is a front-end constraint in its purest form. A shop can be entirely willing and entirely able to do the physical work and still lose the job, because too much effort got consumed before the quote was even ready to send.

Buyers describe the identical problem from the other side of the table. In the IndustryWeek and aPriori survey, only 4% said their average sourcing and procurement process — specs to quote accept — takes a week or less, while most put it at a month or more. Another 60% named complex bid packages as the top reason for sourcing delays. And 79% said the ideal window is 24 hours to two weeks. Sit with that gap for a second. The first serious quote very often wins the work, and the bid package routinely arrives in a form too messy to move at anything close to that speed.

On the ground the failure is boring, which is part of why it survives. A drawing block omits the material temper. The STEP file reflects a newer revision than the print. The inspection plan calls for a CMM report nobody priced. The cert packet requirement is buried three lines deep in the PO notes. Every one of those gaps kicks the RFQ back into clarification, and every clarification burns days before a traveler exists anywhere. NIST has argued that efficient supply chain management reduces costs and delays that can add thousands of dollars to end-product prices, and the supplier-involvement research points the same direction: earlier and better supplier integration improves quality, lowers cost, and cuts time to market. We've talked about this before in this series. Offshore and opaque marketplaces absorb that ambiguity in a way that feels good up front and then reappears later as rework and expense.

So the implication is straightforward. Labor and energy matter enormously, but they are inputs, and inputs are not always the binding constraint. When the quote queue, the qualification gate, and the routing logic are the limiting step, more suppliers buy you more noise, more idle comparisons, and more late quotes. Throughput rises when the RFQ record gets cleaner, the vendor master gets verified, and the first routing decision gets disciplined.

None of which makes supplier count irrelevant. A broader network genuinely improves resilience, coverage, and surge response — once the request is structured well enough that each drawing packet reaches a plausible fit. Capacity added before coordination is relieved mostly inflates the inbox, because the system still can't tell a good option from a bad one fast enough to matter. The control point isn't the size of the directory. It's the quality of the route from RFQ record to serious quote.

That distinction matters for sovereignty. A country can have real machine capacity, real shops, and real workers, and still lose the work if buyers can't turn a drawing packet into a fast, trustworthy domestic quote. The failure mode is latency, and latency makes available capacity look unavailable. Which raises the next question, the one about allocation: once the front end is clean, how should the network decide who sees what work, in what order, and with what consequences when attention is scarce?

Implications

If the bottleneck sits at the quote desk, every additional supplier profile becomes a new demand on scarce attention. You see it as a higher clarification count, a higher no-response rate, and more hours spent sorting weak fits from real ones inside the RFQ record. The network looks larger. Usable capacity doesn't grow at anything like the same rate. The system gets louder before it gets faster.

Relieve the front end and the opposite happens. A cleaner drawing packet, a verified vendor master, and a narrower first route mean fewer dead ends and faster commitment. That doesn't eliminate the need for more labor, more energy, or more machine hours later on. It just puts those inputs behind a request that is actually ready to move.

Questions to Ask

1. Where in the RFQ record do we declare that a drawing packet is complete enough to quote?

1. How many clarification loops does a typical request generate before a quote is sent, and which missing fields in the quote package cause most of them?

1. How often do we route work to a supplier who later turns out to be a poor fit on process, commercial risk, or quality requirements, and how is that failure coded in the closure record?

1. Which facts still live only in inboxes, spreadsheets, or estimator memory instead of in the vendor master, quote sheet, or RFQ history?

1. Before asking for more suppliers, can we show that the current bottleneck is machine time rather than qualification time, routing time, or quote preparation time?