Reputation That Compounds

The promise of this series is to turn soft failure modes into hard control points that both generalists and builders can execute.

Here's the scenario I keep coming back to. A buyer awards a small machined-part PO to a shop they have only heard about through a friend. The drawing package is clean, the STEP opens, the tolerance block is sane, the cert packet requirements are clear. And nobody in purchasing can answer the question that actually matters: when this shop gets busy, do they still ship on the promise date, answer the NCR, and close the loop when something goes wrong? The shop might be excellent. It might be a chronic source of late deliveries and containment work. The buyer cannot tell, because the evidence is trapped inside private inboxes, shop memories, and one-off broker dashboards.

Claim (falsifiable): If supplier reputation is captured as a portable, time-stamped performance record attached to the supplier master record and updated from real work events — RFQ response, PO commit dates, receiving results, NCR closure, dispute history, and repeat awards — then sourcing cycle time to unfamiliar but capable suppliers will fall and avoidable quality escapes will decline, because buyers will stop relearning performance through rumor and start routing work through recorded outcomes. A performance record means a machine-readable history of measurable behavior tied to actual jobs. Not a testimonial page. Not a star rating.

Paper 16 argued that identity has to be attached to the vendor master record so buyers do not waste quote cycle time asking who a counterparty is. This paper extends the same logic from identity to behavior. Once the identity packet answers who a supplier is, the next control point has to answer how that supplier behaves after award — and it can do that out of artifacts the system already touches: the RFQ log, the PO, the receiving record, the traveler, the CMM report, the cert packet, the NCR, and the corrective action closure. Reputation compounds when each job writes back into the supplier master record. The next buyer no longer starts from zero, and the next award does not depend on whether the right planner happens to remember the last fire drill.

The signals already exist

Manufacturing knows which signals matter, and they are not vague. Long-running supplier-selection literature consistently places quality, delivery, and performance history among the most important criteria, and current quality practice still evaluates suppliers through past performance, quality-system maturity, and the ability to meet the required delivery schedule. ISO's own guidance for process monitoring points to supplier performance, on-time delivery and lead times, failure rates and waste, and other measures of conformity, which is another way of saying that serious industrial evaluation has always been multi-dimensional and operational. A single star average takes late shipment, bad parts, weak containment, and poor communication and blends them into one blurred number, when the real work lives in separate fields on a scorecard, a receiving log, and a corrective-action record.

The builder-level evidence is even clearer. When a supplier starts slipping, the first warning never arrives as a bad review on a profile page. It shows up as a failed incoming inspection, an in-process reject, a customer reject, a field failure, or a recurring SCAR (Supplier Corrective Action Request) that keeps bouncing between containment and root cause. Trane's supplier quality manual explicitly ties SCAR issuance to incoming inspections, in-process rejects, customer rejects, field failures, packaging, and labeling issues, and it escalates on a clock — 24-hour acknowledgment, 14-day corrective-action plan, 30-day final report. Stanadyne's manual does the same with NCRs: 24-hour containment response, formal corrective action, third-party sorting for urgent production, and the risk of controlled shipping or a new-business hold if responsiveness and problem resolution remain weak. A reputation record that ignores those artifacts is ignoring the exact data quality engineers and buyers already reach for when a supplier becomes dangerous.

Portable reputation is not a theoretical invention, either. In federal procurement, past performance is formally treated as relevant information for future source selection, and the record includes conformance to requirements, schedule adherence, integrity, and business-like concern for the customer. CPARS is the official source for that past-performance information, and its evaluation areas are supposed to be supported by objective data and customary industry quantitative measures where applicable. DoD's SPRS makes the logic blunter still, tracking on-time delivery scores and quality classifications as part of responsible-award decisions. That is the basic pattern I want for industrial sourcing more broadly: performance should survive the last buyer's memory and travel as a usable input to the next award decision.

It should, because supplier history is predictive rather than decorative. A 2024 empirical study using panel data from 352 suppliers found that poorly performing suppliers were associated with more frequent and longer disruptions, and that disruptions hit quality in particular. NIST's guidance to small manufacturers gets at the same thing from the practical end — a balanced supplier scorecard should use weighted KPIs such as responsiveness and on-time delivery, and those scorecards are useful for measuring performance, justifying which suppliers to keep, developing better partners, and rewarding good performers based on objective data. A supplier who keeps acknowledging RFQs promptly, holds PO dates, ships conforming parts, closes NCRs on time, and earns repeat awards from multiple buyers builds a history that gets harder to fake and more valuable with every completed job.

Irony

Companies tolerate Xometry-style opacity on prototype work, and tolerate even more opacity offshore, because the purchase is framed as a bet on speed rather than a long-term supplier relationship. The buyer uploads a drawing packet and gets one portal, one PO flow, one invoice, and a broker who absorbs the search cost, so the unknown shop on the other side stays comfortably abstract — right up until the cert packet, the first article, or the NCR path gets messy. Offshore sourcing adds a second distortion. Many firms have spent decades treating IP leakage, weak traceability, and quality drift as a normal cost bucket inside a lower piece price, while domestic custom work gets judged shop by shop, where every new supplier feels like a direct personal risk to the engineer or buyer who signed the award. The way I'd even the field is to give domestic sourcing the same surface convenience without the blindfold: the buyer gets the fast path, and the supplier master record carries the identity packet from Paper 16 and the portable performance record from this one, so speed no longer requires ignorance.

Implications

Leave reputation trapped inside a buyer's inbox, a broker's private dashboard, or the tribal memory of one commodity manager, and every new relationship pays the same startup tax. The buyer places phone calls to ask around. The estimator hesitates on a promising RFQ. The schedule board gets padded because nobody trusts the promise date. Good suppliers without the right introductions lose work to familiar but mediocre incumbents. Slower routing, worse allocation, a weaker national supply base — all because proven performance does not travel to where the next PO needs it.

Make reputation a portable performance record tied to the supplier master record and buyers can finally separate unknown from risky. A supplier with clean receiving history, a low dispute rate, fast SCAR closure, stable on-time delivery, and repeat awards can win work faster even when the relationship is new. A supplier with recurring escapes, late cert packets, and unresolved corrective actions starts losing awards before the next major failure instead of after it. That changes the economics of trust. It rewards measured behavior rather than access to a middleman, and it makes the market more legible without pretending that all jobs, all buyers, and all failure modes are the same.

One caution, and it's the thing I got wrong early on. Repeat awards are a consequence signal, not the entire reputation system. A supplier can win repeat work because they are excellent, or because the buyer has no alternatives, a weak memory, or a lazy re-buy habit tied to an old PO. The durable record has to stay decomposed at the artifact level: RFQ acknowledgment time, promise-date accuracy, first-pass quality at receiving, NCR rate, closure time, dispute outcomes, and recovery behavior after a miss. In Paper 18 I'll take that up directly, because once the evidence exists, the next question is how to weight it, how to decay it, and how to keep one bad lot or one lucky streak from distorting the whole system.

A sovereign industrial base cannot coordinate well if it forgets how firms perform. The drawing packet, the cert packet, and the traveler describe the part; the supplier history tells you whether the plan is likely to survive contact with the calendar. The practical failure mode here is misallocation — work goes to whoever is known, not to whoever has earned trust through recorded behavior. The next paper turns that accumulated record into a control system, because portable evidence is necessary and still not sufficient. Evidence needs rules for weighting, thresholds, and escalation.

Questions to Ask

1. Which work events write back to the supplier master record today: RFQ acknowledgment, PO acceptance, promise-date changes, receiving results, NCRs, dispute closures, and repeat awards?

1. Are delivery performance, quality escapes, corrective-action timeliness, and communication responsiveness stored as separate fields, or crushed into one vague rating?

1. When an incoming inspection fails or a cert packet is late, does that outcome change supplier reputation automatically, or does it die inside a quality folder?

1. Can a new buyer see the same performance history that an incumbent buyer sees, or is reputation still locked inside private relationships?

1. What minimum evidence threshold, decay rule, and recovery path will prevent stale history or one-off events from dominating future awards?