A supplier saving isn’t a production saving just because the purchase-order total went down. It becomes a production decision when that supplier is the only party who can put a conforming input in front of your constraint before the line needs it.
Call that test an executable alternate. It isn’t a second name in the ERP. It’s a source that can make the approved part, with the current revision, using available tooling, inside the interval between “the problem is known” and “the remaining work stops.”
That’s a much less glamorous definition of resilience. It’s also the one that keeps WIP from becoming a museum of almost-finished orders.
The Saving
Start by separating your price field from the operating result. A lower quoted unit price can be useful. It isn’t yet realized savings.
Your procurement record needs three different labels:
- Price reduction is the lower quoted unit price against a comparable baseline.
- Cost avoidance is a future cost prevented, such as a proposed increase not accepted.
- Cash realized is the change that has actually reached accounts payable after qualifying purchases.
That distinction matters because the saving can be financed by a larger minimum order, shorter payment terms, buyer-paid freight, looser acceptance criteria, slower replenishment, or a removed service allowance.
None of those moves is automatically wrong. They’re connected terms, and a price report that sees only one of them is just a very optimistic receipt.
NIST’s Manufacturing Extension Partnership separates total cost of ownership from purchase price, and treats critical or custom components differently from standard parts and routine maintenance supplies. That’s the right starting split. A catalog fastener and a drawing-controlled part can carry the same annual spend and completely different continuity risk.
Your procurement scorecard should preserve the baseline, comparable scope, realized quantity, measurement period, and approval trail. It should also preserve the original promise date and any accepted revision.
Measure only against the revised date and you reward a supplier for moving the goalposts with excellent penmanship.
The Alternate
Supplier count is a poor proxy for resilience. Two nominal vendors may share one factory, distributor, raw material, transport lane, cloud region, or other failure domain. That’s two purchasing records and one interruption.
Test every critical input you have as its own dependency. An alternate passes through distinct states:
- Identified means a plausible source has been found.
- Sampled and accepted means the required part has passed your inspection and documentation gates.
- Contracted and recently used means the source, terms, lead time, ordering path, and actual performance remain usable.
The last state is the one you need. Current drawings, tooling access, approved samples, workable lead times, and order history are the evidence.
An alternate last tested against an obsolete revision is lapsed, not qualified. A strong price and responsiveness score can’t repair that gap, because the mandatory gates — conformity, required documentation, and approved-source status — come before any comparative scoring.
Qualification time stays outside ordinary delivery lead time. NIST Supplier Scouting reports that its nationwide supplier search typically takes 30–45 days to return results. That service isn’t your promised sourcing window, but it explains why “we can find someone” isn’t an answer when remaining stock covers the next scheduled work.
Your urgent supplier search begins after the line is already exposed. The qualification file should exist before then.
The Register
Make your critical-input register small enough to inspect and specific enough to change a release decision. It isn’t a supplier master cleanup. It’s a production-control document.
| Critical input | Affected jobs | Supplier share | Lead time and boundary | Approved alternate status | Tooling access | Stockout consequence |
|---|---|---|---|---|---|---|
| Drawing-controlled part | List released orders, backlog releases, and the constrained operation | Share of affected work, not only category spend | Record the starting event and ending event for acknowledgement, material, production, inspection, dispatch, and acceptance | Identified, sampled, accepted, contracted, recently used, or lapsed | Record owner, location, revision, access, and ability to run it | State which release, WIP stage, shipment date, or machine slot stops |
| External process | List jobs whose route cannot bypass the process | Share of affected work and qualified capacity | Separate dispatch, outside processing, return inspection, and acceptance | Record current approval and recent conforming use | Record controlled files, fixtures, gauges, and process approval | State the jobs held between operations and the customer dates exposed |
| Standard but high-runner input | List products and releases that consume it | Record supplier and failure-domain share | Record normal, expedited, and recovery intervals separately | Record the source’s actual capacity before remaining stock is exhausted | Record any special pack, program, or data dependency | State the point at which the constraint is starved |
That table is deliberately not a red-amber-green risk heatmap. “High” doesn’t tell your planner whether a release can start.
The register does. It names the jobs, the source’s status, the technical proof, and the consequence of being wrong.
The Release
You don’t solve a material problem by releasing more work. You hide one.
Before a manufacturing order enters WIP, the release gate checks material availability, drawing revision, tooling, first-operation capacity, customer credit, and any required deposit.
Those fields answer different questions. Material availability doesn’t prove the current drawing is loaded. A fixture on a shelf doesn’t prove the alternate can use it. A free machine doesn’t produce a part waiting for outside processing.
This is where procurement and manufacturing economics touch. WIP starts consuming liquidity when material is bought, labor is booked, or a subcontractor is paid, and it stays a claim on cash through inspection, shipment, invoicing, and collection. Under IAS 2, inventory is measured at the lower of cost and net realizable value. The standard doesn’t turn a physically busy shop into saleable output merely because the job has accumulated cost.
Nor does utilization authorize release. The constraint determines additional saleable output.
Starting a job to keep an upstream resource busy can put more work ahead of the bottleneck, conceal the shortage, and lengthen the interval before any customer order ships.
Protective capacity can look idle in a utilization report while it’s functioning as insurance against late material, a failed tool, inspection rejection, or recovery work.
NIST has documented both sides of that control. One scheduling case cut reported lead time from 12–14 weeks to 5 weeks after correcting routings, configuring work centres, and scheduling around a welding constraint.
Another used a feeder-line kanban and standard work to raise daily output from 75 to 120 units while WIP fell from 717 to 156 pieces.
Neither result is a universal target. Both show that flow improves when release and routing reflect the actual constraint rather than the desire to look busy.
Find the Broken Gate
Use the visible failure to identify your broken stage. Don’t start by adding vendor names. That’s how a database grows while the line stays single-sourced.
One supplier delay stops multiple jobs already on the floor → dependency mapping is broken
Concentration was measured by spend or vendor count instead of the number of production-stopping parts, affected releases, and constrained operations. Map the input to the jobs it can stop and record the share of affected work.
A backup appears in the system but can’t deliver a conforming part → qualification is broken
The drawings, tooling, approved samples, required documentation, workable lead time, or recent-use evidence has lapsed. Treat the source as unqualified until it passes the required gate. A weighted supplier score doesn’t waive conformity.
WIP rises while shipment dates slip → release readiness is broken
Material availability wasn’t linked to the constraint schedule, or the route omitted inspection, external processing, dispatch, return, or acceptance. Follow the traveler’s current operation, not the schedule you hoped for. Shop-wide scheduled utilization isn’t order progress.
Delivery reporting looks good while an essential line is missing → the denominator is broken
On-time and in-full are separate measures, and OTIF requires both. Score at the order-line level when one missing line prevents a whole order from being used, and keep the original commitment beside any revised date.
The supplier score is strong but a defect turns up in use → the acceptance loop is broken
An escape is a nonconformity accepted at receipt and found during use or by your customer. Record it separately from receipt rejection, tied to the purchase-order line, revision, lot or service milestone, disposition, and corrective action.
Savings rise on the report while inventory, freight, or payment pressure worsens → the baseline is broken
The comparison treated volume, specification, service, freight, or payment timing as constants when they changed. Separate negotiated opportunity, comparable unit-price effect, cost avoidance, and cash realized before you call the result a saving.
The Test
The recovery test is the missing step. You don’t need an annual crisis rehearsal with laminated badges. You need one controlled proof that the alternate can take the work before the incumbent becomes unavailable.
Run the test through the same records you use for production:
- Send the current drawing, revision, inspection plan, and required documentation through the alternate’s ordering path.
- Confirm tooling ownership, physical access, program or file control, and the source’s available capacity against the required interval.
- Inspect the sample or limited order, retain the acceptance evidence, and update the register with the recovery-test date and the next condition that would lapse it.
For active vendors, a rolling interval with the underlying counts makes sense. For a supplier that delivers once a year, review after delivery and at renewal rather than producing a monthly score full of empty precision.
NIST recommends a compact vendor-specific scorecard with two or three key measures such as risk, quality, and on-time delivery. Put the continuity evidence beside those measures. Don’t bury it inside a single weighted grade.
FAR 42.1503 uses a similar evidentiary pattern for federal contractor evaluations: objective facts supported by contract-performance data, factors tailored to the contract, and narratives supporting each rating.
You aren’t governed by that system. The practical lesson still survives. A continuity claim without transaction evidence, technical evidence, and a short explanation can’t be relied on at renewal or during a shortage.
The cheapest source is a price. An executable alternate is capacity. A supplier list is administration. A recovery-tested register is production control.
Choose the part that can stop the most of your work. Prove the alternate can make it. Then negotiate the saving.