Whitepaper · Sourcing & procurement architecture
They gave you a tour. You needed an audit.
A supplier visit that produces a good impression has produced nothing. The instrument is not the walk round the floor — it is the tracker that carries every finding from the last visit forward and forces an answer against evidence.
At the first visit the supplier told us the worst was behind them on engineering changes. Four months later, at the second site, a full-time external consultant had been engaged to do nothing but fix the change process. Nobody had lied. The first statement was an honest characterisation by people who wanted it to be true, and it went into the report as a finding because it was said by a senior person in a clean factory during a good conversation.
That is the failure mode of supplier visits, and it is not carelessness. It is that a visit produces impressions at a very high rate and evidence at a very low one, and the two feel identical while you are standing there. The fix is structural. You do not need better instincts on the floor. You need a document that makes an unevidenced claim visible as an unevidenced claim, months later, in front of the person who made it.
The tracker is the instrument
On any visit after the first, the centre of the report is a table with four columns: the issue, what was found last time, what was observed this time, and a trend. Green means improvement confirmed against evidence. Amber means open, or insufficient evidence to close. Red means worse, or unresolved. Everything else in the report supports that table.
The whole value sits in one rule about how the amber cell is used. A verbal assurance given at the last visit that has not been evidenced at this one is amber, not green. Not because the supplier is dishonest, but because the claim has not been tested and the report should say so. Apply that rule twice and the character of the relationship changes: the supplier learns which statements will be checked, and stops making the ones they cannot support.
An unevidenced assurance is amber, not green. “The worst is behind us” is not evidence.
Sections you do not get to delete
Structure is the method. Visit objective stated against a named delivery commitment. Executive summary in three to five sentences with a maximum of four top issues — four is the working limit, because if everything is a top issue then nothing is. Then the tracker, the facility, production flow, the quality system, supply chain, capacity plan against actuals, engineering change control, strategic context, escalation and recovery, action items, next steps, and an appendix of raw notes.
The temptation on a short visit is to drop the sections you did not get to. Do not. A section with nothing in it is itself a finding — write “not observed” or “access refused” and leave the heading standing. Six months later the pattern of what you were never shown is often the most useful thing in the file.
The probes that keep finding something
Across visits, a small number of questions have a much better hit rate than the rest. They are worth asking every time, in the same words.
What to ask, and what the answer means
- First pass yield by stage, and the denominator. Yield below about ninety-five per cent at any stage is a finding. But confirm what the number is measured against — first presentation, or after rework. Suppliers move between the two without saying so, and the comparison to last visit is then meaningless.
- How is conformal coating applied? Hand-brush is acceptable under IPC-A-610 for touch-up and rework. It is not a volume production method. If you see brushes on the line, you have found something the quality system did not report.
- Open change orders — count and age. A rising count is the leading indicator of engineering instability. An outside consultant engaged full-time on the change process is an escalation signal regardless of what the quality manual says.
- The approved manufacturer list at part level. Not the supplier-level vendor list. They are different documents and only one of them tells you whether a component can be substituted.
- Is firmware on the bill of materials? Where is it flashed, and who controls the release? In power electronics the shipped software build is routinely untracked.
- Whose certificate is this site on? Confirm the facility is inside the certification scope rather than riding on another site’s. A change in who owns product listings is worth following up.
The capacity plan is the part most likely to be wrong
Every supplier under pressure will present a capacity model: lines, shifts, days a week, resulting weekly rate, by period. It is usually arithmetically correct and operationally optimistic, and the optimism is concentrated in one cell.
The contract manufacturer’s contribution is the least reliable number in the plan. It arrives as a schedule from a third party who has every incentive to be encouraging and no exposure to your delivery commitment. Validate it against units actually built on the line, not against the schedule. Then name the binding constraint out loud — labour-limited, component-limited or test-limited — because the recovery action is different for each and a recovery plan that does not name the constraint is a hope with dates on it.
The other question to settle before you leave is how much of the supplier’s backlog you represent. Above roughly a third, you are a concentration risk to them, and their behaviour towards you starts being driven by that rather than by the contract. Ownership matters the same way. Pre-IPO, private-equity held and public suppliers each produce a predictable and different set of behaviours around margin protection and customer mix, and those behaviours will show up in your allocation long before they show up in a conversation.
What makes the report administrable
Every action item has a named person. Not a function — “quality” is not an owner, and an action assigned to a department is an action nobody has. Items carried forward from a previous visit are marked as carried forward, and an item that has been carried forward twice goes to the supplier’s executive sponsor without further discussion. That escalation should be automatic and known in advance, so that it reads as a process rather than as an attack.
Then archive the report before the next visit is planned. This sounds like filing advice and it is not. The tracker is only auditable if the prior report exists; without it, the baseline column is memory, and memory is exactly what the tracker was built to replace.
Archive visit one before you run visit two. Without the baseline, the tracker is just memory in a table.
None of this requires a large team or a long visit. It requires the discipline to write down what you were told, separately from what you saw, and to come back and ask about the difference. The supplier who is genuinely improving will welcome that, because it is the only mechanism that gives them credit for the improvement. The one who is not will find the tracker uncomfortable, which is also useful information, and usually arrives earlier than any other signal you were going to get.
That is the work I do.
Where this goes next
Who builds it, and whether the price is right →
The sourcing and contract-manufacturing method — where the audit becomes vendor-list admission, a scorecard and a gate somebody holds.
Right Tack Operational Consulting 