Evidence buyers should request from quality management system audit

By admin
The short answer

The verdict of a supplier audit matters less than the evidence behind it. Buyers should request the procedures, completed records, calibration traces, and corrective action files that show whether documented controls actually operate on the shop floor. A package with dated entries, signatures, and verification traces gives a defensible basis for approving a supplier; a package of blank forms does not.

Key takeaways

  1. Ask for procedures plus completed records with dates and signatures, because blank forms prove intent while filled records prove operation.
  2. Pull records across different months and lines rather than accepting a single tidy week, since sampling across time reveals whether controls held continuously [1].
  3. Check corrective actions for containment, root cause, correction, and verification, because retraining alone rarely prevents recurrence.
  4. Verify calibration certificates behind measuring instruments, as acceptance verdicts depend on equipment that reads correctly.
  5. Confirm that final acceptance evidence resembles factory acceptance practice, where conformance is demonstrated with records before goods are released [2].

Many buyers receive an audit summary with green ratings and assume the factory is safe to approve. The summary is only the auditor's verdict; the value for future orders lies in the underlying evidence, meaning the procedures, completed records, and verification traces that show whether controls operate day after day.

Suppliers sometimes present binders of procedures that look complete while the workshop follows different habits. The gap between paper and practice is where most quality escapes begin: incoming material accepted without checks, process settings changed without approval, or final inspection recorded before testing finished. Requesting the right evidence closes that gap before tooling money moves.

This guide lists the evidence a buyer should ask for after a system audit, explains what each document proves, and shows how to read the package without overstating what it means. It focuses on documents any well-run factory should already hold, so reasonable requests rarely require extra preparation time.

Start with the documents that prove the system runs

The core of any evidence request is the management set: the quality manual or its equivalent, the procedure list, and the master list of controlled documents with revision dates. These pages show whether the factory defines its controls in writing and whether operators work from current versions rather than outdated photocopies pinned to a workstation.

When the evidence package looks thin or inconsistent, many buyers commission a quality management system audit in China quality management system audit in China before placing the first production order, so that gaps surface while switching suppliers is still cheap. The request costs little compared with a failed first shipment, and it sets the expectation that documentation claims must be backed by records.

Read the master list against what the workshop actually uses. If operators follow revision C while the controlled copy is revision E, the document system exists only on paper. Ask how changes reach the floor, who approves them, and how obsolete versions are withdrawn, because version drift is one of the most common findings in otherwise tidy factories.

Records that show control, not just intent

Incoming material and in-process records

Incoming inspection records should name the material, the lot, the criteria applied, and the disposition for each delivery. Spot-check several months and look for rejected lots: a file with zero rejections across a full year usually signals recording pressure rather than perfect suppliers. Rework logs, process monitoring sheets, and first-article approvals complete the trace from raw material to finished unit.

Final inspection and release records

Final inspection reports should tie results to specific lots or cartons, record the sample drawn, and show who released the goods. Where sampling decided the verdict, the plan behind the sample size matters: the established reasoning behind inferring lot quality from limited samples is set out in treatments of acceptance sampling [1]. Release signatures without supporting measurements deserve a follow-up question, not automatic trust.

Illustration: Records that show control Decorative illustration for the section "Records that show control"; visual only, carries no data.

Evidence request checklist

DocumentWhat it provesRed flag
Controlled procedure list with revisionsControls are defined and currentOperators use outdated versions
Incoming inspection records by lotMaterial is checked before useA year with zero rejections
In-process monitoring and rework logsProcesses are watched and containedRework with no recorded cause
Final inspection and release recordsLots are verified before shipmentRelease signatures without data
Calibration certificates and recall listMeasurements can be trustedExpired stickers on final-test gear
Corrective action files with verificationProblems are fixed and checkedRetraining listed as the only fix

How to review the evidence pack without overclaiming

  1. Fix the scope firstConfirm the audit date, scope, product lines, and auditor independence before reading any finding, because a verdict from a narrow or outdated visit cannot support a broad approval.
  2. Pair paper with recordsMatch each procedure to at least one completed record from different months, so that paper controls are tied to operating proof across time.
  3. Test the corrective loopRead corrective files for containment, cause, correction, and re-verification, and downgrade any file that stops at retraining.
  4. Check acceptance disciplineTrace final acceptance records back to calibrated equipment and defined criteria, mirroring the factory acceptance principle that conformance is demonstrated with evidence before release [2].
  5. Record the gaps in writingWrite down which evidence was missing, set a deadline for it, and make continued approval conditional on receiving it rather than on promises.
Illustration: How to review the evidence pack Decorative illustration for the section "How to review the evidence pack"; visual only, carries no data.

Treat missing evidence as a finding, not as an administrative delay. A supplier that cannot produce calibration certificates or corrective files within two weeks usually does not have them, and approval granted in the meantime removes all leverage to obtain them.

Sources

  1. Wikipedia: Acceptance sampling —— Overview of acceptance sampling including AQL concepts used in product inspection.
  2. Wikipedia: Factory acceptance test —— Overview of factory acceptance testing performed before shipment or installation.

Frequently asked questions

How can a buyer tell whether a procedure is actually followed?

Ask for the procedure, a completed example, and proof that the record was reviewed. A blank form proves nothing; a filled form with dates, signatures, and follow-up entries shows the control actually operates. Where the example is missing entries or signatures, treat the control as unproven until the supplier closes the gap.

What calibration evidence matters most for product acceptance?

Calibration stickers alone are not enough. Request the calibration certificates or external laboratory reports behind the stickers, plus the recall list that shows which instruments were due and when. If measuring equipment used for final acceptance has no current certificate, dimensional and functional verdicts from that station carry extra uncertainty.

What does a convincing corrective action record contain?

Look for containment of the suspect stock, a documented root cause that goes beyond operator error, the correction applied, and evidence that the fix was verified on later production. A report that only retrains operators without addressing fixtures, instructions, or incoming material often repeats the same failure within a few orders.

Should buyers review every record or a sample?

Sampling keeps the review affordable and still representative, provided the selection covers different months, shifts, and product lines. Asking for every record of the year usually produces a data dump nobody reads; asking for a defined pull across time periods shows whether the system held up continuously. The reasoning behind inferring population quality from limited samples is the same statistical logic used across acceptance work.