What an evaluation team writes down on site decides how much weight the final report can carry. Procedures describe intentions, but only dated production records, sampling evidence, and test logs show whether those intentions survive contact with real orders. This fifth part of the documentation series explains which records evaluators pull, how sampling and acceptance logic is checked, and what traceability evidence buyers should expect to see referenced in a serious report.
Key takeaways
- Completed production and test records matter more than blank forms, because only dated entries prove the documented process is actually followed.
- Sampling discipline should reference recognized acceptance sampling principles so sample sizes and accept decisions are consistent rather than improvised [1].
- Final acceptance routines gain credibility when their logic resembles structured factory acceptance testing, with defined criteria and recorded outcomes [2].
- Calibration labels, certificates, and out-of-tolerance handling decide whether any measured value in the report can be trusted.
- Traceability from finished units back to material lots determines whether a future failure can be contained to the affected population.
Factory evaluation reports are only as convincing as the documents behind them. A paragraph that says incoming inspection is performed means little unless a dated record shows it happening on real lots with real decisions. Buyers reading a report should therefore pay as much attention to the evidence appendix as to the findings themselves. The documents tell you whether the system described is the system used.
The fifth part of this series continues the walk through on-site documentation, focusing on the records that connect plans to reality. Earlier parts covered procedures, organizational records, and equipment lists; this part turns to sampling evidence, test records, and the traceability links that hold them together. These are the documents that separate a working quality routine from a binder of intentions.
Evaluators cannot read everything in a single day, so they sample documents the way inspectors sample products. The skill lies in choosing records that test the system's weak points: recent lots, changed processes, new operators, and subcontracted steps. This article explains what those records look like and how buyers should interpret what the report says about them.
From procedures to proof
Every factory has procedures; not every factory follows them. The evaluator's job is to close that gap by asking for completed records that correspond to the procedures on display. A work instruction for first-article approval should come with recent first-article sheets bearing part numbers, dates, measured values, and sign-off. A procedure for incoming inspection should come with delivery notes marked accepted or rejected and the test notes behind those decisions.
Timing matters as much as content. Records from last week describe today's factory far better than a showcase file from last year. Evaluators therefore ask for the most recent lots, especially ones that resemble the buyer's products in complexity and risk. A factory that can produce clean, complete, recent records without rehearsal is demonstrating its routine; a factory that needs hours to assemble a file is demonstrating its filing.
Gaps are findings too. Missing records for rework, concessions, or customer complaints often indicate the processes buyers most need to understand. An honest report lists what was requested but not provided, so the buyer can judge whether the gap reflects a quiet week or a blind spot.
How sampling evidence is checked
Almost every in-process and final check in a factory rests on sampling, yet sampling rules are rarely written down with the care they deserve. The evaluation team asks which rule governs each check and whether it is applied consistently. Reviews commissioned through UTS Inspection typically record the stated sampling basis alongside what operators were actually observed doing, so the buyer can see any gap between instruction and practice.
The statistical logic behind industrial sampling is well established: samples are drawn under defined plans, results are judged against acceptance numbers, and the plan's risks are understood by both sides [1]. An evaluation does not re-derive the mathematics, but it checks whether the factory's habits are compatible with that logic. Random draws, defined sample sizes, and written accept and reject criteria all point in the right direction.
Watch for three common weaknesses. First, sample sizes that shrink when production is busy, which means protection is weakest exactly when errors multiply. Second, retesting only the failed pieces until a pass appears, which converts sampling into wishful thinking. Third, mixing accepted and rejected material after the check, which erases whatever assurance the sampling created. Each of these belongs in the report with concrete examples.
Records the team asks to see
| Record group | What good looks like | Warning sign | Why the buyer cares |
|---|---|---|---|
| Incoming inspection | Dated entries per lot with verdicts and test notes | Deliveries booked in without any recorded check | Bad material entering silently becomes bad shipments |
| In-process checks | Hourly or per-changeover sheets with values, not ticks | Identical values copied down the page | Fabricated checks hide process drift |
| Final test logs | Instrument IDs, limits, and pass or fail per unit or sample | Pass stamps with no supporting data | Untestable claims cannot be verified later |
| Rework and concessions | Segregated quantities, causes, and re-verification | Rework disappearing back into good stock | Hidden rework returns as field failures |
| Complaints and corrective action | Closed-loop records with root cause and verification | Complaints logged but never analyzed | Repeat problems are priced into future shipments |
Acceptance logic and functional testing
Beyond dimensions and appearance, many products need functional or safety-related verification before they leave the site. The evaluation checks whether final acceptance follows a defined sequence with recorded criteria rather than a quick power-on and a glance. The structured acceptance routines used for industrial equipment offer a useful mental model: staged checks, agreed limits, and documented outcomes before handover [2].
On site, this means asking who defines the test sequence, how limits were set, what happens on failure, and where the results live. Strong factories keep test programs under version control, record every failure with its disposition, and trend the data for early warnings. Weak factories test with whatever program the last engineer left behind and keep results only when a customer asks.
Test equipment itself deserves attention. Benches crowded with unlabelled instruments, expired calibration stickers, and shared leads patched with tape all undermine the numbers the factory reports. The evaluator notes instrument identities, calibration status, and environmental conditions, because a test result without that context is just an anecdote.
Traceability: the thread that holds evidence together
Traceability is the ability to follow a finished unit backward through production to its materials and forward to its shipment. During an evaluation, the team picks one or two recent lots and asks the factory to reconstruct that thread from records alone. The exercise quickly reveals whether lot identities survive splitting, rework, subcontracting, and repacking.
Strong traceability has visible habits: incoming lots labelled and logged, internal transfers recorded, component reels tracked to work orders, and finished cartons marked with lot or date codes. When a nonconforming batch is found, quarantine records name the affected quantities and the review decision is written down. These habits cost little once established and save enormously when a field issue forces a containment decision.
Where the thread breaks, the report should say exactly where. Common breaks include bulk materials decanted into unlabelled containers, substitute components fitted without record, and mixed date codes in one carton. Each break widens the population that must be suspected when something goes wrong. Buyers should treat traceability findings as pricing information: the weaker the thread, the larger the hidden risk in every shipment.
Sources
- Wikipedia: Acceptance sampling —— Overview of acceptance sampling including AQL concepts used in product inspection.
- Wikipedia: Factory acceptance test —— Overview of factory acceptance testing performed before shipment or installation.
Frequently asked questions
Which production records should be ready for review?
Start with production and test records for current or recent runs that resemble the buyer's products: job cards, in-process check sheets, final test logs, and any rework records. These show whether the documented process is the one actually used. If the factory can only produce blank forms and no completed examples, treat the system as unproven regardless of how neat the procedures look.
How do I judge whether sampling is adequate?
Ask what sampling rule governs each check, how the sample size was chosen, and what happens when a sample fails. A credible answer names the rule, shows it written down, and produces records that follow it. Vague answers such as checking a few pieces whenever time allows suggest that inspection effort varies with workload, which is exactly when defects escape.
What calibration evidence matters most?
Look at whether instruments are uniquely identified, whether calibration labels show dates and due dates, and whether certificates are available and current. Then check that out-of-tolerance instruments are quarantined and that affected product is reviewed. Calibration is the foundation under every dimensional and functional number in the report.
How is traceability checked during a short visit?
Traceability connects finished units back through production steps to incoming material lots and bought-in component batches. Strong sites keep lot identities through splitting, rework, and repacking; weak sites lose the thread at the first decant or substitution. Without traceability, a failure investigation cannot find the affected population.