Numbers Are Only Useful If They Are Defined

Almost every factory can produce a defect rate. Very few can explain what the denominator is. A defect rate of one percent means nothing until you know whether it is one percent of inspected pieces, one percent of produced pieces, or one percent of pieces that reached final inspection after the obvious rejects were removed upstream.

The purpose of a quality data audit is not to check whether the numbers are low. It is to check whether they are defined, consistent, linked to lots, and used to change something. A factory with a high but honest defect rate is easier to work with than a factory with a low number that nobody can reconstruct.

Definitions Come First

Ask the quality manager to define four terms in writing. The answers reveal the maturity of the system faster than any chart.

  • Defect rate: numerator and denominator, and at which inspection point it is calculated.
  • First pass yield: what counts as first pass, and whether rework is excluded.
  • Rework rate: rework events divided by units produced, or hours divided by hours.
  • Complaint rate: complaints per million shipped, or per order, or per customer.

If the definitions differ between departments, the factory has several quality systems pretending to be one.

Defect Rate and Its Denominator

Ask for the raw inspection records for one lot and recalculate the rate in front of the quality manager.

  • Units inspected, units rejected, and the inspection stage for each.
  • Whether inspection is full or sampled, and the sampling plan used.
  • Whether the rate is calculated per defect or per defective unit, since one unit can carry several defects.
  • Whether upstream rejects are included or excluded.
  • Whether the same lot appears in more than one stage and is double counted.

A common finding is that the reported rate covers final inspection only, while most defects are removed at in-process checks and never appear in the number. The rate looks good and the process is not.

Defect Categories and Pareto Discipline

Ask for the defect categories and the last three Pareto charts. Then ask which category received an action.

  • Category list is stable over time, so trends are comparable.
  • Categories map to process causes rather than to symptoms, for example "short shot" rather than "bad slider".
  • Photographs or boundary samples exist for each category, so inspectors classify consistently.
  • Pareto is produced monthly and reviewed with production, not only with quality.
  • At least one category has a documented improvement with before and after data.

A Pareto chart with no action behind it is a decoration. Ask what changed after the last one.

Rework and Its Hidden Cost

Rework is where quality data most often goes missing, because rework is uncomfortable. It consumes capacity, hides defects and distorts yield.

  • Rework log with quantity, defect, operation, operator and date.
  • Whether reworked units are re-inspected and how they are marked.
  • Whether rework hours are charged to production or recorded separately.
  • Whether rework is included in the defect rate or excluded from it.
  • Whether a rework trend is reviewed, and whether the same defect recurs.

Ask for the top three rework causes last quarter and the action taken on each. If the same cause appears in two consecutive quarters, the corrective action system is not closing.

Complaint Handling End to End

Pick one complaint and follow it from receipt to closure. This is the most informative hour in a quality data audit.

  • Date received, customer, part, lot and quantity affected.
  • Containment: what was stopped, what was inspected, and within how many hours.
  • Root cause: whether it was verified or assumed.
  • Corrective action: what changed in process, tooling, inspection or supplier.
  • Effectiveness check: whether the same complaint recurred after the action.
  • Response time to the customer, and whether the customer accepted the closure.

A passing complaint file ends with evidence that the defect did not recur. A failing file ends with a form that says "operator retrained".

Data Integrity and Traceability of Numbers

Ask to see how a number in the monthly report can be traced back to a source record.

  • Reports are generated from a system rather than retyped from paper.
  • Source records are retained and retrievable by lot.
  • Manual entries are signed and dated.
  • Changes to historical data are logged, not silently overwritten.
  • Monthly figures reconcile with shipment and production volumes.

Pick one month and compare total inspected units with total shipped plus inventory change. If the two do not reconcile, the quality data covers only part of the plant.

Feedback Loop and Cadence

Data that is not reviewed on a schedule does not change behaviour.

  • Daily defect review at the line, with a named owner.
  • Weekly quality meeting with production, maintenance and laboratory.
  • Monthly management review with trends and actions.
  • Whether customers receive a quality report, and how often.
  • Whether the factory can show a decision that was changed by the data.

The Actionable Audit Checklist

  • Written definitions exist for defect rate, first pass yield, rework rate and complaint rate.
  • Definitions are used consistently across departments.
  • Defect rate denominator and inspection stage are stated.
  • Upstream rejects are included in the reported number.
  • Defect categories are stable and mapped to process causes.
  • Boundary samples or photographs exist for each category.
  • Rework log records quantity, defect, operation and date.
  • Reworked units are re-inspected and marked.
  • Complaint files show containment, root cause, action and effectiveness check.
  • Reported figures reconcile with shipment and production volumes.
  • Data changes are logged rather than overwritten.
  • A documented review cadence exists with named owners.

What a Passing Answer Looks Like

Passing sounds like this: "Our defect rate is calculated per defective unit at final inspection and includes the in-process rejects, which is why it reads two point one percent rather than the zero point six we used to report." That answer admits the correction and explains the definition. YCC prefers that kind of number, because a supplier that has already corrected its own data is unlikely to be surprised by yours.