Define PCB Defect Codes Before Ranking a Pareto Chart

A Pareto chart is only as useful as its categories. Mixing a symptom such as “open circuit,” a cause such as “insufficient paste,” and a disposition such as “rework” creates bars that do not represent comparable events. Define the defect record before ranking its frequency.

Minitab’s Pareto chart guidance describes ranking defect categories and optionally combining the remainder into an “Other” group. The software can organize counts, but it cannot decide whether the underlying category definitions are consistent.

Separate what was observed from why it happened

Record the observed condition, board location, detection method, suspected cause, confirmed cause, and disposition in separate fields. A suspected cause may later change without changing the original observation. This preserves the evidence while allowing the investigation to develop.

Define the counting unit. One board can contain several defects, and one defect may produce several test symptoms. State whether the chart counts defective boards, distinct physical defects, failure events, or repair actions. Do not switch units between categories.

Create usable category definitions

Check whether inspectors or test software use different names for the same observation. Map aliases through a controlled dictionary instead of merging them manually in each report. Preserve the source code so that the transformation can be reviewed later. Record when each definition became effective and keep ambiguous historical entries separate until they can be mapped honestly; a forced mapping can invent specificity that the original inspector never recorded.

Review the chart’s meaning

1. Confirm the population and counting unit.
2. Check category consistency across the collection period.
3. Inspect the records behind the largest bars.
4. Review what was grouped into “Other.”
5. Choose an investigation based on frequency and engineering consequence.

A frequent low-impact condition and a rare serious failure may deserve different priorities. If severity or cost is used as a weight, label that explicitly; a weighted chart no longer represents simple occurrence frequency. Keep the unweighted counts available for context.

When categories change, maintain a mapping or mark the break in the trend. Splitting one broad category into several narrow ones can make the old problem appear to improve without any change in the product. A meaningful chart should trace back to stable definitions and identifiable records, allowing improvement work to target a real process condition.