Site leadership and quality · Quality metrics

Tracking weld reject rates for a new crew

Count rejected welds against welds examined, by welder, setup and defect type, from the first day, using the weld log as the only data source. The overall percentage matters less than the trend and the split: a few welders or one machine setup usually explain most rejects, and finding which one tells you what to fix.

What to count

  • Welds examined and welds rejected, for the same method and period. Do not mix visual, borescope and NDT results in one figure.
  • Sample or coupon welds separately from production welds.
  • Rejects by welder or operator ID, by machine or head, by material and size, and by defect type.
  • Repairs and cut-outs, linked to the original weld number.
  • The date, so you can see a trend week by week.

Reading the numbers

Patterns and what they usually point to
PatternLikely area to look atNext step
Rejects spread across all welders, same defectProcedure, equipment, purge supply, materialCheck setup and supplies before blaming people
Rejects concentrated in one welderIndividual skill or habitWatch them weld; see the retrain, retest or replace decision
Rejects on one machine or headEquipment calibration or wearCheck and calibrate; compare with another unit
Rejects after a change in size or materialSchedule or technique not adjustedReview the weld schedule; add sample welds
Rate falls over the first weeksNormal learning of site equipment and rulesUse it to relax extra hold points

If one welder keeps failing, the employer’s options are covered in retrain, retest or replace. Defect causes are explained in common orbital weld defects.

Keep the definitions fixed. If you change what counts as a reject, or start counting sample welds with production welds, mark the change in the weekly table; otherwise a trend can appear or disappear because of the counting, not the welding.

Where the owner spec and QA manager decide

Some owner specifications set reject rate thresholds that trigger extra examination or action; some codes and specifications require additional examination when a reject is found in sampled work. Those rules are mandatory and come before your own metric. Your QA manager decides how the metric is defined internally and what triggers a review. Do not publish a target that encourages people to avoid recording rejects.

Setting it up on day one

Add a result column and a defect code to the weld log from the first weld, and agree a short list of defect codes with your QA manager based on the owner’s acceptance criteria. Produce a simple weekly table by welder and setup. It takes little time if the log is kept daily, and it means that when a question comes up in week three you already have the data rather than reconstructing it from memory.

Small numbers

With a new crew, weekly counts can be small, so one reject swings the percentage. Look at counts as well as rates, and at defect types. Use the data to adjust hold points, not as a scorecard for individual pay.

Frequently asked questions

Should every reject become an NCR?

Not necessarily. Many quality plans treat a routine repair of a rejected weld through the repair process and reserve NCRs for systemic issues or work that cannot be repaired normally. Your plan decides.

Can we compare our crew’s reject rate with another site?

Only with care. Different acceptance criteria, examination extent and counting rules make rates hard to compare.

Who should see the reject data?

QA/QC, the foreman and the construction manager. Sharing a welder’s own data with them, privately, helps them improve.

Have a fab scope that needs these trades?

Requests are in preview: you can fill in the request form, but it does not submit yet. Email WELDK with your roles, site country and start window.