Destructive tests such as adhesion, thickness sectioning and salt spray destroy the very parts they protect. Sampling too few hides the fault; sampling too many wastes good parts. Both mistakes come from treating destructive testing as a ritual instead of a statistical decision.
This intermediate guide covers destructive sampling: sample size by batch risk, where to cut samples, and how to read the scatter so the test result drives a process decision.
Common Mistakes and How to Avoid Them
The pitfalls that show up most often in real projects, with the cause and the practical fix.
| Mistake | Why It Happens | Practical Fix |
|---|---|---|
| 1. One sample per batch | Misses the spread | Sample by batch risk |
| 2. Cutting from the easy spot | Hides process faults | Sample across the load |
| 3. Testing only the mean | Scatter hides the worst part | Review the full range |
| 4. No position record | Fault cannot be traced | Label sample position |
| 5. Re-testing until it passes | Fake confidence | Define the re-test rule |
| 6. Sample size by habit | Too few or too many | Size from risk and history |
| 7. Ignoring outliers | Real fault dismissed | Investigate outliers |
| 8. No batch link | Result cannot guide process | Link test to batch data |
Best Practices That Hold Up in Production
The operating disciplines that separate a reliable line from a reactive one.
- Size samples by batch risk and process history
- Cut samples from positions that represent the load
- Read scatter and range, not only the mean
- Record position with every destructive result
- Define the re-test rule before the test
Implementation Roadmap
A practical sequence that can be adapted to your own project.
Process Flowchart
Destructive sampling flow
A step-by-step sequence with notes and cautions so every shift follows the same order.
- Destructive samples are expensive; each one should carry maximum information.
- Position labelling turns a test result into a process map.
- Never re-test the same failure until it passes.
- Never ignore an outlier without investigating it.
Working Data & Formula Notes
Sampling reference
Example guidance for destructive sampling; confirm with your customer and quality standards.
| Component / Parameter | Working Value / Role | What Changes Mean (annotation) |
|---|---|---|
| Sample size | 3-5 per position | Based on batch risk |
| Positions | Across the load | Front, middle, back |
| Risk rule | History of failure | More samples after failures |
| Scatter read | Range and outliers | Not only the average |
| Re-test rule | Defined before test | Prevents testing until pass |
Reference Data
Specifications and references cited in this guide. Confirm final parameters with your line supplier.
Sampling plan example
| Risk level | Samples | Positions | Action |
| Stable history | 3 | 1 | Release if pass |
| Recent failure | 6 | 2 | Contain until pass |
| New process | 9 | 3 | Gate before release |
| Customer return | Full audit | All zones | Stop and trace |
Implementation Cases
Case 1 - the one-sample batch that passed and failed
Situation. A batch passed destructive adhesion with one sample cut from the front of the load. After assembly, parts from the back of the same batch failed on the customer's line.
Approach. The sampling plan was changed to multiple positions with samples labelled by load zone, and scatter is now reviewed with every result.
Outcome. The fault was found at the back position before shipping, and sampling now covers the load.
Case 2 - re-testing until it passed
Situation. A destructive test failed twice, and the team re-cut samples until one passed, then released the batch. A later failure proved the release was wrong.
Approach. A written rule was introduced: a defined number of samples and a stop-and-trace response to any failure, with no silent re-testing.
Outcome. Releases now follow the data and the stop rule, ending false confidence.
Frequently Asked Questions
How many destructive samples do I need?
It depends on batch risk and history; three to nine is a practical working range.
Where should I cut samples?
Across the load, not from the easy spot, so the spread is visible.
Why read scatter, not just the mean?
A good average can hide a worst part that will fail later.
What is a re-test rule?
A written response to failure that prevents re-testing until the batch passes.
How do I label samples?
Record the load position, batch and test time with every result.
What if an outlier appears?
Investigate it; outliers often point to a real process fault.
When should sampling be tighter?
After failures, on new processes, or after process changes.
Who sets the sampling plan?
The quality engineer with the process owner, reviewed after each failure.
What Would You Like to Solve?
If destructive testing is wasting parts or missing faults, send us your test methods, batch history and current sampling plan. We can help set sample sizes, cut positions and scatter rules that make every destructive part count.
Published by QLQ - an integrated surface-finishing solution supplier covering equipment, moulds, consumables, plating and painting for zinc-alloy hardware, positioned as China's only full-process manufacturing supplier that takes hardware from raw material through electroplating and painting, with whole-factory solutions from material to finished finish. Values cited are project references; confirm with your line supplier before specification.