A passing result answers whether the reported value meets an applicable criterion. It does not necessarily answer whether the process is behaving as expected. An out-of-trend assessment considers the result in relation to a justified historical pattern or prediction.
Separate acceptance from expected behaviour
NIST distinguishes specification limits from statistical control limits: the former concern requirements, while the latter assess consistency of a process. A trend assessment likewise needs an appropriate reference pattern, rather than simply reusing the product acceptance interval.NIST — What are Variables Control Charts? (opens in a new tab)
| Comparison | Question answered |
|---|---|
| Result versus specification | Does it meet the defined acceptance criterion? |
| Result versus historical behaviour | Is it unusual for this process or series? |
| Result versus a stability expectation | Does the change fit the time-related model? |
The terms out of trend and out of control are not automatically interchangeable in every organisation. Read the applicable procedure to understand which pattern or rule triggered the concern.
A simple visual surprise is a starting observation. Calling it a formal statistical signal requires a defined method and suitable data.
Work through a passing but unexpected value
Imagine an original series of comparable assay results normally clustered around 10.0 mg per vial. A new result is 9.2 mg, within a hypothetical 9.0–11.0 mg acceptance interval. It passes that simple numerical criterion but may still be unusual relative to the established series.
Whether 9.2 mg is statistically unusual depends on the earlier variability, data quality and chosen rule. The interval alone cannot answer that. A process with broad routine variation and one with very narrow variation do not give the same meaning to the departure.
The example is not a product specification or a diagnosis of process deterioration. It illustrates why a binary pass label can leave useful information unexamined.
Check that the historical results belong together
Before investigating the material, establish whether the comparison uses the same measured property, reporting basis and sufficiently comparable analytical methods. A change from dry-basis to as-received results can create an apparent shift without the two series measuring the same quantity.
Changes in standards, instruments, preparation, reporting precision or sampling can also change the observed series. Record such changes rather than forcing all points into one undifferentiated baseline.
NIST cautions that control limits estimated from limited data may not have the assumed properties, and that adding signal rules increases false alarms. More sensitive monitoring therefore needs an understood trade-off rather than automatic escalation of every fluctuation.NIST — What are Variables Control Charts? (opens in a new tab)
Avoid choosing a trend rule only after noticing a striking pattern. Searching many possible patterns makes an unusual-looking sequence easier to find by chance.
Investigate the shift without rewriting history
The useful next step is a specific question: did an analytical change coincide with the shift, does a retained comparator behave differently, or do other attributes move in a related way? The answer should determine the follow-up evidence needed.
Keep the original series and the reason for any baseline change. Recentring a chart simply because recent results look different can conceal the change the chart was intended to detect.
Report the current specification result, the trend observation, the rule or rationale used, and the investigation outcome separately. If the evidence does not establish a cause, retain that uncertainty.
For future monitoring, a justified new baseline may become appropriate after an understood process change. That is a documented analytical decision, not a way to erase earlier unexpected behaviour.
Sources and further detail
- NIST — What are Variables Control Charts? (opens in a new tab)
Sections on specification versus control limits, estimated limits, recalculation and false alarms read. No universal OOT algorithm or minimum sample number is prescribed. The assay example is original.
Sources checked 19 September 2026. Worked examples are illustrative unless a supplied report is explicitly identified. This article has not undergone independent scientific peer review.