Validation or verification — which one applies?
The distinction decides how much work a new method costs. Verification asks "can we do this correctly?" for a method someone else has already validated. Validation asks "does this method do what it is supposed to?" and has to be answered by the laboratory itself.
| Situation | What is required |
|---|---|
| Published standard method (ISO, EN, ASTM, IEC, DIN…) used as written | Verification (7.2.1.5): show that you achieve the performance stated in the standard with your staff, equipment and environment. |
| New revision of a standard method | Verification of the changes — review what changed and re-verify the affected performance characteristics. |
| Laboratory-developed or in-house method | Full validation (7.2.2.1). |
| Standard method outside its scope (other material, range, temperature, sample type) | Validation of the extension — at least for the characteristics the change can affect. |
| Standard method with modifications (shortened test time, different fixture, other instrument principle) | Validation of the modification, and assessment of its influence on results (7.2.2.2). |
| Customer-specified or manufacturer method | Treated as non-standard unless it is published and validated — validation, with the customer's agreement on the requirements. |
In test laboratories for products — electrical, mechanical, environmental, battery or EMC testing — most methods are standard methods, so verification is the everyday case. Validation comes up mainly with customer test specifications, accelerated or shortened procedures and in-house fixtures.
Performance characteristics to determine
Clause 7.2.2.3 requires that the performance characteristics of a validated method are relevant to the customer's needs and consistent with specified requirements. The standard names typical characteristics in a note; not all apply to every method. Choose and justify the ones that matter.
| Characteristic | Question it answers |
|---|---|
| Measurement range | Over which range of values does the method give valid results? |
| Accuracy / trueness (bias) | How far is the mean result from a reference value, e.g. a certified reference material or a reference method? |
| Precision: repeatability | How much do results scatter with the same operator, equipment and short time interval? |
| Precision: intermediate precision / reproducibility | How much do results scatter across operators, days, instruments — or across laboratories? |
| Measurement uncertainty | Which interval around the result contains the true value with a stated confidence? |
| Limit of detection / quantification | What is the smallest value that can be detected or quantified reliably? |
| Selectivity / cross-sensitivity | Do other components, the matrix or interference affect the result? |
| Linearity | Is the response proportional over the range, or does it need correction? |
| Robustness (ruggedness) | How sensitive are results to small changes in conditions — temperature, fixture torque, timing, ramp rates? |
A validation plan in seven steps
- 1
Define the requirements
Write down what the method has to achieve: measurand, range, required uncertainty or tolerance, the decision the results will support. Agree them with the customer where relevant — this is the yardstick for "fit for purpose".
- 2
Select the characteristics
Pick the performance characteristics that matter for these requirements and justify the ones you leave out.
- 3
Set acceptance criteria in advance
Define before testing what counts as a pass: e.g. bias within ±1 %, repeatability standard deviation below a set value. Criteria defined afterwards are an assessment finding waiting to happen.
- 4
Choose the techniques
The standard lists several in a note to 7.2.2.1: reference standards or reference materials, systematic assessment of influence factors, robustness tests by varying controlled parameters, comparison with another validated method, interlaboratory comparisons, and uncertainty evaluation based on the method's principles.
- 5
Run the experiments
With calibrated equipment, authorised staff and recorded conditions — the validation data are technical records like any other.
- 6
Evaluate and decide
Compare results against the acceptance criteria and write the statement on validity: the method is (or is not) fit for the intended use, within stated limits.
- 7
Release and monitor
Release the method version, authorise staff, and monitor ongoing validity with quality control and proficiency testing (clause 7.7). Re-validate when the method changes.
What the validation record must contain
Clause 7.2.2.4 lists the minimum content of the retained records:
- The validation procedure used.
- The specification of the requirements.
- The determination of the performance characteristics of the method.
- The results obtained.
- A statement on the validity of the method, detailing its fitness for the intended use.
For verification, the standard is less specific, but assessors expect the same logic on a smaller scale: which performance the standard method claims, what you measured, and the conclusion that you meet it. Keep the raw data traceable — which items, which instruments, which calibration status, which operator.
Typical assessment findings
- A standard method was introduced without any verification record, or verification consists only of "method works".
- A modified standard method (shorter dwell time, other fixture) is treated as the standard method and was never validated.
- Acceptance criteria were defined after the results were known.
- The validation report has no statement on fitness for intended use.
- A new revision of the standard was adopted without reviewing what changed.
- Validation data cannot be traced to the instruments and their calibration status at the time.
- Staff perform the method without an authorisation that references the released method version.
Several of these come down to version control: which method revision was released when, and which test jobs were run with which revision. When method versions, test sequences, instruments and results are linked in one system, that question is answered by a query instead of a search through folders.