Process validation: proving a process when you cannot inspect the result

Identifying which processes genuinely need it

The test is whether conformity can be fully verified by subsequent inspection or test without destroying the item. Where it can, control and inspection are enough. Where it cannot, the outcome depends on parameters rather than on a measurement, and those parameters need to be established, held and evidenced. Typical candidates are welding and brazing, soldering, heat treatment, plating and coating, adhesive bonding, moulding of critical features, sealing, cleaning, and sterilisation. The list should be decided explicitly and recorded, because processes drift into this category when a design change makes a previously visible feature internal.

Equipment, operating window, and sustained performance

Work proceeds in layers. First, evidence that the equipment was installed, connected and functions as specified, with its instrumentation calibrated. Second, evidence that the process produces acceptable output across the range of settings it will actually experience, including deliberate excursions towards the edges of the intended window, so that the limits are known rather than assumed. Third, evidence that the process holds over sustained running with production material, production operators and normal changeovers. Compressing these into a single trial under ideal conditions produces a document that describes a demonstration rather than the process you will run.

Challenge the worst case, not the comfortable one

Validation done on the easiest configuration proves the least. The trials should use the material at the limits of its permitted specification, the thickest and thinnest sections, the fullest load in the oven or the chamber, the least experienced qualified operator, and the tooling at the worn end of its life. If the process passes those, ordinary production sits inside the demonstrated envelope. If it passes only at nominal conditions, then every production run outside nominal is unvalidated, which is usually discovered when a batch made with material at the edge of tolerance fails.

What has to stay controlled afterwards

Validation only holds while the process runs as validated, so it creates lasting obligations: parameters locked and change-controlled, instrumentation calibrated on schedule, operator qualification maintained and re-verified, materials from approved sources at the specified grade, and monitoring of the parameters during production with a defined reaction to an excursion. The routine failure is a validated process quietly modified for a practical reason — a different cleaning agent, a substitute gas, a faster cycle to meet demand — with nobody linking the change to the validation that assumed the original condition.

The triggers that require you to do it again

Revalidation is prompted by change and by evidence, and both need owners. Change triggers include new or relocated equipment, a different material grade or supplier, a design alteration affecting the feature, tooling replacement, a modified parameter, and a move of the process to another site. Evidence triggers include a rise in related defects, a field failure attributable to the process, or monitoring data drifting within limits but away from where validation was performed. The scope of the repeat should be argued from the change rather than defaulting to a full repeat or, more commonly, to none.

Frequently asked questions

How many production runs are needed to validate a process?

Enough to show that variation between runs is understood, which means more than one and rarely a fixed count. The determining factors are how much the process varies batch to batch, how severe the consequence of failure is, and how much prior knowledge exists from similar processes. Justify the number in the protocol before running it, using the process characteristics rather than convention, since an unjustified number is the finding an auditor will raise.

What is the difference between validation and qualification?

Qualification generally applies to equipment, utilities and facilities — evidence that a specific item was installed correctly and performs as intended. Validation applies to the process as a whole, including the material, the method, the people and the equipment together, and concerns the output rather than the machine. Qualification is normally a prerequisite: a process cannot be validated on equipment whose own performance has never been established.

Can we validate a process we have already been running for years?

Yes, and retrospective work relies on the accumulated production and monitoring data rather than on designed trials. It is weaker evidence, because nobody deliberately explored the edges of the operating window, and it is only credible if the historical data is complete, the process has been under change control, and the parameter records genuinely exist. Where the history is patchy, the honest route is a prospective exercise on the process as it currently runs.