Three Consecutive Batches: What Process Validation Proves

Three batches is a number everyone recognises and almost nobody can justify. It is a convention that survived because it is convenient, not because three is a property of processes. What a validation exercise actually establishes is narrower: that this process, run inside its normal variability, produces material that meets specification.

The distinction matters because it changes what you ask for. A supplier who says "three batches passed" has told you about outcomes. A supplier who can explain which three, why those three, and what varied between them has told you about control.

The framework moved, and the number did not

The current expectation is not a one-off event but a lifecycle: process design, process qualification, and continued verification during commercial production. Three batches belong to the middle stage, and they are not the end of the story.

Stage	What it establishes	What it produces
Stage 1 — process design	Which parameters matter and how much they can move	Design space, critical parameter list, control strategy
Stage 2 — process qualification	That the commercial-scale process works inside those limits	Validation batches, protocol, report, deviations
Stage 3 — continued verification	That it keeps working as time, people and equipment change	Ongoing data, trend review, corrective action

Most supplier documentation stops at stage two. That is a legitimate stopping point for a conversation, and an incomplete answer to the question of whether the process is currently in a state of control.

What the three batches must represent

Three batches prove repeatability only if they are the right three. Four properties decide that.

They must be consecutive. Batches separated by other campaigns, equipment overhauls or process tweaks are not consecutive in any meaningful sense, even if the numbering suggests otherwise.

They must be at commercial scale. A validation batch at a fraction of commercial size validates the fraction. Scale-dependent variables such as mixing, heat transfer and drying behave differently at size.

They must be at the edges, not the centre. A process validated at the middle of every parameter range has been validated for a condition it will not always meet. Where a parameter range is claimed, the batches should show the process behaving at the limits the file commits to.

They must not be selected after the fact. Choosing the three best batches from a larger campaign and calling them validation is one of the more common ways a report becomes fiction. The selection rule should be written before the batches are run.

What three batches cannot establish

Three data points cannot support a statistical claim about capability. They can show that the process met specification three times under observation; they cannot show the distribution of outcomes, the failure rate, or the behaviour of a parameter that was never moved.

They also cannot establish that the process is stable over time. That is what continued verification is for, and it is the part of the framework that most often has no data behind it when a buyer asks.

And they cannot establish that the material is suitable for your use. Validation is about the process meeting its own specification. Whether that specification matches what your product needs is a separate question, and it is the one a buyer should be asking.

Prerequisites that are often assumed rather than checked

A validation exercise run before its prerequisites are complete produces a document, not evidence. Four prerequisites carry most of the weight.

1. Facility and equipment qualification complete, including the utilities that touch the product.
2. Analytical methods validated, because unvalidated methods cannot distinguish a passing batch from a lucky measurement.
3. Operators trained on the version of the process being validated, not the previous one.
4. Cleaning validated, since carryover between batches is part of what the batch result means.

The fourth is the one most often sequenced late, and the one most likely to change an impurity result.

Deviations during validation are informative, not disqualifying

A validation campaign with no deviations at all is worth a question. Real processes produce events: a temperature excursion, a delayed step, an out-of-trend result that was investigated and closed.

What matters is not the absence of deviations but whether they were investigated, whether the investigation reached a cause, and whether the batch was still valid. A campaign with two well-investigated deviations and a clear conclusion is stronger evidence than a campaign with none and no explanation for why it was so smooth.

For how deviations are classified and what a buyer is entitled to see, see deviation classification and what gets disclosed.

What to ask a supplier

Four questions turn a claim into information.

1. Which batches, and were they consecutive in the production sequence?
2. What parameter ranges were exercised, and were any of the batches run at a limit rather than the centre?
3. What deviations occurred, how were they closed, and did any batch fail and get repeated?
4. What does continued verification show over the last two years?

The fourth question separates suppliers who validated once from suppliers who are still watching. For the trend side of that picture, annual product quality review and batch-to-batch consistency cover what the data should look like when it exists.

Primary sources: EudraLex Volume 4, Annex 15 for qualification and validation, ICH Q7 for API-specific GMP expectations, and PIC/S inspection guidance for how design and control strategy connect to what gets validated.

Frequently asked questions

Is three the required number of validation batches?

It is a long-standing convention rather than a fixed requirement in every case. The defensible position is that the number follows from the process and its variability, and three remains the common minimum for a conventional process.

Can validation batches be sold commercially?

Often yes, if they were made under GMP and meet specification. Whether they may be released depends on the applicable rules and on the holder's own procedures, so it is a question for the qualified person rather than a general rule.

What if a validation batch fails?

The failure is investigated, the cause addressed, and the campaign repeated as needed. What matters for a buyer is disclosure: a failed batch that appears in the record with a closed investigation is more reassuring than a campaign with no record of difficulty.

Does process validation expire?

It does not expire in the way a certificate does, but its conclusions are only valid for the process as it was. Changes to equipment, scale, suppliers or parameters can invalidate the basis, which is why continued verification exists.

What is the difference between validation and qualification?

Qualification applies to facilities, equipment and utilities and establishes that they are suitable. Validation applies to the process and establishes that it produces the intended result. A process cannot be validated on unqualified equipment.

Do I need to see the full validation report?

Usually not, and suppliers rarely release it. A summary covering batch identity, parameter ranges, deviations and conclusion is normally sufficient for supplier qualification, and it is the part that answers the questions above.