Batch Blending: Uniformity Is Not the Same as Consistency

Blending is the oldest recovery technique in powder manufacturing and the most misunderstood. Done as a designed step, it is ordinary manufacturing with a validation file. Done as a rescue, it converts a non-conforming lot into a compliant average and leaves no trace on the certificate.

The distinction is not the act. It is whether the act was in the process description before the batch needed it.

What blending actually changes

A blend averages three things: the assay, the impurity profile and the physical properties. Averaging is not neutral. It improves the composite number and degrades the information, because the certificate now describes a mixture whose components were individually different.

Property	Effect of blending	What the certificate shows
Assay	Weighted mean of the lots	A single number, better than one input
Related substances	Weighted mean per impurity	Within-spec profile that no single lot shows
Particle size	Mixture of distributions	Average D50, wider spread
Traceability	Two or more lot histories in one	One batch number, several origins

The last row is the one that ages badly. Six months later, when a question arises about a batch, the investigation has to open two production histories instead of one, and the answer to "what happened in this batch" becomes "what happened in two of them".

Designed blending versus rescue blending

A designed step has four properties. It is described in the process, with the inputs it permits. The mixing operation is validated, with a uniformity study that sampled the mixture at more than one point. The acceptance rule for combining lots is written down in advance. And it happens on schedule, not only when a lot is in trouble.

A rescue has the opposite properties. It appears when a lot misses a limit, its inputs are whatever is available, its uniformity is asserted from the blending time, and its record is thinner than the surrounding batch records. The result can be identical powder. The regulatory positions are opposite, and the general expectation in the pharmaceutical framework, set out in the ICH Q7 guidance and the European GMP text, is that a non-conforming lot is investigated rather than diluted.

Uniformity is a property of the operation

The most common blending failure is not inadequate mixing. It is the assumption that adequate ingredients guarantee a uniform blend. Segregation during discharge, static cling in the blender, and a fines fraction that migrates during handling all produce a mixture that was uniform in the vessel and is not uniform in the drums.

That is why the uniformity study samples across the discharge rather than inside the blender, and why the batch history matters more than the single result: a blend operation that produces drum-to-drum variation will show it across batches as a wider spread on every parameter.

It also connects back to the physical side of the specification. Particle size distributions average arithmetically, but the flow behaviour of a blend is not the average of the flow behaviours, and a customer whose process is sensitive to flow will find the difference before the certificate does.

The stability question

Stability data attaches to a process history. A blend has at least two. Three defensible arrangements exist: the blend itself goes on stability; every component lot is covered by existing data and the blend is composed only of lots within their own history; or the blend is restricted to lots from the same campaign with the same process state.

What is not defensible is silence. A certificate for a blended lot whose stability basis is unstated is asking the customer's file to absorb an assumption, and the file will be asked about it at the worst possible time.

What the mass balance shows

A blend can be detected from its own certificate in one situation: when the profile is smoother than the process usually produces. Related substances that normally vary by a factor of two between lots suddenly sit in a narrow band, assay lands in the middle of the historical range, and every number is comfortable. Comfort is not evidence of blending, but a profile that is suspiciously smooth deserves the question.

The mass balance is the other tell. A blend of two lots each carrying a different residual solvent produces a total that is the average, and an investigation that assumes a single process history will not find the second solvent.

What belongs in the quality agreement

Three clauses close the topic. Whether blending is a permitted step for the material supplied. If it is, what the inputs may be, and whether the customer is notified of the component lots. And an explicit statement that non-conforming material is not blended into conforming lots. The last clause costs nothing and is the one that prevents the conversation nobody wants to have.

Where the process changes to introduce blending where there was none, that is a change like any other, and it belongs in the notification categories rather than in a production meeting.

Where this leaves a supply-chain partner

We distinguish the two cases explicitly: a blending step that exists in the process, with its validation file, and everything else. Where a lot fails, the lot is dispositioned on its own record. Where a customer asks whether a batch is a blend, the answer is a fact about the batch rather than a policy about disclosure.

The question worth asking any supplier is the fourth one in this note: is blending a permitted step for this material, and may we see the rule that governs it? The answer is short, and it describes the site's discipline better than most of the certificate.

How a blend should be validated

A designed blending step earns its place in the process with three pieces of evidence. The mixing study, which demonstrates that the operation produces a uniform mixture within a defined time, sampled at more than one point in the vessel and again at discharge. The input rule, which states which lots may be combined: same process state, same specification version, and no lot carrying a failed parameter that the blending is expected to dilute. And the stability position, which states whether blends go on stability or rest on the coverage of their components.

The general expectations for such steps sit in the ICH Q7 framework, the European interpretation in the GMP text, and the compendial perspective on mixture uniformity in the USP general chapters. None of them blesses the rescue version, and all of them treat the documentation as the difference.

What a customer's file needs to know

A file written around a single-lot process describes raw material sourcing, impurity carry-forward and stability on that basis. A supply that includes blends changes two of those three assumptions quietly. The sourcing becomes two lots per batch; the stability basis becomes a composition question; the traceability trail gains a fork.

None of that is objectionable where the blending is designed and disclosed. All of it is objectionable where the customer learns about it from a question they asked about something else. The notification framework is where the disclosure lives, and "introduction of a blending step" is one of the cleanest examples of a change that deserves its own line in the agreement rather than a clause about general compliance.

Reading the batch record of a blend

A blend's record should let a reader reconstruct the composition in one page: which lots, what quantities, what each lot's own results were, and what the blend's results are. The comparison between the component results and the blend result is where the information lives, because a blend whose components all sat comfortably inside specification is a routine operation and a blend whose components straddled a limit is a decision that deserves its own justification.

The second case is not automatically wrong. There are legitimate operations that combine a low-assay lot with a high-assay one under a written rule established before either lot existed. The difference from a rescue is entirely documentary, which is why the rule's existence matters more than the numbers in any individual case.

One question closes the topic for a buyer: may we see the rule that governs which lots may be blended, and was it written before or after the batch in question? The answer is a single date, and it says more about the site than any audit report.

Why the temptation exists

Blending exists as a rescue because the economics of a failed lot are severe: material, capacity, a delivery date and a customer relationship all sit on the same decision. The pressure to find a route that keeps the batch alive is real, and it is strongest exactly when the documentation discipline is weakest, which is at the end of a quarter.

Recognising that pattern is what makes the designed version worth building. A site with a written blending step, a validated mixing operation and a pre-agreed input rule can recover yield legitimately and defend it. A site without those three has a choice between absorbing the loss and improvising, and improvisation under pressure is how the practice becomes invisible.

None of this is unique to powders. The same mechanics appear in solvent recovery, in second crops and in rework, and the common thread is that each is legitimate when it is in the process before it is needed.

What a buyer can rely on

Two facts, if the supplier will state them. Whether blending is permitted for the material, and what the input rule says. Everything else follows: the stability basis, the traceability trail, the notification duties. A supplier who answers both is describing a process. One who answers neither is describing a practice that has not been written down, which is the case where the certificate and the batch record drift apart.

The question is short enough to belong in the first technical exchange, and it is far cheaper there than in a deviation report eighteen months later.

Frequently asked questions

Is blending batches allowed?

Where blending is a designed step in the approved process, with defined inputs, a validated mixing operation and a uniformity check, it is ordinary manufacturing. Where it appears only to recover a failing lot, it is generally treated as prohibited, because it dilutes a non-conformity instead of correcting it.

What does blending do to the impurity profile?

It averages it. Two lots that individually carry different related substances produce a blend whose profile is the weighted mean. That can bring a lot inside specification arithmetically while the individual lots tell a truer story about the process.

Does a blend change the stability picture?

It complicates it. Stability data attaches to a process history, and a blend has two. The blend should either be placed on stability itself or be composed of lots that are individually covered, and the file should say which.

How is blend uniformity demonstrated?

By sampling the mixture at multiple points, usually across the discharge, and testing for the parameters most likely to vary. Uniformity is a property of the operation, demonstrated during validation, not an assumption inherited from the ingredients' compliance.

What does the certificate of a blended batch show?

The tests on the blend. The component lots appear in the batch record, and whether the customer sees them depends on the quality agreement. A blend whose components would individually have failed is the case where the question matters most.

Should a customer be told a batch is a blend?

Where blending is a normal process step, disclosure is a documentation question, and the agreement can settle it. Where the blend was assembled to recover non-conforming material, the customer should be told, because the batch they receive is not the one their file describes.