Before choosing what to measure, you need to know what carries the quality of the product. That chain is climbed backwards: from the intended use to the properties that guarantee it, then to the steps that create them.
PAT-INDUSTRY works that chain with manufacturers in pharma, biopharma, chemicals, cosmetics and food. The vocabulary comes from pharma; the reasoning applies wherever a specification exists.
Three links, in this order
What the product has to do
The target profile: the performance expected by the user. In pharma, the QTPP. Elsewhere, the customer specification or the product promise.
A tablet that releases its active substance in thirty minutes. A cream that stays stable for eighteen months.
The properties that guarantee it
The measurable characteristics of the product on which that performance depends. In pharma, the critical quality attributes.
For release in thirty minutes: hardness, porosity, dose uniformity, particle size of the active substance.
The steps that create them
The unit operations where those properties are decided — not every operation in the process.
Hardness is decided at compression, particle size at granulation, uniformity at blending.
It is this third link that points to where you measure. A measurement placed on a step that carries no critical property produces a curve, not a decision.
And the incoming material, in this chain?
A critical property does not always come from the process. It often comes from what enters it: the particle size of an excipient, its water content, the origin of a lot. Those properties have a name, critical material attributes — CMA.
The concept is ICH’s, even if the acronym is usage: the annex to ICH Q8(R2) devotes its § 2.3 to “linking material attributes and process parameters to drug product CQAs”, and ICH Q11 does the same in § 3.1.5 for the drug substance.
The reference needs that precision: ICH Q8(R2) comes in two separately numbered parts, and the § 2.3 of the guideline itself is Manufacturing Process Development. It is the annex’s that carries this subject.
This is the link that explains why a model drifts when a supplier changes: the instrument has not moved, a material attribute has left the domain the model learned. An unidentified CMA becomes an unexplained deviation two years later.
How a property is judged critical
The word critical is not a matter of experience, it is the outcome of an analysis. In pharma the definition is the one in ICH Q8(R2): a physical, chemical, biological or microbiological property or characteristic that should be within an appropriate limit, range, or distribution to ensure the desired product quality.
That definition says what is critical; it does not say where to start. To decide where to place a measurement we add a priority criterion: a property calls for an in-line measurement when its variation, within the range actually observed, degrades the expected performance of the product.
And the method that produces that decision has a name too. You first build an input‑process‑output matrix for the operation — what goes in, what the step does, what comes out — then put the outputs through a risk analysis, usually an FMECA (failure mode, effects and criticality analysis). ICH Q9(R1) names it among its tools, in annex I.3. We do not run your risk analysis — that is your teams’ work; we bring to it what the measurement can see, and what it will not see.
Two parts of that priority criterion deserve attention. Within the range actually observed: a property with a large theoretical influence that never varies in your plant calls for no in-line measurement. Degrades the performance: the question is not whether the property is interesting, but whether the product stops keeping its promise when it moves.
In pharma this analysis belongs to quality risk management, framed by ICH Q9(R1) and built into development by ICH Q8. It leaves a trace: what was examined, what was set aside, and why. That trace is also what makes the choice of measurements defensible in front of an auditor.
And the level of formality is not imposed. It is one of the additions of the R1 revision: “Formality in quality risk management is not a binary concept (i.e. formal/informal); varying degrees of formality may be applied…” (§ 5.1). The text makes that level depend on the uncertainty, the importance and the complexity of the question — high formality calls for tools, a cross-functional team and a stand-alone report; lower formality sits inside the existing elements of the quality system. Tracing the criticality of a single operation therefore does not require the full apparatus.
Outside pharma the same exercise runs under other names — FMEA, functional analysis, control plan review — with the same useful output: a short, ranked list of what actually matters.
Why climb this chain before choosing a sensor
This is the order most often reversed. A project starts because a technology was seen at a trade show, or because a competitor uses one. The question becomes “what can we do with this sensor”. It should be “what makes the quality of my product vary”.
The two routes do not arrive at the same place. The first produces an installation that works and that nobody knows what to do with. The second produces a measurement whose output has a recipient and a consequence.
In practice, climbing that chain occupies the first phase of our engagement, and it is done with your experts: formulation, process, quality. We bring the method and the knowledge of what each measurement family can see; you bring the product and its history.
Frequently asked
Do you need a full QbD exercise before starting?
No. The chain described here can be climbed on a single operation, the one that concerns you, without reopening the whole development. Many projects start that way: a recurring issue, a suspect step, and the question of which product property is actually decided there.
What if the critical attributes cannot be measured in line?
That is common, and it is not a dead end. You then measure a related quantity, provided the link is established rather than assumed. Dissolution is not measured in line; porosity and hardness, which govern it, can be approached — and are measured at-line rather than in line. The work is to document that link, not to postulate it.
One precision that changes the feasibility: how easy that detour is depends on the product. For a highly soluble drug substance, bioavailability may depend only on the disintegration rate — a surrogate is then simpler to develop and validate, and the disintegration test itself can serve. For a poorly soluble substance, dissolution keeps a part of its own, and the surrogate is harder to demonstrate. The BCS class of the substance is therefore the first question to ask, before looking for which quantity to measure in its place.
After USP, Technical Guide: In-vitro Dissolution Modeling for Oral Solid Drug Product Continuous Manufacturing, 2024, United States Pharmacopeial Convention, Rockville, MD. USP presents these technical guides as a practical resource that “do not constitute official USP standards”: they are cited as state of the art, not as an enforceable reference.
Does this vocabulary hold outside pharma?
The reasoning does, entirely. The acronyms do not: nobody in food manufacturing talks about a QTPP. We use the vocabulary of your sector, and regulatory terms only where they are expected. The glossary gives the correspondences.
Which property of your product concerns you?
Name it, and the step where you suspect it is decided. Forty-five minutes are enough to climb the chain on that one case and see what would be measurable.