Stratified Sampling
Stratified sampling is the deliberate collection of in-process samples from defined positions across a blender, discharge stream or compression run — designed to detect non-uniformity and segregation that random or single-point sampling would miss.
Read the full summary
In solid-dose manufacturing it is the regulator-preferred sampling approach for blend uniformity and dosage-unit content uniformity.
On this page · 6 sections
How does Stratified Sampling apply to your shop floor?
Pick your industry and scale — Ask V5 rewrites the definition in your context, gives a worked example, and shows what V5 does on day one.
01What stratified sampling is
Stratified sampling subdivides the population (blend, discharge stream, compression run) into strata and samples from each. The strata are chosen to capture the modes of variation that matter: position in the blender, time during discharge, location across the press turret. The sampling plan is fixed in advance and documented in the protocol or SOP, not chosen by the sampler in the moment.
- Blender BUE — 10 spatial locations (top/middle/bottom of multiple columns) sampled at end-of-blend.
- Discharge stratified — first/middle/last portions of drum or IBC discharge sampled separately.
- Press stratified — n tablets at the beginning, middle and end of compression, plus during disturbances (speed change, hopper refill).
- Stratified weight — samples grouped by tablet weight band to detect die-fill variation.
02The FDA stratified guidance
FDA's 2003 draft guidance on Powder Blends and Finished Dosage Units remains the working playbook. It introduces stratified in-process sampling at the press as the alternative or complement to BUE — recognising that thief sampling in the blender is sometimes unrepresentative and that the press is the dosage-unit-of-truth. Stratified sampling forms the basis for content uniformity acceptance under criteria more stringent than USP <905> alone.
| Stage | Strata | Typical n per stratum |
|---|---|---|
| Blend (in-blender) | 10 locations × top/middle/bottom | 1–3 |
| Discharge | Beginning / middle / end | 10 |
| Press routine | Start / middle / end of run | 20 |
| Press extended (disturbances) | After speed change / hopper refill / shift | 20 each |
03Designing a stratified plan
- Identify the modes of variation — blender geometry, discharge order, press dwell, environmental shifts.
- Choose strata that resolve those modes — not arbitrary times or locations.
- Set sample size per stratum on statistical power, not convenience.
- Specify when to sample relative to events (post-mix, post-refill).
- Document acceptance criteria per stratum and across the run — both matter.
04Common mistakes
- Treating stratified sampling as 'more random samples' — it is the opposite, it is structured.
- Letting the operator choose locations on the day — destroys statistical defensibility.
- Skipping disturbance samples — missing the worst-case content uniformity moments.
- Reporting only the mean — hiding individual-stratum failures inside aggregate statistics.
- Using the same plan for every product without considering API loading, dose mass or segregation tendency.
- Failing to qualify the sampling thief or stream port — invalid samples invalidate the conclusions.
05Cross-industry examples
- Tablet manufacturing — stratified press sampling drives content uniformity per USP <905>.
- Encapsulation — stratified end-of-fill weight + assay to catch hopper-bridging effects.
- Powder-fill biologics — vial-position stratified sampling across a fill cycle.
- Dietary supplements — stratified blend + finished-product sampling for label-claim defence under 21 CFR 111.
- Veterinary medicated feed — stratified mixer outlet sampling for medicated article uniformity.
- Cosmetics — stratified pull from large mixers and fillers for MoCRA stability batches.
06How V5 Ultimate handles stratified sampling
Frequently asked questions
Q.What's the difference between random and stratified sampling?+
Random sampling gives every unit an equal chance; stratified divides the population into meaningful subgroups and samples each. Stratified is more powerful when you expect or want to detect subgroup differences (segregation, drift).
Q.Is stratified sampling required?+
FDA's guidance is a draft but it is the de facto expectation for solid-dose products. Most warning letters citing inadequate sampling refer back to its principles.
Q.How many tablets at the press?+
Routinely 20 per stratum (start/middle/end), with extended sampling (20 more per disturbance) during PPQ and after process changes.
Q.Can PAT replace stratified sampling?+
Continuous PAT covers every dosage unit in principle — but the validation requires demonstrating equivalence to stratified discrete sampling first.
Q.Do I need to stratify for liquids?+
Less critical when fully homogeneous, but stratified pull at fill start/middle/end is still good practice for content uniformity on suspensions and emulsions.
Primary sources
Further reading
Want to see how Stratified Sampling could fit into your own records and workflows? Explore the related V5 pages or talk to our team about what applies to your operation.