A 500 kg batch goes into a ribbon blender, and two kilograms of it is a vitamin premix. That is 0.4 per cent of the load. Every capsule filled from that batch is either within specification or not, and the difference is decided by whether those two kilograms went everywhere.
Blending looks like the simplest operation in the building. It is the one that fails most quietly, because a blend can be uniform in the sample and segregated in the drum.

A service with a boundary
Dietary supplement contract blenders do one thing: they combine powders to a specification and hand back a homogeneous blend. They may also mill, sieve, and pack into drums or foil-lined bags. They usually do not fill capsules, press tablets or run a packaging line, and that boundary is the first thing to understand about buying the service.
The boundary matters because of who is on the label. In the European Union, a food business operator can have material blended at one site and filled at another; the operator named on the pack carries the responsibility for the finished unit either way. In the United States, a facility that manufactures, processes, packs or holds a dietary supplement registers with FDA, and the preventive controls rule under 21 CFR 117 attaches to each facility for the operations it performs.
So a blend certificate covers the blend. It does not cover the capsule. When a customer complaint arrives about a capsule that assayed at 80 per cent, the question is not whether the blend was homogeneous. It is whether anybody measured the blend in a way that would have caught it. The finished-goods side of the same project is set out in our note on contract manufacturing.
Micrograms beside grams
The reason supplement blends are harder than food blends is the ratio between the smallest and largest ingredient. A B12 dose is measured in micrograms; a calcium dose is measured in hundreds of milligrams. That is six orders of magnitude in the same batch.
You cannot weigh 2.4 micrograms on a production scale, and you cannot disperse a few grams of a potent ingredient through 500 kg by dropping it in. The standard answer is a pre-blend: the active is diluted onto a carrier in stages, each stage bringing the concentration down by a factor of ten, until the quantity is large enough to weigh and small enough to disperse.
The carrier is not neutral. Maltodextrin, silicon dioxide, calcium carbonate and rice flour behave differently in a blender and differently in the finished product. A pre-blend built on a coarse carrier in a fine formula will segregate; one built on a carrier that picks up water will cake in the drum three weeks later, which is the same problem powder caking presents in storage.
Why blends drift apart
Mixing is not a one-way process. Powders mix toward uniformity and then, if you keep going, separate again. The mechanism is physical, and three properties drive it.
Particle size is the strongest. Fine particles percolate down through the voids between coarse ones, so a blend of 50-mesh and 200-mesh material will show the fine fraction at the bottom of the drum after handling, no matter how good the mixer was.
Density is next. A mineral salt at 2.7 g/cm³ and a light botanical at 0.4 g/cm³ do not stay together when a drum is dropped, vibrated or conveyed. This is the practical reason why mineral premixes are often granulated before blending.
Surface behaviour is the third. Fine powders pick up static and cling to the shell of the blender; hygroscopic powders form small agglomerates that survive mixing as soft lumps and only reveal themselves when a capsule is opened.
The operational consequence is that mixing time has an optimum rather than a minimum. Too short and the blend is not uniform; too long and it de-mixes. The optimum is specific to the recipe and the equipment, which is why a blender quoting a fixed twelve minutes for every job is telling you something about how it works.
How uniformity is proven
A uniformity result is only as good as its sampling. The usual method is a thief sample taken from a set number of positions in the blender or the drum, each analysed separately rather than composited.
Ten points is a common number. What matters is where they are: top, middle, bottom, centre, and near the shell, because the shell is where fine material clings and where the bottom is where it collects. A result built from three samples taken through the top hatch describes the top hatch.
The acceptance criterion is usually expressed as relative standard deviation across the points, with a limit in the region of 5 per cent for a well-behaved blend, and a tighter limit for a low-dose active. A composite assay of 100.2 per cent of label tells you the average. It says nothing about whether one capsule in ten is at 60 per cent.
That is the argument for sampling per point, and it is also the argument for asking who reads the result. If the blender signs the uniformity report and the filler signs the finished goods report, the two documents can both be green while the batch is not.
Cleaning, allergens and the claim on the front
A blender that runs six products a week has six opportunities to leave something behind. Cleaning validation is the evidence that it did not.
The practical questions are what is swabbed, what the limit is, and which allergen is used as the marker. A dairy protein is easy to detect at low levels and is a sensible marker for a dairy-containing product; a soy or gluten marker needs its own method. A cleaning regime validated only against the previous product's colour is not validated against its protein. Under GMP the record of that validation is the first thing an auditor asks to see.
This is where a clean label claim becomes a manufacturing claim rather than a marketing one. "No cross-contact with allergens" is a statement about changeover, scheduling and cleaning records. A site that runs allergen-containing products on the same line can still make the statement, but it has to be able to show the sequence, the clean and the verification. That is also the substance behind traceability: not a barcode, but a record that connects a drum of powder to a batch, a batch to a cleaning event, and a cleaning event to a signature.
From drum to batch
A blend record should allow one question to be answered in a few minutes: which lots went into this batch, and where did this batch go.
In practice that means lot numbers recorded at dispensing rather than reconstructed later, a yield reconciliation that explains any loss above the normal handling allowance, and a retained sample from the finished blend kept for the shelf life of the product it went into. A blend that cannot be traced to its lots cannot be recalled with any precision, only withdrawn wholesale.
Yield reconciliation is the least glamorous and most useful control. If 500 kg went in and 492 kg came out, the eight kilograms went somewhere: dust extraction, spillage, or a line that was not emptied. A site that records 498 kg every time without explaining it is not measuring.
When blending-only is the right buy
A vitamins and supplements contract manufacturer will often quote blending and filling as one price, and there are three situations where splitting them is better.
The brand already owns a filling line or has a packer it trusts. Buying blend in 25 kg foil-lined drums and filling locally removes a transport leg and often a week of scheduling.
The product is a powder sold as a powder. A greens blend or a protein blend does not need a capsule line, and the blender is the whole supply chain, which is also why the dosage form decides whether blending is a step or the product.
The active is the brand's own IP and it does not want to hand the full formula to a filler. Blending at one site and filling at another splits the knowledge, at the cost of splitting the responsibility too.
In every other case, the argument for one site is the one that decides most supplement projects: a single party accountable for the unit that a consumer opens. The blend is a step, not a product.
Ding Shaohua, Foreign Trade Manager at WANHUACAO, works with brands and distributors on supplement programmes from brief to first shipment, in bulk and in private label, across gummies, capsules, powders and tea. The plant operates under SGS CN23/00007035 for gel confectionery and SC11434160205012 covering eight dosage forms, sports nutrition supplements among them. Reach him on WhatsApp, TikTok, LinkedIn and Facebook.
FAQ
Can a blender supply finished capsules?
Some can, many cannot, and the distinction is worth checking before a quotation is compared. A blending-only site holds the equipment and the records for powder. Filling needs a capsule machine, a polishing step, a metal detector and a packaging line, and in most jurisdictions the facility registration and the preventive controls apply to whoever performs those steps.
What uniformity result should I ask for?
Ask for the sampling plan before the result. A report that states an RSD without saying how many points, where they were taken and whether each was assayed separately cannot be evaluated. Ten points across the blender, each assayed, with the acceptance criterion written into the specification, is a report that means something.
Does blending-only save money?
On the blend, usually yes, because filling, capping, labelling and cartoning are charged per unit and the blender does not carry them. On the project, it depends on volume. Below a few hundred kilograms a year, the second set of MOQs, the extra freight and the split retention samples often cancel the saving.
Who is responsible if a finished batch fails assay?
The party that released the finished unit, which is normally the filler and the brand named on the label. A blend certificate protects the blender, not the product. This is the practical reason to agree in advance on who holds retail-unit retention samples when the work is split across two sites.
