Most specimen collection kits are assembly problems. The components exist, the preservation chemistry is known, the container is a tube, and the work is sourcing, cleanroom assembly, documentation, and shipping.
Occasionally a program arrives where none of that is true. The company has a collection format that is not a tube and not a swab. No supplier catalog has the part. No procedure covers the transfer step. And the preservation chemistry that works in a laboratory has not been shown to work in the hands of an untrained person at a kitchen table.
The company needs all of it settled well enough to run a study, then settled again, differently, to sell the product at retail. This is what that engagement looks like from the manufacturing side.
The problem was the transfer, not the assembly
The collection device worked. The question was what happens between the device and the laboratory.
A specimen held on an absorbent substrate has to get into a preservation medium, and the person doing it is not a phlebotomist. They are at home, following a printed instruction sheet, once. Every step you add is a step that gets skipped or done wrong, and every failure lands as a rejected sample and a support ticket.
So the first work was not assembly. It was working out a transfer path that a first time user could complete without thinking about it, that did not contaminate the sample, and that did not require a component nobody manufactures.
The answer involved an adhesive backed element that created a controlled pouring surface from the collection substrate into the media vessel. Simple in description. Not simple to arrive at, because the adhesive has to bond to the substrate without interfering with the specimen, hold under the force of the transfer, survive the temperature range the kit will actually see in transit, and not be the thing that makes the kit feel cheap when a consumer opens it.
That is four constraints on one component, and two of them are chemical rather than mechanical.
Media selection ran in parallel
The preservation medium is chosen by what the assay needs, and that is the client’s call and the laboratory’s call, not ours. Our job is everything around that choice: fill volume, vessel compatibility, closure integrity, whether the medium survives the shipping profile, and whether a consumer can open, use, and reseal the vessel without spilling a preservative.
Those questions are answerable, but only with material in hand. Which brings up the part of these projects that nobody quotes for.
Samples, in both directions
We sent the client physical samples of every candidate material, in the configurations we thought would work, so their team could run their own evaluation. Not renderings. Not spec sheets. The actual components, assembled the way the kit would assemble.
Then we ran our own bench testing on the same configurations, so that when we said something worked we had our own evidence and were not repeating a supplier’s claim. Two independent looks at the same question, and where they disagreed we found out why before anything scaled.
This is the step that gets compressed when a program is in a hurry, and it is the step that determines whether the pilot tells you anything. A pilot built on an unverified component tests the component, not the program.
Then the paperwork became the product
Once the physical build was settled, the constraint moved.
Instructions for use went through revision after revision, because the instruction sheet is the user interface. Label proofs went under document control with quality approval before anything printed, which sounds bureaucratic until you have shipped a lot with a wrong revision on it and have to decide what to do with the units already in the field.
Kit identity had to be resolved: what identifier each unit carries, when it is assigned, how it is scanned, and how it maps to whatever the receiving laboratory generates on its side. Get that wrong and traceability is a spreadsheet exercise, which works at a hundred units and collapses at ten thousand.
The return path had to be settled: shipping method out, shipping method back, and how the returned specimen stays compliant in transit.
Pilot, then pilot again
The program ran as two pilots rather than one, deliberately.
The first pilot answered whether the kit worked as designed. The second incorporated what the first taught, at higher volume and closer to real conditions, and answered whether it worked when the people handling it were not the people who designed it.
Alongside those, we built volunteer kits and ran a hands on training session with the client’s team. Not a slide deck. Physical kits, in a room, with the people who would be answering support calls. You learn more about an instruction sheet in twenty minutes of watching someone use it than in three rounds of review comments.
What actually made it work
There was no clean handoff in this project, and that was the point. Materials moved in both directions. Testing happened on both sides. The client owned the assay and the clinical question. We owned the build, the documentation, and everything about whether it survives contact with a real user and a real shipping network.
Two teams, one objective: get the product through its study and onto a shelf.
What to take from this if you are building the device
- Get physical material into both laboratories early. Spec sheets do not tell you whether an adhesive bonds to your substrate. Two weeks of samples saves two months of pilot.
- Ask your manufacturer to test independently. If the only evidence is a supplier claim, you have a single point of failure in your evidence, not just in your supply chain.
- Treat the instruction sheet as part of the device. For an at home collection product, user error is a design defect, not a user problem.
- Put label and instruction revisions under document control before the first pilot, not before the first commercial lot. The habit is what protects you, and habits are cheaper to build early.
- Budget for two pilots. The first one tells you what you got wrong. That is what it is for.
- Decide unit identity before volume. Where the identifier comes from, when it is assigned, and how it reconciles with the laboratory’s own numbering. This is trivial to design and painful to retrofit.
Related reading: how to choose a preservation medium covers the chemistry decision in more detail, and fulfillment API and ecommerce integrations covers unit level identifiers and how they are returned.
