Sampling is quoted as a duration — typically one to two weeks for a variant of an existing platform — and buyers treat it as waiting time. It is the opposite. It is the only window in which changes are cheap, and almost every expensive mistake in hardware sourcing comes from treating it as a formality.
The short answer: use the sampling window to freeze the decisions that become unchangeable later. Once tooling exists, changing the enclosure is a new tool. Once safety certification is filed, changing components invalidates documentation. Once production starts, changing anything requires the line to stop. None of those are expensive in week two, and all of them are expensive in month three.
What gets locked, and when
There are four things that must be firm by the time you approve samples, because each gates something downstream.
The cell or battery model. This one is the trap discussed elsewhere in detail, and it belongs here because it is frozen at sample approval. Changing it later invalidates transport documentation and restarts weeks of lead time that cannot be compressed by expediting anything.
The enclosure in its final material and finish. Surface texture, gloss level, colour standard, and any applied marking all get validated on physical parts. Fixtures, printing screens and colour references are made against this decision. If the surface standard is not agreed here, there is no reference for what production output should look like, and disputes about acceptability become unresolvable because nothing was ever defined.
The labelling and marking content. Regulatory markings, ratings, and any required symbols have requirements tied to the target market. These are not aesthetic. Getting them wrong means goods that cannot be sold in the intended market, discovered at customs rather than at the factory.
Port layout and mechanical interfaces. Which connector goes where, what orientation, what clearance around it. Users discover these immediately and reviewers mention them permanently, and they require tooling changes to fix.
The three sample stages, and what each one is for
Most programs produce more than one round, and each round has a different job. Conflating them wastes time.
Appearance and fit samples establish that the industrial design works. Does it feel right in the hand, do parts align, do seams sit evenly, is the weight acceptable. These questions can only be answered by holding the object. Nothing about how a rendering looked tells you whether the thing feels cheap.
Engineering samples establish electrical and thermal behaviour under real conditions. This is where you run sustained load and measure temperatures, verify actual output rather than claimed output, check behaviour when input is out of spec, and test the edge cases that only appear after twenty minutes of continuous operation rather than twenty seconds.
Pre-production samples establish that the production process can repeat what the prototype did. This is the most-skipped round and it is where defects that are invisible in hand-built units show up — because a prototype was assembled carefully by a technician, and unit number three thousand will be assembled quickly by someone on a line.
Skipping the third stage to save two weeks is how programmes discover a problem at volume that they could have caught in a day.
Four tests worth doing yourself
You do not need a laboratory. You need to actually use the thing rather than inspect it.
Run it at full rated load long enough to reach thermal equilibrium and then measure surface temperature with anything that reads temperature. Thirty minutes of sustained load answers questions that no amount of specification reading will.
Use it with the worst input you can find, not the best. An ageing adapter, a long thin cable, a source at the edge of specification. Products designed only around ideal inputs fail constantly in the field where inputs are never ideal.
Charge and discharge it repeatedly for several days and watch for behaviour changes. Anything that degrades over cycles — capacity, heat, negotiation reliability — shows up quickly at the consumer timescale.
Drop it once, from about waist height, onto something hard. Then check every function. This is not a certification test. It is the cheapest possible way to find a connector that will not survive normal ownership.
The document to produce before approving
Before you sign off on samples, produce one short internal spec listing what "acceptable" means, covering at least these:
Maximum surface temperature at full load, and where it was measured.
Minimum sustained output, stated as a number rather than a percentage.
Acceptable cosmetic standard, ideally referencing a physical sample retained by both parties. This single item prevents more disputes than everything else combined, because most disagreements about quality are disagreements about an unstated standard.
Required markings and their exact content.
Then keep one approved physical sample, labelled and dated, on both sides. Every later batch should be compared against that object. Without it, "it looked different" is unanswerable in both directions.
The mistakes that are expensive
Approving based on photographs or video rather than holding the unit. This is becoming more common and it removes exactly the information sampling exists to provide.
Changing your mind after approval because someone senior finally looked at it. Every change at this point is a tooling or documentation event, and the schedule impact is disproportionate to how small the change feels.
Sampling one unit and treating it as representative. Sample three, test all three. Variance between units tells you about process control, and one unit tells you nothing about variance.
Treating the sample as the product. It is not. It is evidence about what the production process is capable of, and it is only trustworthy if enough other units agree with it.
If you are planning a first order, the three pages worth reading first are MOQ and pricing structure, packaging decisions and compliance timeline. Programme details: ODMKit.
Top comments (0)