The short answer: unit count is set by destructive boundaries, not by the number of test items
Manufacturers almost always open with "how many units should we send?" The question is framed the wrong way round. Whether the request form lists a dozen items or two dozen has no direct bearing on how many samples you prepare. What actually sets the number is how many of those items drive a sample into an irreversible condition, and whether those paths collide with one another.
For a wheelchair programme organised around the ISO 7176 series and the GB/T 18029 series, the items sort into groups according to how much of the sample they consume, and you can read the unit count straight off the table below.
| Item category | Effect on sample condition | Can it share a sample? | Sampling rule |
|---|---|---|---|
| Dimensions, mass, marking, accompanying-document checks | Leaves the sample essentially unchanged | Yes, and these should run first | Shares one unit with the performance items |
| Performance and handling (static and dynamic stability, braking, gradient and obstacle climbing, driving performance and similar) | Condition is normally recoverable, but wear and battery drain accumulate | Mostly yes, provided the sample is restored to its initial condition between items | Shares the unit with the check items; bring spare consumables |
| Strength, impact and fatigue (the scope covered by ISO 7176-8) | Plastic deformation, cracking, weld or fastener failure is likely | No; the sample is generally scrap afterwards | Split by irreversible path, one dedicated unit per path |
| Items relating to use as a seat in a motor vehicle (the scope covered by ISO 7176-19) | Involves restraint and dynamic conditions; neither sample nor hardware is reusable | No | Dedicated unit, complete with the full set of restraint hardware |
One line to close it out: check items and performance items can share a single sample; every irreversible strength, impact or fatigue path takes a sample of its own; items relating to use as a seat in a motor vehicle take another dedicated sample. Applied to a product in a normal configuration, that lands in the range of "one general-purpose sample plus however many units the destructive paths require" — not the single unit that many manufacturers assume will do.
Why stacking destructive items onto one sample breaks the data
Trying to save samples by running several strength and impact sub-items back to back on the same unit is something we see constantly. The data that comes out has two structural problems.
First, the result stops representing the design and starts representing what the previous item left behind. Once the first item is finished, residual deformation and redistributed stress are already present in the structure, even when nothing is visible from the outside. The load path has changed. If the next item passes, that does not show the product meets the requirement; if it fails, you cannot tell whether the cause is a design weakness or damage inherited from the previous run.
Second, failure attribution is broken. The report says a crack appeared at a given location, and the reviewer will ask which item caused it. With a shared sample you have no answer, so the work is repeated — new samples, a new slot in the queue, a new invoice, and the whole project slides by a full cycle.
The working rule is plain: if an item can plausibly leave plastic deformation, cracking, weld damage or fastener failure behind, it takes a sample of its own. That is not conservatism, it is what keeps the conclusion in the report traceable.
Sequencing: three tiers ordered by irreversibility
Test order does not follow the running order of the standard's table of contents. It follows the damage done to the sample, from lightest to heaviest.
| Tier | What goes in it | Entry conditions | Cost of getting the order wrong |
|---|---|---|---|
| Tier one: no change | Dimensional measurement, mass measurement, marking and accompanying-document checks, appearance and assembly inspection | Sample is in production condition with all accessories present | Run after the destructive items and the dimensions and appearance are already distorted, so the data is unusable |
| Tier two: recoverable | Stability, braking, handling and driving performance, energy consumption and range-related items | Tier one complete and recorded, consumables are new parts | Mixed in with destructive items, the performance data is contaminated by structural damage and the whole block has to be repeated |
| Tier three: irreversible | Strength, impact, fatigue and other destructive items | Confirmed that this sample will carry no further items | Placed ahead of other items, one sample writes off the entire submission |
There is one detail here that regularly gets missed: tier two needs internal ordering as well. On powered products the performance items keep drawing down the battery, and drift in battery condition makes the later data unstable. The sound approach is to group the items that are sensitive to battery condition into one block, finish them together, and handle battery condition consistently between blocks rather than topping up as you go.
Can this sample carry the next item: three self-checks
The lab records sample condition after every item. Manufacturers should run their own screen as well, so that a sample carrying hidden internal damage does not keep moving down the list.
- Structural re-check: no visible deformation, cracking or coating loss on the primary load-bearing members, welds, or the folding and adjustment mechanisms; the travel of every adjustment mechanism matches its initial condition.
- Fastener re-check: once every fastener has been re-torqued, none of them shows a tendency to loosen repeatedly. Repeat loosening means the mating faces have been crushed or the threads are damaged, and that sample should not carry any further items.
- Consumables re-check: tyres, castors, cushion and backrest, battery and similar parts have been replaced with new parts of the same specification, and the replacement is written into the record.
Fail any one of the three and you change samples. The test is not "can it still be used" but "can the data still represent the product".
Accessories and spares: leave these out and you rework anyway
| What travels with the sample | Why it has to be there | What happens if it is missing |
|---|---|---|
| Battery and charger matching the declared configuration | Several performance and safety items on powered products depend on the real supply condition | Data produced with a substitute battery is not accepted and the work has to be rescheduled |
| Full set of cushions, backrest, armrests, footplates and legrests | Dimensions, mass distribution and stability all depend strongly on these parts | Missing parts distort the mass distribution and the stability conclusions are void |
| Instructions for use, marking, packaging labels | Accompanying-document and marking checks are made against the physical items | The check items cannot be carried out and the report has gaps |
| Optional accessories for optional configurations | Configuration-family coverage has to be backed by physical parts | Adding a configuration afterwards means opening a new submission |
| Common tools and spare fasteners | Re-torquing and restoration are needed between items | The initial condition cannot be restored on site and the whole schedule slips |
Three scenarios that actually bite
Scenario one: one sample only, impact run before stability. The sample picked up frame deformation during the impact item that was not easy to spot, and the stability data shifted with it. The reviewer saw that the two data sets contradicted each other and the whole report was thrown out. What the manufacturer paid was not just the fee, it was a full scheduling cycle.
Scenario two: the destructive sample and the performance sample were not the same configuration. To save time the two units were pulled from different production batches, with differences in cushion and footplate. The two halves of the report described products in different states, the reviewer asked for retesting on a harmonised configuration, and half the work was written off.
Scenario three: an engineering prototype was submitted. The structure differed from production parts and the fastening was still hand assembly. The results looked good, but during a customer audit they were judged not to represent the production product, and the whole submission was repeated.
What these three have in common is that the problem was never the testing. It was a decision taken before the samples shipped. Fixing unit count and sequence at the planning stage costs nothing; changing them once the samples have arrived means new samples and a new place in the queue.
Working out the unit count before you ship
Sort the requested items into the categories above and count how many irreversible paths there are. Confirm whether the destructive items need separate samples for different structural locations. Work out the base count as "one general-purpose sample plus one per irreversible path", then add a dedicated sample for the items relating to use as a seat in a motor vehicle if your request includes them. Once you have the number, write the unit count, the configuration and the accessory list into the test request form so that nothing has to be added after arrival. How to fill in that form so it is not sent back is a topic of its own; our expectations for request information are set out in the testing services overview.
Sampling priorities differ by product family as well. Electric wheelchair testing adds the question of keeping battery and control system condition consistent across samples. Manual wheelchair testing turns mainly on the reuse boundary of the frame and the folding mechanism. Mobility scooter testing calls for attention to how whole-product mass distribution knocks on to the stability items. As for a product that has already been tested and then has its motor, controller or upholstery changed, the retest scope depends on which path the change lands in — a separate discussion that we will not open up here.
Ask us to run the numbers for your product
Send us the product category, the configuration list and the target markets, and we can propose a sample count, a test sequence and an accessory list against your actual item list, so that nothing has to be reworked once the samples arrive. SUNGO Mobility Testing Lab is the dedicated wheelchair and mobility aid testing lab within the group, accredited by CNAS, CMA and IAS (USA), with laboratories in Shanghai and Hefei. Technical enquiries: +86 132 4819 8029, or request a quote and we will come back with scheduling at the same time.