Validation workflow
How a dataset is built
Validation is arithmetic before it is evidence. Every plan resolves to a real
device count, every draw runs beside an approved method, and every rejection is recorded.
The output is a quality file — not a verdict.
Validation matrix
device count = validation events × samples per event
50
Starter · 12 weeks · method check
250
Pro · 24 weeks · powered comparison
100
CE readiness · 12 months · file build
01
Fix the matrix, not a vague allotment
You commit to a number of validation events and the samples drawn at
each. Fifty devices is a method check; two hundred and fifty runs a powered study. The
device count is the product of the two — agreed up front, not estimated later.
02
Compare against the approved method
Each RUO draw runs in parallel with an approved reference collection.
We measure agreement, bias and limits against the method that already holds — never the
device judged against itself.
03
Characterise sample stability
Time-to-analysis, temperature and handling are logged for every
sample. Stability windows are measured, not assumed, so the dataset reflects how the
device behaves outside a perfect bench.
04
Track every failure and rejection
Hemolysis, underfill, clotting, mislabelling — rejections are counted
and categorised, never quietly dropped. The failure rate is part of the result and travels
with the data.
05
Compile a QA file you can hand over
Protocols, raw data, comparison statistics and rejection logs assemble
into a single quality file — the evidence a notified body, lab partner or CE route expects.
We build the file; certification stays with the bodies that grant it.
06
Where the file goes next
A completed dataset is a starting position for a CE route — it is not
certification, not a clearance, and not permission to report patient results. The device
stays validation-only unless a CE route or an accepting clinic/lab adopts a valid route.