V&V gets written as one word on most medical device teams, said in one breath, tracked in one column of the trace matrix. That blur is where audit findings live. Verification vs validation asks two different questions that demand two different kinds of evidence, and when a team treats them as a single activity, one of the two usually ends up under-proven.
Here is the whole distinction in one line. Verification asks did we build the device right. Validation asks did we build the right device. Everything else is detail hanging off that sentence.
The one-sentence test, unpacked
Verification checks your output against your input. You wrote a design input that said the infusion rate shall be accurate to within a stated tolerance. Verification is the bench test, the calculation, the inspection that proves the design output, the actual pump behavior, meets that specification. It is internal. It compares the thing you built to the requirement you wrote for it.
Validation checks your device against the user's real need in real use. The design input traces back to a user need: the clinician needs to deliver a dose safely at the bedside under time pressure. Validation proves the finished device, in the hands of an actual intended user in an actual or simulated use environment, meets that need. It is external. It compares the thing you built to the world it has to work in.
The reason this matters is that a device can pass every verification test and still fail validation. You can build exactly to spec and discover the spec was wrong, or incomplete, or right on the bench and dangerous at the bedside. Verification cannot catch that. Only validation can, because only validation asks the question verification never asks: was this the right thing to build at all.
Verification vs validation: different questions, different evidence
Because the two ask different questions, they need different proof, and this is where teams that conflate them get into trouble at audit.
- Verification evidence traces to the specification. Test protocols against design outputs, results with pass or fail against defined acceptance criteria, inspections and analyses. The trace is design output to design input. For software this is where IEC 62304's unit, integration, and system verification activities sit, each checking the code against its requirement.
- Validation evidence traces to the user need. Usability and human factors data, clinical evidence where the pathway requires it, simulated-use studies with representative users. The trace is finished device to user need and intended use. This evidence usually cannot be gathered on the bench alone, because the whole point is real use.
- The two are not interchangeable, and one cannot stand in for the other. A stack of beautiful verification reports does not validate the device. A successful summative usability study does not verify that the pump meters within tolerance. Auditors look for both, traced separately, and a program that documents one thoroughly and the other by assumption has a gap whether or not anyone has found it yet.
If your trace matrix has a single V&V column, you have almost certainly hidden this gap from yourselves. Split it. Make the verification trace and the validation trace visible as two chains, because they resolve two different risks.
Where risk sets the depth
None of this means every requirement gets the same treatment. FDA's design controls, defined in the design controls provisions the agency enforces under its quality system requirements at 21 CFR Part 820, and now harmonized with ISO 13485, expect verification and validation to be planned and documented, but they do not demand identical rigor across every line item.
This is where risk-based thinking scopes the work. A requirement tied to a serious hazard earns deep verification and, where the hazard touches use, validation with real users. A low-risk cosmetic requirement still gets a test case, but the depth of coverage is proportionate. The move that keeps this honest is tying V&V planning to your risk analysis under ISO 14971, so the rigor of each proof matches the harm it guards against, rather than spreading effort evenly and thinly across everything. That is also the spirit of FDA's Computer Software Assurance thinking for software: assurance effort scaled to risk, not to page count.
Where the confusion actually costs you
The practical cost is not usually a failed audit line, though it can be. It is later and more expensive than that. A team that treated validation as a formality, that ran a token usability check and called it validated because verification was thorough, is the team that ships a device which does exactly what the spec said and still causes use errors in the field. The verification was real. The validation was assumed. And assumed validation is how a technically correct device becomes a recall.
Get the two questions straight, trace them separately, and let your risk analysis set how hard you push on each. That is most of what good V&V is.
Where Nopmark fits
The verification vs validation distinction is foundational to medical device software work, and it is exactly the kind of thing that is obvious once someone draws the two chains for you and quietly expensive until they do. Nopmark's Software Validation course, part of the Medical Device Bundle, walks through design verification, design validation, and how IEC 62304 and your risk file fit together, taught by accredited instructors. If your team says V&V as one word, that course is where it stops being one word.
FAQ
What is the difference between verification and validation in medical devices?
Verification confirms the device meets its specification, did we build it right. Validation confirms the device meets the user's real need in real use, did we build the right thing. They need different evidence and cannot substitute for each other.
Can a device pass verification but fail validation?
Yes, and it is the case that matters most. Building exactly to specification does not help if the specification was wrong or incomplete. Validation is the only activity that tests the device against the actual user need rather than against your own requirement.
Does IEC 62304 cover both verification and validation?
IEC 62304 structures software verification across unit, integration, and system levels against requirements. Validation of the finished device against user needs and intended use sits in the broader design controls and usability engineering activities, not in IEC 62304 alone.
Where does risk management fit into V&V?
Risk analysis under ISO 14971 sets how much rigor each requirement's verification and validation earns. High-harm requirements get deep proof and, where use is involved, validation with representative users. Low-risk items still get a test case at proportionate depth.
Is one trace matrix column enough for V&V?
No. A single V&V column hides the fact that verification traces to the specification and validation traces to the user need. Keep them as two separate chains so both risks are visibly covered.
Ready to make verification vs validation something your whole team gets right by reflex? Nopmark's Software Validation course covers the full design controls picture, taught by accredited instructors with worked medical device examples.