Device classification is a strategic decision, not a paperwork exercise at the end of development. The class determines your conformity assessment route, your clinical evidence burden, your Notified Body involvement, and effectively your burn rate for the next two years. Decide the class before the first prototype, architect the product to defend that class, and revisit the decision only when the intended purpose genuinely changes. Founders who classify after they build almost always discover they built for the wrong class — and by then, the cost of correcting course is measured in quarters, not weeks.

By Tibor Zechmeister and Felix Lenhard. Last updated 10 April 2026.

TL;DR

  • Classification under Article 51 and Annex VIII of Regulation (EU) 2017/745 should be fixed before hardware and software architecture decisions — not reverse-engineered from a finished prototype.
  • The class drives conformity assessment route, clinical evidence scope, QMS depth, technical documentation depth, Notified Body involvement, and total budget. Decide it early or pay for the consequences later.
  • The lowest defensible class — backed by a specific Annex VIII rule and a stable intended purpose under Article 2(12) — is the strategic target. Not the lowest wished-for class. Not the highest cautious-consultant class.
  • Intended purpose is the single highest-leverage lever. A correct, disciplined intended purpose can move a product from Class IIa to Class I, or from medical device to wellness product, without changing the technology.
  • Reclassify only when the intended purpose genuinely changes, a new MDCG guidance alters interpretation, or the Notified Body formally disputes the class. Do not reclassify to chase a lower number.

The Graz wellness pivot — months wasted on the wrong path

A Graz-based team came to us eighteen months into development. They had a working prototype, a pitch deck, a small team, and a regulatory strategy that assumed a Class IIa path. They had spent most of a year building the QMS around that assumption and planning a clinical evaluation around that assumption. The product was technically competent, the team was experienced, and the regulatory plan was internally consistent.

The problem was not the plan. The problem was the premise. When we pulled the intended purpose apart — paragraph by paragraph, claim by claim, against the Article 2(1) definition of a medical device — it turned out that the genuine use of the product was wellness, not medical. Nothing about the technology changed. Nothing about the user experience changed. What changed was the honest regulatory positioning: the product, as actually used, was not a medical device at all. With a disciplined re-framing of the intended purpose, the product moved out of MDR scope.

The months of QMS work, clinical strategy work, and Notified Body prep were not recoverable. That is what "classifying after you build" costs when it goes wrong. The decision that could have been made in the first week of the project — for the price of a careful reading of Article 2(1), Article 2(12), and Annex VIII — was made in month eighteen, after the runway had already paid for the wrong project.

This is not a rare story. It is the most common classification pathology we see. The fix is the strategic approach: decide the class first, build the product to defend that class, and revisit only when something material changes.

Why the class drives everything

Classification is not one decision among many. It is the decision the rest of the regulatory project inherits from. Under Regulation (EU) 2017/745, the class determines:

  • The conformity assessment route. Class I (non-sterile, non-measuring, non-reusable surgical) is self-declared. Everything above Class I requires a Notified Body. The cost step between self-declaration and Notified Body involvement is the single largest in the MDR landscape.
  • The clinical evidence burden. Clinical evaluation is always required, but the depth, the need for a dedicated clinical investigation, and the acceptability of equivalence or literature data all scale with class.
  • The QMS scope and depth. A Class IIa or higher project needs a QMS audited by a Notified Body. A pure Class I project has QMS obligations but no full Notified Body QMS audit.
  • Technical documentation depth. Class IIb and Class III documentation is broader and deeper than Class IIa, which is broader and deeper than Class I.
  • The timeline. The gap between a Class I self-declaration and a Class III full assessment can be two years or more.
  • The budget. Every line item in the budget is class-dependent. Classifying wrong does not just shift a number — it shifts every number.

If the class drives all of this, classifying last means building blind. The founders who build first and classify second are not saving time. They are betting the runway on a guess.

The early decision points

The strategic approach to classification fixes four things before the first prototype is committed.

1. The intended purpose. Written as one paragraph that will hold up on the label, in the IFU, on the website, and in the clinical evaluation. Article 2(12) of Regulation (EU) 2017/745 defines intended purpose as the use supplied by the manufacturer across these documents. If the paragraph cannot be written consistently across all four surfaces, the intended purpose is not ready.

2. The qualification question. Is the product actually a medical device under Article 2(1)? This is the question that eliminates the Graz wellness pivot scenario up front. If the genuine use is wellness, lifestyle, or general fitness, saying so in month one is cheap. Saying so in month eighteen is expensive.

3. The target rule in Annex VIII. Which classification rule do we expect to apply, and why? The answer is not "we think it's IIa." The answer is "Rule 11, second paragraph, because the software is intended to provide information used for therapeutic decisions, and the decisions do not carry the escalation triggers in the subsequent paragraphs." A specific rule. A specific reading.

4. The architecture implications. Given the target class, what features, modules, or claims would push us into a higher class? Which of those are essential, and which are nice-to-have? If a nice-to-have feature would push the product from Class I to Class IIa, the strategic question is: is that feature worth a Notified Body certificate, six to twelve additional months, and the corresponding cost? Sometimes yes. Often no.

These four decisions, made in the first weeks of a project, are worth more than any later optimisation. They are the inputs to every spreadsheet, every hiring plan, every investor conversation that follows.

The lower-class-defence reality

Classification is sometimes a lookup and sometimes a judgment. Rule 8 on long-term implantables is close to a lookup — the device either contacts the central circulatory system or it does not. Rule 11 on software, Rule 10 on active diagnostic devices, and several of the special rules involve genuine interpretive judgment. Where judgment exists, a defensible lower-class position backed by evidence is worth preparing.

We worked on a combined hardware-software device where the software side was unambiguously Class IIa and the hardware side was disputed. The question was whether the hardware carried a measurement function in the regulatory sense, which would have pulled it into the Class I measuring (Im) subcategory with the corresponding Notified Body involvement for metrological aspects. The analysis was not trivial — it required a careful reading of the intended purpose, of how the device actually reported values, and of whether those values were used for clinical decisions or for user orientation. A well-documented argument that the hardware was Class I without measurement function held up under Notified Body scrutiny. This is not gaming the rules. It is applying them precisely, with evidence, where they allow judgment.

The lesson for early-stage strategic classification is this: when a rule allows judgment, the judgment is yours to make, provided it is defensible and documented. The founders who assume the highest plausible class "just to be safe" often pay for Notified Body scope they did not need. The founders who assume the lowest plausible class without evidence fail the Notified Body review and have to rebuild. The strategic answer is the lowest class you can defend with a specific rule, a specific intended purpose, and a specific argument — prepared before construction, not scrambled together afterwards.

MDCG 2021-24 (Guidance on classification of medical devices, October 2021) is the authoritative reference for interpretive classification decisions. MDCG 2019-11 Rev.1 (June 2025) is the reference for software qualification and classification. Both are required reading before the early class decision is locked.

When to commit

Commit to a class when three conditions are met: the intended purpose is stable and written consistently across the four surfaces (label, IFU, promotional materials, clinical evaluation outline); the Annex VIII rule has been identified specifically, including any implementing rules and special rules that might apply; and the architecture implications of that class are understood and accepted by the engineering team.

In practice this means the commit happens after the Purpose Pass but before the Architecture Pass — somewhere in the first four to eight weeks of a serious regulatory project, not in month six or later. The commit is documented with the specific rule citation, the specific intended purpose paragraph, and a brief rationale. That document is the anchor the rest of the project is built on.

When to reclassify

Reclassify when one of the following happens, and only then:

  • The intended purpose genuinely changes. A new indication is added. An existing indication is removed. The patient population shifts. The decision-support claim changes. Any of these can move the class.
  • A new MDCG guidance or Borderline Manual entry changes the interpretation of a relevant rule. MDCG documents get revised. When they do, re-read the rule that applies to your device.
  • The Notified Body formally disputes the class. If the NB says the class is wrong, argue the point if you have grounds, but accept the outcome if you do not.
  • A design change crosses a rule boundary. For example, a previously non-invasive device that becomes short-term invasive, or a previously transient device that becomes long-term.

Do not reclassify because the current class feels expensive. Do not reclassify because a competitor chose differently. Do not reclassify because a new consultant has an opinion. Reclassification has to follow the same discipline as the original classification: a specific rule, a specific intended purpose, a documented rationale.

The cost of discovering the wrong class late

Discovering the wrong class late has three distinct costs, and founders usually only think about the first.

The rework cost. Every artefact built to the wrong class has to be rebuilt. QMS procedures scoped for Class IIa when the device is actually Class IIb. Technical documentation that omits sections a higher class requires. Clinical evaluation built on equivalence data when the higher class needs investigation data. This cost is visible and painful.

The runway cost. The rework takes months. Months the startup does not have. Months the investors did not budget for. Months the team spends rebuilding rather than progressing. This cost is larger than the rework cost itself because it extends the zero-revenue period.

The credibility cost. Investors, partners, and Notified Bodies form opinions about a team that misclassified. Not permanent opinions, but real ones. The team that classified correctly in week three is telling a different story than the team that classified correctly in month fifteen. Credibility is harder to rebuild than documentation.

The total cost of discovering the wrong class late is typically six to twelve months of runway and a significant hit to trust. The cost of classifying correctly early is a careful few weeks of intended-purpose work, Annex VIII reading, and documented commitment. The math is not complicated.

The Subtract to Ship angle on strategic classification

The Subtract to Ship framework treats strategic classification as the decision that enables every subsequent subtraction. Once the class is fixed, every activity, document, procedure, and feature in the project is checked against the class. If a Class IIa project is carrying QMS procedures only required at Class IIb, those procedures are subtracted. If a Class I project is carrying a clinical investigation plan only required at Class III, that plan is subtracted. Subtraction is only possible when the class is known — without a class, subtraction has no reference point.

The discipline is to subtract only what the class does not require, never what the class does require. This is the line between lean and reckless. A Class IIa project that subtracts required Class IIa activities is not lean — it is non-compliant. A Class IIa project that subtracts activities only required at Class IIb is lean. The class is the ruler.

This is also why strategic classification is the second of the four passes in the framework — after the Purpose Pass (is this a medical device, and what is the intended purpose?) and before the Evidence Pass and the Ship Pass. Classification without purpose is guesswork. Evidence and shipping without classification is waste.

Reality Check — Is your classification strategic or accidental?

  1. Can you write your device's intended purpose in one paragraph that is consistent with what your label, IFU, website, and clinical evaluation will all say?
  2. Have you fixed the class before committing to hardware and software architecture, or are you reverse-engineering the class from a finished prototype?
  3. Can you cite the specific Annex VIII rule that governs your device — by number and paragraph — and quote it from the actual MDR text?
  4. Have you checked MDCG 2021-24 and, if your device is software, MDCG 2019-11 Rev.1 for current interpretation of the rule that applies to you?
  5. Have you identified which nice-to-have features would push you into a higher class, and have you decided which of those are worth the higher class and which are not?
  6. Do you have a documented rationale for your class that a Notified Body auditor could accept on first reading, without additional explanation?
  7. Do you have a trigger list for when you would reclassify — and is "a new consultant has an opinion" not on it?

Frequently Asked Questions

When should I classify my medical device under the MDR? Classify before you commit to hardware and software architecture. In practice, this means in the first four to eight weeks of a serious regulatory project, after the intended purpose is stable and before significant engineering investment. Classifying after the prototype is built inverts the decision chain — the class should constrain the product, not the product dictate the class.

Can classification really change how I build the product? Yes, significantly. The class determines QMS scope, clinical evidence requirements, technical documentation depth, Notified Body involvement, and timeline. Features that would push a product from Class I to Class IIa, or from Class IIa to Class IIb, create material cost and time differences. Knowing the class early lets founders decide which features are worth the higher regulatory burden and which are not.

What is the lowest defensible class, and how is it different from the lowest wished-for class? The lowest defensible class is the lowest class that can be justified by a specific Annex VIII rule, a specific intended purpose under Article 2(12), and an argument a Notified Body will accept. The lowest wished-for class is whatever the founder hopes for. The first is strategic. The second is fantasy. The Notified Body will eventually enforce the difference.

Should I ever reclassify to a lower class mid-project? Only if the intended purpose genuinely changes, a new MDCG guidance alters the interpretation of the rule that applies, or a documented analysis of the original classification reveals a defensible error. Reclassifying because the current class feels expensive, or because a competitor chose differently, is not a strategy — it is a shortcut waiting to collapse at the Notified Body review.

What is the single most common classification mistake founders make? Classifying after the product is built, instead of before. The second most common is copying a competitor's class without reading Annex VIII against their own intended purpose. Both mistakes share the same root cause: treating classification as paperwork instead of strategy.

Sources

  1. Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices, Article 2(1) (definition of medical device), Article 2(12) (intended purpose), Article 51 (classification), Annex VIII (classification rules, definitions, implementing rules, Rules 1-22). Official Journal L 117, 5.5.2017.
  2. MDCG 2021-24 — Guidance on classification of medical devices, October 2021.
  3. MDCG 2019-11 Rev.1 — Guidance on Qualification and Classification of Software in Regulation (EU) 2017/745 — MDR and Regulation (EU) 2017/746 — IVDR, October 2019, Revision 1 June 2025.

This post is part of the Device Classification & Conformity Assessment series in the Subtract to Ship: MDR blog. Authored by Felix Lenhard and Tibor Zechmeister. Strategic classification is the decision every later regulatory decision inherits from. When the intended purpose is genuinely unstable, or the rule interpretation is genuinely contested, a working session with a sparring partner who has walked dozens of teams through the same decision is usually faster than another month of internal debate.