Where expert time actually goes

The first essay in this series established that medical-device conformity assessment is under structural load (The certification system under load). That diagnosis answers how large the pressure is. It does not yet answer the more operational question: where does scarce expert time actually go once a TD submission dossier is open?

If the answer were simply “not enough people,” the remedy would be linear: designate more Notified Bodies, hire more reviewers, wait for training pipelines to fill. Those steps matter, and staffing rose for years, but Team-NB’s 2025 survey also records the first decline in conformity-assessment headcount in more than a decade. Headcount is neither infinite nor permanently expanding. More importantly, linear hiring misses how assessment work is structured day to day. Much of an expert’s calendar is consumed by necessary work that does not require their highest judgment: finding material across large packages, reconstructing context after revisions, checking completeness, coordinating clarifications, and stitching a defensible record from fragments.

Capacity is not only a headcount problem. It is, to a large extent, an information-design and workflow problem: usable expert time is lost to waste work around judgment.

The reviewer as integration layer

A technical-documentation assessment is not a single document read in sequence. It is a web of claims, requirements, test reports, clinical evaluation, risk management, labeling, and change history, often delivered as heterogeneous packages that evolve through deficiency rounds. Someone must continually reconnect those pieces. When tools and process do not do that reconnection, the reviewer becomes the integration layer by default.

• Evidence hunting: locating the right report, annex, or version among hundreds of pages and supporting files.
• Context reconstruction: re-learning what changed after a revision, which finding was closed, and which claim still lacks a source.
• Completeness and structure friction: discovering late that required content is missing or hard to navigate, before substantive review can even settle.
• Coordination: questions, manufacturer responses, internal handoffs, and the administrative path from draft assessment to issued certificate.
• Reporting labor: turning reviewed evidence into a consistent, inspectable assessment record under institutional templates and quality rules.

None of that is frivolous. Safety depends on it. The problem is that these loops scale with package size, revision count, and organizational fragmentation faster than human attention does, and they often occupy the same people whose clinical and technical judgment is the scarcest input in the system.

Completeness as a capacity tax

In 2025, about 65% of responding Notified Bodies reported half or fewer MDR submissions complete enough to start review, with 23% under 25% and only 6% over 75% (Team-NB Survey 2025). The completeness check is a blunt signal before content review - the severe under-25% band improved versus prior waves, but the structural tax remains.

Incomplete or poorly structured submissions are not a moral indictment of manufacturers as a class. They are an operational tax on expert time: cycles spent requesting missing pieces, waiting, re-opening packages, and re-orienting after each exchange. Structure is not cosmetic. It is how scarce attention finds the claim-evidence path without becoming a full-time search function.

Much of the calendar is not pure document review

Manufacturer surveys reinforce that total certification time is not identical to “hours an auditor spends reading.” MedTech Europe’s 2024 report finds that, of total time spent on conformity assessments for IVDs and MDs, the majority of calendar time sits in pre-review and certificate-issuance phases; the remainder, less than half, covers actual documentation review (MedTech Europe 2024). Delay is a multi-phase system problem, not only a reading-speed problem.

Why headcount alone cannot keep up

Notified Body staffing grew for years, then contracted. Team-NB reports roughly 3,358 full-time-equivalent staff dedicated to conformity assessment among members in the medical-device sector in 2025, with about 8% fewer internal conformity-assessment employees and 21% fewer contractors year-on-year, the first such decline in over a decade (Team-NB Survey 2025). Training a reviewer still takes on the order of twelve to eighteen months. Capacity that is cut under a temporary workload dip cannot be restored on a product-sprint cycle.

When institutional knowledge is trapped in personal habits rather than shared structure, every new hire pays a long ramp, and every departure exports capacity the organization thought it had. Hiring is necessary and still insufficient if the information environment multiplies waste work. The binding constraint is not only seats. It is how much of each expert hour is available for judgment rather than search, rework, and orchestration.

The reframe

• How much reviewer time is spent reconstructing what a well-structured dossier would make obvious?
• How often does incomplete intake force loops that pure judgment never required?
• How much institutional knowledge depends on individual memory rather than designed process?
• What would change if evidence, requirements, findings, and dispositions lived in one navigable environment instead of across inboxes and file trees?

Speed and rigor are not opposites. Waste and rigor are. Expanding capacity without diluting scrutiny means reclaiming expert time from waste work, not automating the judgment that makes certificates trustworthy.

Structure submissions so claims attach to sources. Make review state visible so teams do not re-discover ownership and open findings from scratch. Preserve context across revisions. Surface uncertainty instead of burying it in confident prose. Keep disposition with qualified professionals. Expert capacity is consumed by the environment of the work. Improve that environment, and the same people can apply more rigor to more of the decisions that protect patients, without pretending the decisions themselves can be outsourced to software.

For how improvisation costs compound under binding stage ceilings, see the next essay (Binding assessment timelines).