Why MDR Assessments Still Take 13–18 Months — and What Actually Shortens Them

Overview

The Commission's 18th Notified Bodies Survey, published in March 2026 with data as at 31 October 2025, describes a system that is working harder and not getting materially faster.

The headline figures: 33,175 MDR applications submitted against 17,549 MDR certificates issued. For IVDR, 3,634 applications against 2,194 certificates. Certificate volumes are growing — MDR up 18% and IVDR up 23% against the June 2025 snapshot — but the gap between applications lodged and certificates issued leaves an estimated 15,000-plus applications in the pipeline across 51 active Notified Bodies.

On duration, the survey is blunt. 71% of Notified Bodies report that securing an MDR product certificate takes more than 13 months. For IVDR the figure is 90%, with 40% reporting 19 to 24 months. The typical band is 13 to 18 months, extending to 19–24 months where both product and quality management system certification are required.

These numbers have been broadly stable for several survey cycles. That stability is the interesting part. Notified Body capacity has increased substantially since 2021 — more designations, more assessors, more throughput. Duration has not fallen proportionally. That tells you the binding constraint is not primarily assessor headcount.

What the survey says about causes

The survey's own diagnosis points repeatedly at application quality rather than assessment complexity. Manufacturers submit technical documentation missing critical sections, referencing superseded standards, or failing to address all applicable MDR and IVDR requirements. Classification disputes and incomplete applications recur as themes across editions.

This is worth stating precisely, because it is routinely misread as blame-shifting. The claim is not that manufacturers are careless. It is that the elapsed time of a certification is dominated by round-trips, not by review effort.

The arithmetic is straightforward. A technical documentation assessment might represent six to ten weeks of actual reviewer working time. A 15-month elapsed duration therefore contains something on the order of nine to twelve months in which the file is not being actively assessed. That time is spent waiting: for the manufacturer to respond to a deficiency, for the reviewer to become available again once the response arrives, for a second question set that could have been asked alongside the first, for an expert panel or EU reference laboratory opinion.

Each round-trip has a cost far exceeding the time to answer the question. A manufacturer takes four to eight weeks to compile a response. The file then re-enters the reviewer's queue and waits for a slot. If the same reviewer is unavailable, context has to be rebuilt. A file that generates four sequential question sets can easily spend a year in this cycle without anyone behaving unreasonably.

Why EU 2026/977 does not fix this on its own

Commission Implementing Regulation (EU) 2026/977, which applies from 25 February 2027, sets binding maximum timelines: 30 days for application review and contract, 120 days for QMS auditing, 90 days for product verification, 20 days for decision and certification.

Read quickly, that looks like an eight-month cap. It is not, and the reason is Article 3.

Article 3 permits the Notified Body to interrupt the timeline where the manufacturer must address non-compliances or answer justified questions — four times during QMS auditing, four times during product verification, once during application review. Interruptions for EMA, regulatory authority, expert panel or EU reference laboratory opinions are not counted against those caps at all.

The clock-stops are exactly the round-trips that generate the 13–18 months. The Regulation caps how many there can be and requires each to be justified, agreed in duration and confirmed in writing. It does not eliminate them, and it does not constrain their length beyond requiring agreement.

Article 2(4) closes the obvious escape route: expiry of a maximum timeline, or exhaustion of the permitted interruptions, is explicitly not sufficient reason to refuse a certificate. Notified Bodies cannot manage to the number by rejecting difficult files.

What 2026/977 genuinely changes is visibility. From 1 January 2028, Article 4(4) requires each Notified Body to publish an annual report by 30 April giving the percentage of assessments completed within the maximum timelines, the median duration from application to certification, and median total cost. Elapsed duration stops being an industry aggregate in a Commission survey and becomes a published, comparable, per-Notified-Body figure.

The interruption caps do create one real forcing function. With four permitted interruptions on a product verification, the cost of a poorly-scoped question set rises sharply. A reviewer who issues deficiencies serially as they encounter them will exhaust the budget. The Regulation makes consolidated questioning an operational necessity rather than a matter of style.

What actually shortens the elapsed time

Four things move the number. Only one of them is about reviewing faster.

1. Front-loaded completeness checking

The highest-leverage intervention happens before assessment starts. An application screened against the full Annex II and Annex III structure at intake — every section present, every referenced standard current, every applicable general safety and performance requirement addressed or justified as inapplicable — either enters assessment complete, or generates its deficiencies in one batch on day one rather than in month four.

The difference in elapsed time between "this file is missing its usability engineering file" discovered at intake versus discovered eleven weeks into review is not eleven weeks. It is eleven weeks plus a full round-trip cycle plus a queue re-entry.

Article 1 of 2026/977 pushes in the same direction, requiring a defined information set before a quotation is issued. That is a completeness gate at the front of the process, and it exists because incomplete intake is understood to be the problem.

2. Consolidated rather than serial deficiency reporting

The single largest avoidable cost in a technical documentation assessment is the second question set that could have been part of the first.

This is harder than it sounds. A reviewer working sequentially through a large file naturally encounters issues in document order, and there is real pressure to send an early finding to the manufacturer so they can start working on it. That instinct optimises the wrong variable: it starts one clock earlier at the cost of guaranteeing an additional cycle.

Complete-file analysis before the first deficiency is issued is the discipline that matters, and it is genuinely difficult to do manually across a file running to thousands of pages in multiple documents with cross-references between them. This is the specific point where systematic gap detection across the whole file has real leverage — not because it reviews faster, but because it makes a single consolidated question set achievable.

3. Parallel rather than sequential phases

Article 2(2) of 2026/977 now requires QMS auditing and product verification to run in parallel where the assessment proceeds under Annex IX, provided the technical documentation input feeds the audit programme. Organisations that scheduled these sequentially will find this is the most immediate operational change the Regulation forces.

4. Reviewer continuity

Every queue re-entry after a clock-stop carries a context-rebuilding cost that is invisible in effort accounting and substantial in practice. A reviewer returning to a file after six weeks re-reads to re-establish position. Where continuity cannot be preserved, the handover cost is higher still.

The mitigation is that the assessment record itself has to carry the context — which requirements are satisfied, by which evidence, with what open questions — in a form that survives the reviewer stepping away. Assessment state that lives in a reviewer's working memory and a set of personal notes does not survive it.

What does not shorten it

Two things are worth naming because they are frequently proposed.

Reviewing faster. Reviewer working time is not the dominant term in the elapsed-duration equation. Compressing six weeks of analysis into four saves two weeks against a fifteen-month cycle, and it does so by reducing the depth of the assessment — which is both regulatorily indefensible and, given Article 2(4), unnecessary.

Adding assessors without changing the process. Capacity growth since 2021 has not produced proportional duration improvement. Additional assessors increase the number of files in progress; they do not reduce the number of round-trips per file.

The measurable version

From April 2028, every Notified Body will publish its median duration from application to certification and its percentage of assessments completed within the statutory maximum timelines. Manufacturers will compare them. Designating authorities will compare them.

The organisations that look good in that comparison will not be the ones whose reviewers work fastest. They will be the ones that catch incompleteness at intake, issue consolidated rather than serial deficiency sets, run phases in parallel where the Regulation permits, and maintain assessment state that survives an interruption.

For the evergreen treatment of these levers, see how to reduce technical assessment turnaround time. For the documentation problems that generate the round-trips in the first place, see common MDR documentation gaps that delay technical assessment. For the process this all sits within, see the technical documentation assessment process.

Conformo builds AI-assisted review infrastructure for Notified Bodies, Certification Bodies and TIC organisations. Every finding is traceable to its source in the technical file, and every conformity decision remains with a qualified reviewer.