Medical device product development: the process from idea to production, phase by phase (2026)

The phases, from idea to production

This new product development process follows the order of decisions in Hardware Bible, Lisa Voronkova’s book. It holds for any medtech or biomedical device, and for healthcare product development in general; the device class only changes how much evidence each phase needs.

1. Clinical need and intended useWho uses the device, where, and why.
2. Regulatory route and payerThe class, the FDA pathway and who pays.
3. Concept and feasibilitySeveral concepts, and the riskiest physics tested.
4. Design and prototypesElectronics, firmware and mechanics, built in rounds.
5. Design controls and riskThe design file and the risk file, kept as you go.
6. Verification and validationProof against the specifications and the user needs.
7. Clinical evidence and submissionA study only if your claims need one, then FDA.
8. Transfer and pilot buildA process a factory can repeat.
9. Launch and post-marketHospital sales and the loop back into design.

1. Clinical need and intended use

Everything starts with a need written in one sentence that names the patients, the problem and the outcome you will change. Hardware Bible asks eight questions here, from who will use the device and where to why users will prefer it to the alternatives, and the answers become the user needs document and the intended use statement. Write that statement before the first prototype, because its wording decides the device class.

At OVA this sits in discovery, 3 to 8 weeks and 100 to 336 engineering hours (median about 180), which ends with requirements, an architecture, a risk list and an estimate for the next phase. The mistake the book meets most often is what it calls “the technology looking for a problem”. The free development readiness check shows where your project has gaps.

2. Regulatory route and who pays

The class and the pathway set your timeline, budget and evidence, so they are design inputs from the first week. Most Class II devices reach the US market through a 510(k), FDA’s premarket notification, which shows that the device is substantially equivalent to one already sold, the predicate. A new low to moderate risk device with no predicate goes through De Novo classification, and most Class III devices need premarket approval (PMA). FDA clears a 510(k), grants a De Novo request and approves a PMA; the FDA class and pathway finder and the 510(k) clearance time by product code tool show where devices like yours landed.

Payment is a separate question, and Lisa has seen “more good devices die of a missing billing code than of a failed test.” The phase ends with a short regulatory and commercial strategy. One founder in the book learned the route too late, after three years and $4 million, when it turned out to require a randomized trial.

3. Concept and feasibility

Concepts are generated in numbers and killed against criteria written in advance. The book scores each problem from 1 to 10 on frequency, severity, workaround difficulty and willingness to pay and drops anything whose product is under 1,000, and Wizard of Oz tests, where a person stands in for a function the device will automate, check usability before anyone pays for electronics.

When one physics question decides the device, OVA proposals price a feasibility gate at 400 to 1,000 hours, typically 400 to 560. The phase ends with a concept definition and a candid feasibility assessment. The book puts early technical feasibility at $50,000 to $150,000, a good fit for SBIR and STTR grants (federal small business research funding), and the SBIR and STTR fit check shows whether you can apply. The common mistake is settling on the first solution that works instead of testing three different approaches against each other.

4. Design and prototypes

Electronics, firmware and mechanics are designed together, in prototype tracks that each answer one written question about feasibility, clinical use, usability or manufacturing. In OVA proposals, design and the first prototype took 700 to 1,600 engineering hours and each further round 200 to 1,000, and taking a concept or proof of concept to 3 to 10 working prototypes, without design for manufacturing, came to 1,000 to 2,500 hours across eight proposals. A cellular radio on a battery got its own phase of about 500 hours in the one proposal that had it; the wearable battery life calculator checks a power budget against datasheet currents.

The usual mistakes are processes that will never scale and patient-contact parts in materials that differ from production. The phase ends with working prototypes and a golden sample, the pre-certification unit that becomes the reference for production.

5. Design controls and risk management

Design controls are the rules for how you develop the device itself: user needs become measurable design inputs, outputs are checked against them, and reviews sit between stages. Since February 2, 2026, FDA’s Quality Management System Regulation (QMSR) has incorporated ISO 13485:2016, the international quality standard for medical devices, and its design clause covers Class II and III devices and Class I devices automated with software. Risk management under ISO 14971, the risk standard, runs alongside.

OVA proposals for medical programs put documentation and verification preparation at 400 to 1,250 hours, 15% to 31% of the hours in the four Class II and sterile proposals that broke it out. You walk away with a design and development file (the old design history file), a traceability matrix that links each need to its tests, and a risk file. The book warns against backfilling, writing the records after the device is built. For a device with software that can reach the internet, the cyber device check shows whether section 524B, FDA’s cybersecurity law for connected devices, applies.

6. Verification and validation

Verification shows that the design meets its specifications; validation shows that the device meets users’ needs in realistic use. Verification usually covers:

• electrical safety to IEC 60601-1, the standard for medical electrical equipment
• electromagnetic compatibility (EMC) to IEC 60601-1-2
• biocompatibility to ISO 10993, on the finished, packaged and sterilized device
• software to IEC 62304, the software life cycle standard

Approve acceptance criteria before each test and justify the sample: 59 passing units support 95% reliability with 95% confidence. For the summative human factors study, the formal test of real users on the final design, FDA recommends at least 15 participants from each distinct user group, none of them your employees. The book plans 6 to 12 months for this testing on a moderate-risk device, and the mistake it names is writing pass criteria after the data comes in.

7. Clinical evidence and the submission

Most 510(k) devices are cleared on bench, animal, software and usability data; De Novo devices often need a clinical study, and a PMA almost always does. If you need one, match each claim to an endpoint and ask FDA through a Pre-Submission (Pre-Sub), its written feedback program, before you lock the protocol. The book allows 12 to 36 months for a study and warns against starting one before the bench work is done.

A 510(k) received in fiscal year 2025 took 141 calendar days on average to a decision, 76 FDA days plus 66 days of the company’s own time (510(k) review times), and recent De Novo requests averaged 248 to 278 days (De Novo review time). For FDA’s fiscal year 2027 the 510(k) fee is $28,653, or $7,163 for a qualified small business; the FDA user fee calculator adds up the rest.

8. Transfer to manufacturing and the pilot build

A prototype proves function. Production needs an assembly sequence, tolerances, test points, fixtures, second sources and a cost target, and each can change the design. Design transfer turns design outputs into production specifications, and the pilot build proves the process, with installation, operational and performance qualification (IQ, OQ and PQ) for steps you can’t fully check afterward, such as a seal.

Three OVA proposals put design for manufacturing, transfer and the pilot build at 1,000 to 2,300 hours; Lisa’s published breakdown puts design for manufacturing for an average product at 2,500 to 4,000. The book’s warning is to start the pilot before the submission, because one startup with a cleared, hand-built device lost 9 months to redesign. OVA builds batches of up to 200 units and hands volume production to your contract manufacturer; the cost calculator shows design for manufacturing as its largest single phase.

9. Launch and post-market surveillance

Clearance lets you sell a device; it doesn’t make a hospital buy it. The book maps everyone who can say no, from the value analysis committee that approves new products to IT security and the lead nurse, and allows 3 to 6 months for launch preparation.

Many teams treat launch as the finish line, but surveillance starts with the first unit shipped. A remark about a malfunction made to a sales rep in a corridor is a complaint, and if it suggests a death or serious injury, FDA’s Medical Device Reporting rules give you 30 calendar days to report. Post-production support runs 200 to 400 engineering hours a year in OVA’s published project sizes. A legacy device sold in the EU under an old directive certificate keeps its transition period only without significant changes; the EU MDR transition deadline check shows its end date.

Engineering hours by phase

The phase estimates below come from OVA’s hours benchmark, 35 proposals written from July 2024 to October 2026. They are estimates made before the work, not hours spent, and they exclude materials, test labs and FDA fees. The cost calculator turns hours into a budget: at $125 per hour, 1,000 hours cost $125,000.

Phase	What you walk away with	Engineering hours (OVA benchmark)
1. Clinical need and intended use	User needs, intended use; requirements, architecture and risk list from discovery	100 to 336 for discovery (median about 180)
2. Regulatory route and payer	Class, pathway, predicates, payment codes	Depends on device
3. Concept and feasibility	Concept definition, feasibility assessment	400 to 1,000 for a feasibility gate (typically 400 to 560), only when one physics risk decides the device
4. Design and prototypes	Working prototypes, then a golden sample	700 to 1,600 for design and the first prototype; 200 to 1,000 per further round
5. Design controls and risk	Design and development file, traceability matrix, risk file	400 to 1,250 for documentation and verification preparation (15% to 31% of a Class II or sterile program)
6. Verification and validation	Verification, validation and human factors reports	In the row above; lab tests are paid separately
7. Clinical evidence and submission	Study data where needed, engineering part of the submission	Depends on device
8. Transfer and pilot build	Production specifications, validated processes, pilot units	1,000 to 2,300 (adds about 30% to 60% on top of the prototype work, 2 proposals that priced it separately)
9. Launch and post-market	Complaint, reporting and field action procedures	Depends on device
Project management, all phases	Plans, coordination, reviews	6% to 18% of hours, usually 10% to 15%

What changes with the device class

Questions

Related pages

• Development servicesWhat OVA designs, builds and documents
• Development for startupsWhat to build at each funding stage
• Hours benchmark 2026Engineering hours from 35 proposals
• Cost calculatorEngineering hours and cost, phase by phase
• All free toolsFDA class, fees, 510(k) times and more

Sources

By Lisa Voronkova, CEO of OVA Solutions. Updated on October 23, 2026.