There's a moment in every device company's life that feels like crossing a finish line. The clearance letter arrives, the champagne gets opened, and the whole organization exhales and shifts into manufacturing mode. Development is "done." The engineers who lived in design reviews and verification protocols start thinking of their work as production support: tweak a fixture, adjust a cycle time, swap a supplier, keep the line running.
And that's where many post-market quality problems are born. Launch is a phase change, not a finish line, and the teams that get burned after clearance are the ones who started treating manufacturing changes as routine engineering work the day the letter arrived. They're not routine. They never become routine. And I just had a conversation on the podcast that makes this exact case..
Mike Dolphin is the CEO of GuideStar Medical Devices, where he is bringing an epidural safety device to market, a closed-loop system he describes as automatic braking for the needle. Before medtech he engineered at NASA's Jet Propulsion Laboratory and ran scientific research and experimental design at Ernst & Young, and this is his second device through the full concept-to-clearance gauntlet. His framing for the episode was a question: is manufacturing a step you complete after development, or is it part of development? His answer, and mine, is that manufacturing is development. It takes real engineering and real know-how, and you do not have a fully developed product until the manufacturing process that produces it is developed too.
So what's the point? The point is, if manufacturing is R&D, then changes to manufacturing are design activity. And design activity comes with design control discipline, before clearance and (critically) after it.
What manufacturing-as-R&D actually looks like
Two stories from Mike's own line make this concrete.
The first is about tolerances. GuideStar's parts carry very tight dimensional requirements, in the tens of microns. Their injection molder's standard answer was that nothing tighter than 100-micron tolerance could be guaranteed. Mike pushed on it from first principles: the mold itself is cut metal, and cut metal can hit 50 microns, so the uncertainty is in shrinkage, not machining. The solution was to build three molds stepped roughly 20 microns apart, shoot parts from each, and keep the one that landed on target. It was the middle one (naturally). And here is the payoff: once the right mold existed, every part produced landed within about 10 microns of the next, hundreds of thousands of units of repeatability. That's experimental design, hypothesis and iteration and measurement.
That's R&D.
The second story is equally valuable. One of GuideStar's production runs of parts came out brittle. Same plastic, same mold, same spec on every document. But parts kept breaking in their hands. The cause: a different operator ran the line and shortened the cycle time to save a few minutes, the plastic cooled differently, and the material properties changed. Nothing in the paperwork changed, but something in the product had definitely changed.
Think about what that story means for the "manufacturing is routine now" mindset. A parameter that appeared on no drawing, controlled by a person who appeared in no change order, altered the mechanical behavior of a medical device. As Mike put it, the entire process needs to be well-defined and consistent, because a validated process is only validated for the way it was actually run. Every knob on that line is a design input that needs to be controlled.
Why teams get casual after clearance
The temptation is real. Before clearance, change is the job. Mike's team iterated molds, stress-tested parts off the line on their own bench, and treated every failed dimension as information. Change felt cheap because learning was the product.
Then two things flipped simultaneously. First, the regulatory stakes. Mike said it clearly: once you have clearance on a final finished product, you sign off on everything, and you make it very hard to make changes, because it is really important that you don't make them casually. Your 510(k) was cleared on a specific device produced by a specific validated process. Drift from that process and you are, in a meaningful sense, shipping a device that hasn't been reviewed.
Second, the economics flip, and Mike used examples of hard numbers. Working with a contract manufacturer, every change ran through his quality team and engineers, then through the manufacturer's quality team and engineers, duplicated review at a marked-up rate. A change that should have cost a thousand dollars in internal effort landed closer to ten thousand. Each change stretched from one week to two or three. Ten changes in a year added something like thirty weeks of timeline, with payroll and overhead running the whole time. His summary: it was literally costing millions. That experience is a big part of why he suspects that, with the funding, building manufacturing in-house from the start would have been faster and cheaper.
Put those together and you get the picture. Post-clearance, changes are simultaneously more consequential and more expensive to process, so teams start routing around the process. The tweak goes in as "maintenance." The supplier swap gets a verbal okay. The cycle time drifts because nobody wrote it down as a controlled parameter in the first place. This is how change control gaps start, and it's why what good change management looks like across the device lifecycle is a discipline and not a form: the cost of the process tempts you to skip it when skipping it is most dangerous, and the message is that the discipline has to get cheaper to follow than to avoid.
Where Mike and I don't necessarily see eye to eye
Let's talk about the most provocative thing Mike said, because I pushed back on it during the episode and I'll push on it a bit here. His view is that before clearance, beyond documenting decisions and signing off on design reviews, you don't need signed-off documents, a quality management system (QMS), and all the design control that goes with it. His team carried heavy QA sign-off overhead during manufacturing development, while things were still changing daily, and it cost them real time and money. Lock things in when things need to be locked in, he argues, and for them that lock point was clearance.
And, to a large extent, I get where he's coming from. What Mike is describing as unnecessary is not design control. It's ceremony misapplied to the wrong phase. Because listen to what his team actually did from the beginning: wrote down every design decision and the reasoning behind it, recorded test results and how they informed the next choice, held and signed design reviews, verified dimensions in-house before paying for formal testing. That's the substance of design control. The thing that burned him was bolting a full production-grade approval workflow onto work that was still exploratory, which is a right-sizing failure, not an argument against the discipline.
Where I part ways with him is on the letter of it: for a device headed to a 510(k) submission, design controls are a regulatory requirement during development, not an option you exercise at clearance. Postponing the formal system until the end usually means reconstructing your design history file backwards, from memory, under submission pressure, and I've watched that archaeology project eat more time than the sign-offs ever would have. The fix for QA overhead that slows development is a system sized to the phase you're in, light where iteration is the focus, rigorous where decisions become commitments, not the absence of a system.
But that doesn't mean Mike and I don't agree, because it's the whole point we were trying to make: after clearance, the lock is real and the discipline is non-negotiable. His own brittle-plastic run is proof. Post-market design control is not bureaucratic residue from development. It's the only mechanism you have for making sure the device in the field stays the device that was cleared. Every post-clearance manufacturing change deserves the same questions a pre-market design change would get. What design outputs does this touch? What does it do to risk? Does it invalidate a validation? Does it rise to the level where the Food and Drug Administration (FDA) needs to hear about it, or is a documented letter-to-file justification the right call? An engineering change order that walks through those questions is the difference between the brittle run getting caught at incoming inspection and the brittle run getting caught in a patient.
And Mike's requirement-creep warnings affect both directions here, which I love. There's no regulation demanding a fireproof filing cabinet, no rule that packaging must be tested exactly sixty times, and his advice to read what the requirement actually says and stop doing things that are not needed is the same right-sizing principle applied post-market. The goal is not maximum process, but a change control system lean enough that engineers actually use it and rigorous enough that a shortened cycle time can't slip through.
From clearance to control
If manufacturing is R&D, then it never stops being R&D. The experiments just get more expensive and the subjects are in the field. The teams that internalize this treat the clearance letter as the moment the design freezes and the discipline transfers from the lab to the line.
So if your device is cleared, or close to it, here's the self-check. Are the process parameters that matter, the cycle times and cure profiles and torque values, actually captured as controlled specifications, or do they live in an operator's habits? Does every manufacturing change, however small it feels, get evaluated against design outputs, risk, and validation status through a real engineering change order? And is your change process cheap enough in time and friction that your engineers route through it instead of around it?
If any of those answers made you think, they're not going to close themselves. The phase change is coming either way. The only question is whether you cross it with a system that keeps the device you ship identical to the device you cleared, and not out in the field, with parts breaking in your hands.
Keep reading
If you are building out your post-market change control process, these related guides go deeper on the specific components:
- The engineering change order process, step by step
- The complete guide to change management and change control
- What comes after launch: a guide to post-market maturity
- How to master the CAPA process
- Predetermined change control plans, explained
- How risk management and design controls connect
Design control discipline does not end at clearance. It just changes address, from the design history file to the device history record. Get a demo of Greenlight Guru to see how connected design and change control keep the device you ship identical to the device you cleared.
BONUS RESOURCE: Change Order Template
