Your manufacturer will build a thousand copies of a board that doesn't work.
And they won't warn you first.
That sounds like an accusation, but it isn't, because catching your design mistakes was never their job in the first place.
Most product creators assume the exact opposite, that since the manufacturer is the expert, surely they'll flag anything wrong with the design before production starts.
And honestly, it's a fair assumption, because a good one really does know more about manufacturing than you ever will.
But that one assumption is also how people end up paying for an entire production run of perfectly assembled boards that simply don't work.
So let's break down what a manufacturer actually checks before building your PCB, the design mistakes they will never catch, why the system works this way on purpose, and what you need to do before handing off any design.
A DFM review sounds like just the kind of safety net you'd want.
It stands for design for manufacturability, and nearly every manufacturer runs one on your PCB before production.
The name alone causes half the confusion, because design for manufacturability sure sounds like somebody reviewed your design.
When your factory comes back and says your design passed DFM, it feels like a stamp of approval, like a team of experts combed through your work and gave it a clean bill of health.
But a DFM review only asks one question.
Can we assemble this board on our production line with good yield?
Yield just means the percentage of boards that come off the line put together correctly, and it's the number your manufacturer cares about most, because bad yield costs them money.
So a DFM review is really a yield check for them, not a design review for you.
Nobody in that process is judging whether your circuit does what it's supposed to do.
The actual checklist your factory runs during a DFM review is all about their machines and their process.
They'll check that your component footprints match the packages you chose, at least for the obvious mismatches like a part that's physically bigger than its footprint.
Your pad sizes and solder mask openings get checked against the limits of their assembly process, and they build the solder paste stencil to match their own equipment.
They'll also confirm your components have enough spacing and clearance for their pick and place machines and their reflow oven.
Then come panelization, fiducials, and tooling holes, which are basically the features their equipment needs to grab, align, and build your board.
Parts that are out of stock or getting discontinued will get flagged too, though mostly because availability affects their sourcing and their quote.
And finally, they'll check your assembly drawing for orientation and polarity markings so parts don't get placed backwards.
Now look at that list one more time, because every single item on it protects their yield, and not one item tells you whether your product will even work.
Your manufacturer will never check whether your circuit actually works.
Schematic errors, wrong pinouts, missing pull-up resistors, none of that is anywhere in their scope, and most factories never even see your schematic because the manufacturing files are all they need.
Some will run a netlist check, which just confirms your board matches the connections in your design, not that your design is right, so a mistake in the design still gets built.
Component selection is the same story.
If you put the wrong value resistor in a voltage regulator's feedback divider, which is the pair of resistors that sets the output voltage, they'll place that exact wrong part on every board with impressive precision.
I designed voltage regulators at TI for years, and I can promise you a regulator will faithfully lock onto whatever wrong voltage those resistors tell it to, even if it's so high it causes downstream damage.
The same goes for a part that's electrically wrong for the job, like a transistor that can't handle your load current, since a part number is just a part number to them.
Thermal design mostly gets ignored too.
The one exception is that your manufacturer wants your board and its parts to survive the reflow oven.
That's why they'll look at things like thermal reliefs, which are the little copper spokes that keep a large copper area from pulling heat away from the pad during soldering, along with the overall copper balance that keeps the board from warping.
But that only matters during assembly, not for how hot your product runs in normal use.
Your factory will happily build a board that's going to cook itself inside your enclosure, because operating temperature has nothing to do with how the board gets assembled.
That one usually shows up as a product that works great in open air on your bench, then starts rebooting or dying once it's sealed inside a case on a warm day.
Certification readiness is another thing your manufacturer will never check.
If your design has an EMC problem, meaning it radiates too much electrical noise to ever pass FCC testing, they'll still build every board flawlessly, and this exact thing happens every single day.
Firmware is a similar story, though it depends on who's building your product.
If you're only having bare boards assembled and shipped to you, nobody ever loads your firmware, so any software bug is yours to find later when the boards arrive.
But a manufacturer that builds your complete product will usually flash your firmware to run a functional test on each unit, and loading your code and running it is nothing like reviewing it.
And that functional test only exists if you define the tests, pay for the test fixture, and spell out exactly what to check, test by test.
A functional test will only ever catch what you told it to catch.
And bring-up itself is always your job either way.
Bring-up just means getting a brand new design to power on and talk to all its parts for the first time, and that's development work that happens on your bench long before production ever starts.
Every manufacturer operates on a principle called build to print.
That means their contract is to reproduce your data package exactly as you sent it, with no interpretation and no improvements.
The moment they modify your design, or even bless it as good, they start owning its failures.
So if they suggest a change and your product later fails in the field, guess who gets the angry phone call and maybe the lawsuit.
Staying out of your design protects them, which makes silence the safe business decision every time.
The business model points in the same direction, because manufacturers make their money on PCB assembly volume, not on engineering hours.
A real design review would be expensive engineering labor they have no good way to bill for.
And any factory that carefully reviewed every customer design would have to charge more for assembly, which means losing business to the shop down the street that doesn't.
Some manufacturers do offer design review as a separate paid service, but even those reviews lean heavily toward manufacturability and sourcing rather than whether your circuit works.
And none of this makes your factory lazy or shady.
It's a clean division of responsibility that keeps assembly prices low, and it means the design side of that fence belongs entirely to you.
A design mistake reaching production almost always plays out the same way.
Your first article, meaning the very first boards off the line, comes back looking beautiful.
Every solder joint is shiny, the automated optical inspection passes, and the factory proudly reports a clean run.
So you green-light the full quantity, the boards ship, and then reality sets in.
Maybe none of them work at all, or worse, they work fine on your desk and start failing in the field a month later.
Whatever the mistake was, maybe a swapped connection or a wrong part or an unverified footprint, your manufacturer built every board correctly, and that's the painful part.
You pay for the scrap, you pay for the rework, and you eat the schedule slip, all while they send you an invoice for a job well done.
Protecting yourself from a bad production run comes down to five steps you take before the design ever leaves your hands.
Step #1: Get an independent review of your schematic and PCB layout before you release the data package.
At TI, every design I created had to survive a formal design review where a room full of engineers picked apart my work in excruciating detail, and that process existed because even the most experienced designers miss things in their own work.
You don't need a room full of engineers, but you absolutely need at least one qualified set of eyes that isn't yours.
That review can come from a design firm, an experienced engineer you trust, or a service that does design reviews all day long.
Step #2: Validate your prototypes under real conditions instead of just on the bench.
Run the product inside its actual enclosure, at its real operating temperature, on its real power source, and for longer than feels necessary, because cold boards behave differently than warm ones, and batteries sag in ways your bench supply never will.
Step #3: Define what working actually means by writing a functional test spec, then budget for the test fixture that goes with it.
If you can't describe the pass and fail criteria for your own product, your manufacturer definitely can't test for them.
Step #4: Deliver a complete data package, meaning your Gerbers, your centroid file that tells the machines where every part sits, your BOM with approved alternate parts, your assembly drawing, and that test spec.
Every gap in that package turns into a guess that someone at the factory makes for you.
Step #5: Inspect the first article yourself before green-lighting the full run.
If your production is overseas, insist they air-ship you a real sample, not just a report with photos, because a photo can't tell you whether the board powers up or runs your firmware.
Verify the part orientations, confirm the board powers up, load your firmware, and exercise every major function before you tell them to build the other 999 boards.
If anything looks off, this is the cheapest moment you'll ever have to stop and fix it.
Your manufacturer's job is to build exactly what you send them, and your job is to make sure that what you send them is right.
Get those two responsibilities straight, and a good factory becomes one of the most valuable partners you'll ever have.
Talk soon,
John Teel
Predictable Designs
P.S. My Hardware Academy Open House closes Sunday night. For $1 you get full access, including asking our engineers and product experts whatever you're stuck on. Get in for $1.