What It Actually Costs to Change One Component After You've Placed the Order
A single component swap after your PO can cost more than the whole part order. Here's the four-bucket breakdown — scrapped material, respin, requalification, schedule — with real numbers.
PRODUCT DEVELOPMENT
Peakingtech
8/23/20268 min read
The samples were fine. That's the part people remember.
A client sent us a design where one MOSFET had been validated on ten hand-built boards. Everything passed. We built the first fifty units on the production line and two of them failed thermal rise. Not dead — just outside spec, on a part that had behaved perfectly in prototypes. Batch-to-batch variation in a part nobody thought was risky.
Changing that MOSFET cost roughly forty times the price of the MOSFETs themselves.
This is the cost that first-time hardware founders never budget for, and it isn't a line on any bill of materials. It's the cost of changing your mind after you've committed. Below is how that cost is actually built up, what it looks like at each stage of a project, and how to decide whether to change a part or ship with it.
The four buckets of a change
Every component change resolves into the same four cost categories. They don't all fire every time. The skill is knowing which ones your change will trigger before you approve it.
Scrapped material — components you've already bought and can no longer use
Respin — new bare boards, new stencil, new setup, and the engineering time to produce them
Requalification — the testing and certification that has to be repeated
Schedule — lead time that restarts, and whatever the delay costs your business
Most founders only see bucket 2, because it's the one that arrives as an invoice. Bucket 1 is usually the largest and the most invisible. Bucket 4 is the one that kills companies.
Bucket 1: Scrapped material — the reel problem
Here's the mechanic that surprises almost everyone.
Electronic components are not sold in the quantity you need. They're sold in MOQ (minimum order quantity) and MPQ (manufacturer pack quantity) increments. A specialty inductor might only ship as a full 3,000-piece reel. Your build needs 200. You buy 3,000.
If you then change that inductor, 2,800 pieces become dead stock. They aren't returnable. Most distributor stock is sold under NCNR terms — non-cancellable, non-returnable — for anything that isn't a standard catalogue passive, and for factory-ordered parts NCNR is close to universal.
So the real question when you change a part isn't "what does the new part cost." It's:
How many were bought, versus how many the build consumes?
Were they bought as cut tape, or as full reels?
Were they NCNR?
Have they already shipped, or can the order still be pulled?
Who owns them under your contract — you, or the factory?
That last one is where relationships get tested. Under a turnkey arrangement, the factory bought the parts against your PO, and the exposure is contractually yours. Under consignment, you bought them and you already know. Either way, someone is holding a box of parts nobody will use, and the invoice reflects it.
A useful rule: your material exposure on any component is the full purchased quantity, not the used quantity, until that build ships. Ask your factory to state MOQ exposure per line item at quotation, not after the change.
There's a second, smaller layer here. Line attrition — parts lost to feeder loading, pick-and-place drops, and test scrap — means the factory ordered 3–5% more than your build quantity to begin with. That overage disappears with the change too.
Bucket 2: Respin — cheaper than you fear, if the footprint matches
A component change only forces a board respin if the footprint, pinout, or thermal envelope changes. This is why the single highest-leverage decision in a hardware project happens at the schematic stage: choosing a part whose approved alternates share a footprint.
If the footprint holds, your respin cost is close to zero. If it doesn't, you're paying for:
(Ranges are typical for low- to mid-volume builds; your actual figures depend on board complexity and quantity.)
For a small hardware brand, a footprint-compatible swap and a footprint-breaking swap are two completely different events. It's worth asking your engineer which one you're looking at before anything else.
Bucket 3: Requalification — what has to be redone, and what doesn't
This is where founders either panic unnecessarily or get blindsided. Both are avoidable, because the rules are reasonably clear.
Usually does not require retesting:
A same-value, same-package passive from a different manufacturer
A pull-up resistor value change on a low-speed line
Cosmetic or silkscreen changes
Usually does require some retesting:
Anything in the power path — regulators, MOSFETs, inductors, bulk capacitance
Anything that changes switching frequency or edge rate (EMC re-test territory)
Any change to the antenna, RF front end, or ground plane geometry (radio re-certification)
Any change to the battery, charger IC, or protection circuit (safety, and UN 38.3 for shipping)
Any change affecting creepage, clearance, or insulation in a mains product
The expensive ones are radio and safety. A wireless product that needs a fresh FCC/CE radio test after a layout change is looking at a five-figure bill and a multi-week lab queue — and the lab queue is often the bigger problem.
The practical implication: the same change costs radically different amounts depending on where it sits in the circuit. A capacitor swap in a digital section is a non-event. The same swap in the feedback loop of your DC-DC converter can pull EMC back onto your critical path.
Bucket 4: Schedule — the one that actually hurts
Change a part and the clock restarts. Not on your project — on the supply chain.
The new part has its own lead time. If it's 4 weeks, you've lost a month. If it's a semiconductor with an allocation problem, you've lost a quarter. Meanwhile:
Your booked production slot goes to another customer, and the next one may be weeks out
Freight bookings and container space, if arranged, get rescheduled
Retail or crowdfunding delivery commitments slip, in public
Your team keeps burning payroll on a program that isn't shipping
For a funded startup, a two-month slip is often the single largest number in this entire article — larger than material, respin, and testing combined. It just never appears as an invoice, so it doesn't feel like a cost.
What a change costs at each stage
The same swap gets dramatically more expensive as the project matures. This is the table worth internalising:
A worked example
A 2,000-unit connected consumer device. After the pilot build, thermal testing shows the buck converter runs hotter than acceptable, and the team decides to change the regulator IC and its inductor.
Scrapped material. Regulator bought at MOQ 2,500 — 2,500 scrapped. Inductor bought as a 3,000-piece reel — 3,000 scrapped. Combined material write-off: $1,900.
Respin. New footprint, so new boards ($520 for the batch), new stencil ($160), line re-setup ($250), and a test fixture pin relocation ($400). $1,330.
Requalification. Switching regulator change means EMC pre-compliance again, plus a partial radiated emissions re-test at the lab. $4,200, with a three-week lab queue.
Schedule. New regulator lead time is 8 weeks. Production slot lost. Launch slips one quarter. Direct payroll and overhead burn during the slip: $28,000 — before counting the missed selling season.
Total identifiable cost: roughly $35,000, on a component change where the parts themselves were worth under $2,000.
The BOM saving that motivated choosing the original regulator, incidentally, was about $0.11 per unit — $220 across the run.
(Figures are illustrative of a typical small-batch program, not a quotation.)
When to change, and when to ship with it
Not every problem justifies a change. Use these five questions:
Is it a safety or compliance issue? Then it isn't a decision. Change it.
Does the failure rate compound in the field? A thermal margin problem gets worse in a hot warehouse in July. A cosmetic issue doesn't. Marginal thermal, mechanical stress, and electrolytic ageing all compound — treat them as change-now problems.
Can you screen for it instead? If 4% of units fail a specific test and the rest are solid for the product's life, 100% functional test plus a scrap allowance is sometimes cheaper than a respin. Run the arithmetic before assuming otherwise.
Is this the last batch on this design? If a revision is already planned for six months out, batch the change into it rather than paying the respin twice.
What is the warranty exposure if you don't? A return costs far more than a unit — shipping both ways, diagnosis, refurbishment or scrap, and the review that follows. At consumer scale, an extra 2% failure rate will usually dwarf the cost of fixing it now.
How to not be here in the first place
Most late changes trace back to three decisions made months earlier.
Approve alternates at the schematic stage, not at the crisis. Every critical line item should carry at least one engineer-approved, footprint-compatible second source, documented in the BOM before release. This single practice removes most of bucket 2 and all of bucket 3 for the changes it covers.
Stage your purchasing. Don't buy production quantities against a design that hasn't survived a pilot build. Buy pilot quantities, run them on the real line at the real speed, and only then release the reel-quantity PO. The extra per-unit cost on the pilot batch is cheap insurance against MOQ write-off.
Get MOQ exposure stated in writing at quotation. A good quote flags every line item subject to MOQ or MPQ, states the attrition allowance, and identifies which parts are NCNR. If your quote doesn't do this, you're carrying risk you can't see. Ask for it — any competent factory can produce it.
Do the thermal and margin work before the pilot, not after. Most late changes we see are power-path problems that a proper derating review would have caught on paper.
FAQ
Can I change a component after the PO is placed? Yes, but what it costs depends entirely on whether parts have been ordered, whether they've arrived, and whether they're NCNR. Tell your factory the moment you suspect a change is coming — a purchase order that hasn't been released to the supplier can often still be pulled at little or no cost.
Do I have to pay for unused components if I change the design? In almost all cases, yes. Components bought against your PO are your exposure, and most are non-cancellable and non-returnable. Under a turnkey arrangement the factory bought them on your behalf; the cost still lands on your invoice.
What happens to the leftover parts? They sit in inventory. If they're a common part, they may be usable on a later build or resold on the excess market at a steep discount. Specialty and custom parts are typically written off.
How much does a PCB respin cost? For a low-volume build, the direct cost is usually in the low hundreds to low thousands of dollars — new boards, stencil, and line setup. The direct cost is rarely the problem. Requalification and lead time are what make respins expensive.
Does changing a resistor require re-certification? Usually not, if it's a same-value, same-package part in a non-critical circuit. Changes in the power path, RF section, or safety-relevant circuitry often do require retesting. When in doubt, ask before you approve — a lab will tell you the scope for free.
What does NCNR mean on a BOM? Non-cancellable, non-returnable. Once the order is placed, you own the parts regardless of whether you use them. Most factory-ordered and specialty components are sold this way.
When is the cheapest time to change a component? Before layout. The cost curve is roughly exponential from there: near-zero at schematic, moderate before the PO, high once parts are in stock, and severe after certification.
Where we come in
Most of the changes described above were preventable at the review stage. Peakingtech's BOM and DFM review flags MOQ exposure, missing second sources, NCNR risk, and power-path margin before anything is purchased — so the decision to change a part stays cheap.
If you're approaching a production release and want a second set of eyes on the BOM before the PO goes out, get in touch.
SERVICES
Peakingtech © 2025. · Shenzhen, China · Privacy Policy · Terms of Service
Contact Form
Metal Parts
Peakingtech® is a registered trademark of Peakingtech Co in the United States.
COMPANY
CONTACT