Choosing a Contract Manufacturer (From the Inside)
What the CM shortlist actually looks like, what to negotiate before the quote, and the failure modes that never show up in a capability deck.
Every contract manufacturer has the same capability deck. SMT, through-hole, conformal coating, potting, final assembly, box build, ISO 9001, IPC Class 2 or 3, ESD floors, a wall of logos. The decks are interchangeable. The factories are not.
The real variance does not show up in the deck. It shows up in who responds to your quote email, what they ask for before they price, whether they push back on your DFM, and how the sales engineer answers when the answer is “we can’t do that.” Choosing a CM from marketing is a coin flip. Choosing one from the inside is a different exercise.
The CM is the most expensive decision in a hardware program after architecture. A wrong CM will not show up at quote time. It shows up at first pilot, when yield is forty percent and the sales engineer who answered your emails is no longer on the account. By then the tooling is cut, the NRE is spent, and the program is a year late. Choose on the signals that predict the sixth month, not the first call.
The CM you want is not the biggest CM
Most founders, on first contact with manufacturing, ask for introductions to the largest CMs they can find. Foxconn-adjacent names. Tier-one EMS. The instinct is usually wrong.
A large CM quotes off a standard cost model. Your program, at ten thousand units a year, is a rounding error. You get a capable account team, a standard MSA, and a line slot you share with whatever else is running that week. You do not get engineering depth. You do not get pushback on DFM. You get the account manager, who gets the line engineer on a call next week.
What you usually want is a mid-size CM — eighty to four hundred people, two to six lines, where your program is meaningful but not their entire revenue. At that size, the sales engineer who quotes you is the operations lead or sits two meters from them. DFM feedback is real because it comes from the person who will build the thing. Minimum viable runs are a few hundred to a few thousand units. Tooling amortization is negotiable because they care about keeping the line loaded.
The exception is consumer programs targeting a million units a year from day one. Those belong at tier-one CMs. Everything else — industrial, medical, connected consumer, pro-grade — is better served by a mid-size CM that will engage with the engineering.
Five signals of a CM worth engaging
Capability decks grade everyone at ninety-five percent. The real shortlist comes from five signals visible only in the first engineering conversation. Not the commercial one. The engineering one.
A capable CM sends notes on your drawings before the quote. Tolerances too tight for the process. Assembly steps that cost more than they need to. Panelization they would redraw. A footprint that does not match the land pattern their line is tuned for. If the first response is a clean quote with no questions, they either did not look or cannot.
A serious CM wants to see what they will source before they price. They check lead times, MOQs, package availability. They flag parts at end-of-life or on allocation. If the CM prices off the drawing without reviewing the BOM, the quote is fiction. The real price surfaces when the first PO hits and three line items are on six-month lead.
End-of-line test is the hardest part of the transition. A CM that has built test jigs, written sequences, and integrated them into MES within the last twelve months has the muscle memory. One that says “we’ll work with your fixture” and has never built one will cost you six weeks at PVT. Ask to see examples. Ask who wrote the sequence.
Most CMs quote any volume to win. What matters is whether that volume is economical for them. A CM whose typical run is fifty thousand units will price your five-hundred-unit pilot as a favor — and treat it as one. A CM whose typical run is two thousand will set up the line for yours. Ask directly: what is the smallest run where you start to make money.
Not the logo wall. An actual reference in the same product class — connected consumer, wearable with a silicone seal, medical device with a battery. Someone they finished a program with, not just quoted. If the CM cannot produce one, they have no pattern match. They will learn on your dime.
What to negotiate before the quote, not after
The biggest mistake with a CM is treating the quote as the output of the negotiation. It is the input. Everything that matters — tooling, NRE amortization, BOM ownership, line allocation, quality responsibility — gets decided before the first unit price lands. Negotiating after the quote is arguing with a number already written down.
Negotiate these, in order, before you see the quote:
Tooling ownership and amortization. Who owns the tools, where they live, what happens if the relationship ends. Amortized tooling feels lighter on cash flow but traps you — the tools functionally belong to the CM until amortization clears. Up-front tooling with a title-transfer clause in the MSA costs more in month one and is cheaper every month after.
BOM ownership. If the CM buys, they mark up parts five to fifteen percent on landed cost to cover financing, inventory risk, and handling. If you consign, you carry inventory risk and lose their volume pricing. Under a million units a year, consigned BOM with CM-handled kitting is usually right. Above that, CM-purchased with audited cost files and a locked markup schedule.
MOQ and run size. Minimum run, changeover cost for smaller runs, lead time between PO and first unit. A two-thousand-unit MOQ on a product selling a hundred a week is an eighteen-week cash trap.
Capacity commitment. Units per month at the line speed you need, without competing with their other customers. Contractual or best-effort. Notice period if they need to move you.
Quality terms. Who pays for rework on yield failures. The first-pass yield floor. The RMA window for workmanship versus design failures. Write these down before the first PO.
Change order process. How ECOs are priced, lead time, sign-off. The most commonly forgotten clause. The one that hurts most in ramp.
If the CM will not discuss these before the quote, they are not a partner. They are a vendor.
Red flags that never show up in a capability deck
These separate a CM worth engaging from one you will spend a year unwinding from. They appear in the first three conversations if you know where to look.
The sales engineer answers BOM questions without deferring to operations. A sales engineer who quotes lead times, yields, and process capability off the top of their head is either wildly experienced or bluffing. In our experience, the latter. A trustworthy CM has a sales person who says “let me check with ops and get back to you.” That sentence, early and often, is a buy signal.
No line engineer on the DFM review call. If DFM review is run by sales and a document gets emailed back, there is no engineering depth in the room. Real DFM review has the line engineer, the SMT process engineer, and the quality lead on the call arguing about tolerances. That argument is what you are paying for.
They quote faster than they ask questions. A clean quote inside forty-eight hours of sending a BOM is a red flag. Either they did not cost the BOM, or a salesperson is closing a number before due diligence. The right response for a complex electromechanical product is one to three weeks.
They have no opinion on your test strategy. A CM that shrugs at your proposed coverage and says “whatever you want” has not thought about test. They will not catch the defects your test misses. You will.
Their MES is a spreadsheet. Every mid-size CM should track builds with per-unit traceability in an actual system. If that requires the line supervisor to check a binder, your warranty, field failure analysis, and firmware update workflows fall apart at scale.
Quality reports arrive on request, not on schedule. If you have to ask for first-pass yield, RMA rates, and defect Pareto data, they are not watching them. Nobody is catching the drift that turns a ninety-five percent yield into seventy-two percent over four months.
How to structure the relationship
A CM relationship runs on three documents: the MSA, the quality agreement, and the SOW per program. Most founders read the MSA and skim the other two.
The MSA covers the commercial frame — payment terms, pricing adjustment mechanisms, tooling ownership, IP assignment on anything the CM co-develops (test jigs, process parameters, firmware flashing scripts), confidentiality, dispute resolution. Negotiate the tooling transfer clause and the change-order language.
The quality agreement defines what “acceptable” means — first-pass yield floor, RMA rates, defect classifications, corrective action timelines, FAI requirements, Cp/Cpk targets. It is the document that gets cited when yield slips, and the one most programs never sign. Figuring it out as you go means arguing with your CM in the middle of a yield crisis with no shared definition of acceptable.
The program SOW covers the specific build — volumes, timelines, milestones, NRE scope, deliverables. Line allocation and capacity commitments live here.
The relationship also needs cadence. Weekly operations call during ramp. Monthly review at steady-state. A named account engineer on their side, a named production lead on yours. A shared issues log. Not nice-to-haves. How the relationship survives its first real yield problem. We cover the DFM side of this loop in DFM: The Decisions That Don’t Show Up on Drawings — the DFM dialogue starts during selection, not after contract signing.
When to walk away mid-pilot
The hardest CM decision is not selection. It is leaving a CM that is not working, mid-pilot, when leaving feels more expensive than staying. It usually is not. The cost of staying compounds. The cost of switching is bounded.
Walk when any of these are true:
- First-pass yield below your floor for three consecutive builds with no convergent root cause. If the CM cannot tell you why after three tries, they will not on the fourth.
- The line engineer who set up the program has left the account. The rebuild on their replacement costs more than switching CMs.
- Communication has degraded to weekly status emails. A CM that stops getting on calls has written off the program. They will still ship. They will stop fixing problems.
- They miss a delivery and do not offer recovery options. “We’ll catch up next month” rather than a recovery plan means they do not treat your ramp as a commitment.
- Unit cost creeps past agreed pricing without ECO documentation. Quiet drift rather than formal change orders means the commercial relationship has broken down.
Switching mid-program costs three to five months of re-qualification, re-tooling, and rebuilt line documentation. It is still almost always cheaper than finishing with a CM that is not going to get there.
At Rheo Dive — a consumer dive computer with over-molded silicone and a custom PCBA — CM selection took four months and rejected six vendors before we found the one that would engage with the optics integration and silicone tooling as a coherent program rather than two separate line items. That front-loaded work saved a year of downstream unwinding. The parent context sits in From Prototype to Production and How to Launch a Hardware Product.
This is what choosing a CM from the inside looks like. Not a capability deck exercise. An engineering conversation about whether this factory, this line, this process engineer, and this program fit each other — before the quote, before the MSA, before the first PO. The fit either exists or it does not. Finding out which is the work.
Selecting a contract manufacturer for a hardware program right now?
We sit on the program side of CM selection — writing RFQs, running DFM reviews with the line engineer, negotiating MSAs and quality agreements. Every engagement starts with a fixed-scope definition phase — no open-ended billing, no ambiguous timelines.
Start a Conversation