Switching CNC machine shops mid-program works when you treat it as a controlled transfer, not a hard cutover. Freeze the current revision and inspection plan. Pull a full quality package from the incumbent: material certs, FAIR (First Article Inspection Report) or AS9102 (the aerospace first article inspection standard), routers, and special process certs. Dual-run one or two high-risk components at the new shop before you cancel the old PO. Complete first article at the new source before volume moves. Overlap at least one lot so you are never single-threaded. The shops that fail this are the ones that quote fast, ask few questions, and cannot show a sample quality package.
Quick facts
- Do not move the whole book of business on day one
- A new process source usually needs a new first article
- Ask for 12-month on-time delivery, not a slogan
- Confirm what is in-house versus sent out: heat treat, anodize, passivation, NDT
- A named project manager and a set update cadence matter as much as the machine list
- ITAR, CMMC, AS9100, and ISO 13485 only count if the certificate scope matches the work
Why shops fail after first article
First article approval feels like the finish line. It is not. It proves a shop can make one good lot while everyone is watching. It does not prove the same process holds when the release quantity goes up and second shift is running it.
That is where programs break.
The shop has no spare spindles. There is no pallet pool or 5-axis plan for the next volume step. Travelers stop at “send out.” CMM reports come back with no balloon numbers. On-time delivery lives in someone’s head. Nobody owns the schedule except the person you already email twice a week.
The cost does not show up as a line item called “bad supplier.” It shows up as a down line, expedite freight, a missed customer ship date, or findings in your next audit. A lot of work arrives here the same way: the last shop got through first article, then ran out of hours.
If you are still building the production process with a new source, that is a different job than pulling a live program off an incumbent. Start with the path from first article to repeat orders: Medical Device Production Machining: How to Scale from First Article to CNC Machined Repeat Orders.
When switching is cheaper than staying
You do not change shops because a quote came in 8 percent lower. You change shops when staying is the bigger risk.
| Signal | Why it matters |
|---|---|
| On-time delivery under about 90 percent for two quarters | That is a pattern, not a bad month |
| No written plan if volume doubles in 90 days | You will eat the ramp, not them |
| Quality package is verbal or incomplete | The next audit becomes your problem |
| No named project manager, and updates only when you chase | Schedule risk stays invisible until the part is late |
| “We can get certified if you award the work” | The process is not in place now |
| Critical operations are brokered with no visibility | You bought a middleman |
One missed lot can be a mill or a coating queue. Two quarters of missed dates and no capacity plan is when you start the transfer, not after the next miss.
The mid-program transfer playbook
Give the move a gate at each step. Skip one and you will feel it on the dock.
1. Freeze the revision and the inspection plan
Lock the drawing revision, the CTQs, and the inspection method before either shop cuts chips on the transfer lots.
If an ECN is coming, decide who owns the cut-in. Do not let the incumbent run Rev C while the new source quotes Rev D and incoming inspection still uses last year’s ballooned print. That mismatch creates scrap that looks like a new-shop problem when it is a configuration problem.
2. Pull the incumbent quality package
Ask for the file that should already exist:
- Material certifications and heat or lot traceability
- FAIR or AS9102, or the medical equivalent your quality agreement requires
- In-process and final CMM reports
- Router or traveler
- Special process certifications and the approved sources behind them
- Nonconformance history on that part family
If they cannot produce this, you already know why you are leaving. You also know you will be rebuilding process knowledge from the print and the part, not from a clean handoff. Watch for certs that show up as photos of photos, and for a “FAIR” that is just a spreadsheet with no ballooned print.
3. Move one or two high-risk components, not the whole book
Do not send the new shop your entire book. Send the component that is already late, or the one that will sink you if it goes late. Leave the boring, stable parts where they are until the new source has a bought-off process.
Dual source means two live processes for a period. It does not mean an announcement that everything is moving next month. The second source has to earn the rest of the book.
4. Run first article at the new shop before you cut the old PO
A new source is a new process. In most aerospace and medical programs, that means a new first article, full or partial. A prototype lot or “we made three that look good” is not production approval.
Define “first article complete” in writing:
- Ballooned print against the frozen revision
- Material and special process evidence
- Dimensional report from the inspection method you will use in production
- Any capability data your quality agreement actually requires
Keep the incumbent PO alive until that package is accepted.
5. Overlap lots so you are never single-threaded
Plan one or two overlapping releases. Write down who ships which lot and when. Buy material so you are not waiting on the same mill lead time twice.
The overlap costs extra for a short window. A single-source gap during a transfer costs more. This is also the moment the incumbent can go quiet. Pay for those overlap lots and be direct about the timeline. Do not announce a full exit before the new source has a bought-off first article.
6. Map outside processes before chips hit the floor
Heat treat, anodize, passivation, coatings, NDT, and laser mark are where transfers stall. Anodize queues and mill-cert titanium stretch timelines more often than the machining itself.
Name the spec, the approved vendor list, and the cert flow for each outside process. Decide whether the new shop uses your AVL or theirs. If NADCAP or a prime-approved processor is required, confirm that before the first article clock starts.
If finishing is “we have a guy,” get the processor name, the spec, and a sample cert before you award anything. People skip this because machining felt like the hard part.
7. Set the communication rules in writing
You want a named project manager, a weekly update even when the news is bad, an escalation path if a date slips by an agreed number of days, and a clear way engineering questions get answered.
If the new shop only talks when you call, you traded one blind spot for another.
Questions to send in writing before you move metal
Put these in the RFQ. You want the answer in writing. A good call is easy. A complete package is not.
- What was your on-time delivery for the last 12 months on similar work? Not a target. Actual.
- What is current utilization, and what happens if this volume doubles in 90 days?
- Can you send a sample FAIR or quality data package on a comparable component?
- Which machines would actually run this job?
- What release quantity versus annual volume are you quoting, and have you run that cadence before?
- What is done in-house versus sent out, by process?
- What certifications do you hold, and what is the scope on the certificate? AS9100, ISO 13485, ITAR, and CMMC only count if they cover this work.
- Walk through a nonconformance from detection to CAPA.
- Who is the project manager, and what is the update cadence?
- How do you handle a mid-program print change?
- Can you share two references in the same industry and volume class?
A fast quote with no questions is a warning, not a win. Shops that will run regulated work ask about CTQs, finishes, inspection level, annual volume, release quantity, and whether the print is ITAR-controlled.
If you need a deeper filter for flight hardware, use the certification and traceability checks in our aerospace CNC machining guide. For device work, start with medical CNC machining services. Defense programs should confirm data handling, not just a logo. That is the point of CMMC readiness.
Red flags that should end the conversation
- Instant online price and no questions about the drawing
- Quote ignores inspection level, finish spec, or annual versus release quantity
- They want every part number before they have finished one
- No CMM, and “we inspect everything” with hand tools only
- Sample “FAIR” is a spreadsheet with no ballooned print
- Certification logos with no certificate, or a scope that does not cover this work
- Will not discuss utilization
- Offshore material with no mill pedigree on regulated work
- “We will figure out finishing after we win the PO”
Keep looking. The transfer is hard enough with a shop that already runs this class of work.
What “the new shop can scale” has to mean
Scalable does not mean they said yes. It means they can add capacity without inventing a new quality system at the same time.
Look for capacity they can add, not a hope to buy a machine someday. Look for a process that survives second shift. The setup that worked when one programmer babysat first article will not hold when the job is tribal knowledge. And look for proof, not promises: material tied to lot, CMM data that matches the print, certs that travel with the shipment. You should be able to pick a finished component and walk it back to incoming material.
We get this work when the last shop cleared first article and then ran out of hours. What we can commit to is a named project manager, a weekly update that still goes out when the news is bad, and a quality package you can hand to an auditor. AS9100D, ISO 13485, CMMC, and ITAR only matter if the scope on the certificate matches your program. If a shop cannot tell you how inspection and finishing keep up when volume moves, they are not a second source. They are another first article.
How long a transfer usually takes
Ranges, not fake precision. Simple aluminum brackets sit on the short end. Titanium, multi-process aerospace, or medical components with special processes sit on the long end. What usually slips is not the cycle time. It is the anodize queue, the mill cert, or the ballooned print the incumbent swore they had.
| Phase | Typical window | What stretches it |
|---|---|---|
| Quote and gap review | A few days to 2 weeks | Missing 2D, unclear CTQs, ITAR questions |
| Material and tooling | 2 to 8 weeks | Mill-cert alloys, custom workholding |
| First article and buyoff | 2 to 6 weeks | Special processes, ballooned print gaps |
| Overlap / dual run | 1 to 3 releases | Schedule fights with the incumbent |
Build the overlap into the calendar. If first article at the new source lands the same week the incumbent’s last lot was supposed to ship, you do not have a transfer plan. You have a gap.
FAQ
Do I need a new first article if I change machine shops?
Yes in most aerospace and medical programs. A new source is a new process. AS9102 and OEM quality agreements treat that as a full or partial FAI (first article inspection), not a paperwork shuffle.
Can I dual source ITAR-controlled parts?
Yes if both shops are registered and handle technical data correctly. Confirm registration, data control, and visitor rules before you send a print.
Should I move prototypes and production together?
Usually no. Move the production-intent process. Prototype habits (tribal setups, no traveler) do not survive a volume ramp.
How many parts should I transfer first?
One part family or two components that are actually hurting you. Prove the process, then expand.
What should be in a quality data package?
Material certs tied to lot, FAIR or equivalent, dimensional reports, process traveler, special process certs, and a way to trace a finished component back to incoming material.
Is a cheaper quote a reason to switch?
Not by itself. A lower piece price that misses two deliveries costs more than the incumbent.
Can a job shop that did our prototype take production?
Only if they can show capacity, inspection, and a documented process. Prototype skill and production discipline are different jobs.
How do I keep the current shop from walking out mid-transfer?
Be direct. Pay for the overlap lots. Do not announce a full exit before the new source has a bought-off first article.