CNC Northern Ireland: How Buyers Judge a Shop

• 1
  Tolerance Is a Stack, Not a Number
• 2
  Where 5-Axis Helps and Where It Hurts
• 3
  Material Choice Changes the Process, Not Just the Price
• 4
  Volume Decides the Process, Not the Other Way Around
• 5
  Where Lead Time Actually Goes
• 6
  How to Verify a Shop Before the First Order
• 7
  Which Route Fits Your Part

The first question engineers ask a CNC Northern Ireland supplier is usually about tolerance. The honest answer is that tolerance is not one number. It is the sum of machine positioning error, thermal drift, tool deflection, workholding movement, and measurement uncertainty. A shop quoting ±0.005 mm on a part that will be clamped three times is quoting a capability, not a promise.

Positioning accuracy on a modern machining center sits well below the tolerance you need. The error that bites is elastic: a long thin wall pushed by a 12 mm end mill will deflect far more than the ballscrew error. That is why a shop with 16 simultaneous 5-axis machining centers can hold tight tolerances on a complex part yet still struggle on a flimsy one. Geometry matters more than spec sheets.

Thermal drift is the second hidden term. Aluminium 6061 grows about 23 µm per meter per °C. If the shop is at 20 °C in the morning and 27 °C by mid-afternoon, a 500 mm feature moves roughly 80 µm on its own. Good shops hold the finishing cell within a couple of degrees, or they finish in one continuous pass so the part and the machine drift together.

Measurement closes the loop. A ±0.005 mm callout measured with calipers is not verified. It needs a CMM or a micrometer in a temperature-controlled room, and the report needs to state the temperature. When we quote, we state which features can hold ±0.005 mm and which should be opened up. That is a more useful answer than a blanket claim.

The practical rule: keep tight tolerances for the features that carry function — bearing bores, mating faces, dowel holes — and release everything else to ±0.1 mm. Every tight callout adds cost, adds inspection time, and adds a place for the process to fail.

• 1
  Function-first calloutsTight only where the fit matters.
• 2
  Watch aspect ratioThin walls deflect; support them or relax them.
• 3
  Ask for temperatureA tolerance report without a temperature is incomplete.

Multi-axis work is often sold as universally better. It is not. A 5-axis machine wins when the part has features on several faces, when a single setup removes a stack of positional errors, or when an undercut cannot be reached by a straight tool. Those are real gains. A part with all features on one face gains almost nothing from the extra axes.

The gain is mostly in setup reduction. Every additional fixture adds a datum shift. On a part with four machined faces, going from four 3-axis setups to one 5-axis setup can remove three positional errors and several hours of labor. The travel envelope also matters: our large 5-axis centers carry 4,000 × 400 × 150 mm, so long extrusions and frame rails fit without re-fixturing.

Two conditions make 5-axis the wrong call. First, when the part is simple and the volume is high — a dedicated fixture on a 3-axis machine with a 27-machine pool will beat a 5-axis cell on cycle time and on cost per part. Second, when the part is small and rigid. A 500 × 500 × 450 mm envelope 3-axis machine holds a small bracket just as accurately for less money.

There is a tool-access tradeoff too. Simultaneous 5-axis motion needs a short, stiff tool. Long reach tools chatter, and chatter shows up as poor surface finish rather than a dimensional error, so it passes inspection and fails in service. If your part needs both deep pockets and a fine finish, expect the shop to split the operation.

Ask the shop which features drive the axis count. A clear answer names the specific undercut, the angled face, or the number of setups. A vague answer about advanced technology tells you the salesperson has not looked at your drawing.

• 1
  Good fitFeatures on 3+ faces, deep undercuts, one-setup parts.
• 2
  Poor fitFlat plates, high volume, small rigid parts.
• 3
  Ask whyThe shop should name the feature that needs the axes.

Material selection is usually treated as a cost line. In CNC work it changes feeds, speeds, tool wear, and sometimes the whole process route. Aluminium 6061-T6 machines fast and holds a good finish, which is why it dominates prototypes. Aluminium 7075 is stronger but gummier; it wants sharper tools and lighter chiploads to avoid built-up edge.

Stainless steels split into two families on the shop floor. Free-machining grades like 303 and 17-4PH in the annealed state cut cleanly. Austenitic 304 and 316 work-harden under a dull tool, so a light pass on a worn insert hardens the surface and the next pass skims over it. That is a process error, not a material defect, and it shows up as tool breakage mid-run.

Titanium TC4 (Ti-6Al-4V) is where thermal management decides the outcome. Its thermal conductivity is low, so heat stays at the cutting edge. Flood coolant, moderate surface speed, and a rigid setup are required. Inconel pushes this further; it is machinable but the cycle time and tool cost move by a large factor, so quote it as a separate line.

Plastics behave differently again. POM and PEEK hold tolerance well but move with temperature and moisture. ABS and PC are easy to cut and easy to scratch. Carbon fibre is abrasive; it wears tools quickly and needs dust extraction. If the part is a prototype that will be injection molded later, the material choice here should match the final resin family so the DFM feedback stays valid.

The practical check: tell the shop the function and the environment, not just the grade. A marine bracket and a medical instrument housing may both be 316L, but the fixture, the finish, and the inspection plan will not be the same.

• 1
  Aluminium6061, 2024, 5052, 7075, ADC12.
• 2
  Stainless303 and 17-4PH cut cleanly; 304 work-hardens.
• 3
  TitaniumTC4 needs coolant and rigidity; budget more time.

A common mistake is choosing a process and then fitting volume to it. Volume should choose the process. One to roughly 50 parts is CNC territory, where no tooling investment is needed and design changes are cheap. Above that, the comparison shifts toward die casting, vacuum casting, or dedicated fixturing, because setup cost gets divided across more parts.

CNC never becomes uneconomical — it just becomes outcompeted on unit price if the geometry allows another route. A machined bracket at 500 pieces is often still cheaper than a cast one once you add tooling, finishing, and the weeks of lead time for a mold. At 10,000 pieces the same comparison usually flips. The crossover point depends on geometry, not on a universal number.

Setup amortization is the mechanism. A 3-axis run on a simple part might need two fixtures and one program. That fixed cost is the same whether you make 5 parts or 5,000. So the unit price curve is steep at low volume and flat at high volume. If your quote jumps between quantity 1 and quantity 100, that is setup, not material.

Fixtures are worth questioning. For a run of 200 parts with a 40-minute cycle time, a soft jaw set and a simple plate are usually enough. For 2,000 parts, a dedicated hydraulic fixture pays back through cycle time and consistency. Ask the shop where their break-even sits on your part; it is a legitimate question and a useful sanity check on the quote.

There is no minimum order quantity in our model, so a single prototype and a 10,000-part run use the same quote path. What changes is the fixture plan, the inspection sampling, and whether we run lights-out overnight.

• 1
  1–50 partsCNC, no tooling, easy revisions.
• 2
  50–1,000Compare CNC against casting with tooling cost included.
• 3
  1,000+Dedicated fixtures and casting routes start to win.

Lead time is usually quoted as one number, but it has four parts: quoting, DFM feedback, machining, and finishing plus inspection. Only one of them is cutting metal. If a shop quotes ten days and cannot say how those days split, the number is a guess.

Quoting and DFM are the fastest to compress. We return a quotation and a free DFM analysis within 12 hours, and production can start within 24 hours of approval. That is not the norm for every shop, but it shows the mechanism: fast quoting comes from having the material on the shelf and the programmer available, not from rushing the machine.

Machining time is the part most sensitive to design. A 40-minute cycle on one part becomes 4,000 minutes across a 100-piece run. Every unnecessary tight callout, every deep pocket, and every difficult material adds directly to that number. Reducing cycle time at the drawing stage is the cheapest lead time you will ever buy.

Finishing and inspection are the quiet bottleneck. Anodizing, plating, and powder coating are outside processes with their own queues. Final inspection on a complex part can take longer than the last machining pass. If your schedule is tight, ask which finish steps are external and add their queue time to the plan.

Shipping is the last variable, and it is the one buyers fixate on. For a CNC Northern Ireland buyer, an air shipment from a partner in Dongguan is often measured in days, while a domestic ground shipment inside the island is measured in hours. The realistic comparison is total landed time from drawing release, not transit alone.

• 1
  Quote and DFM12 hours on our side, with material in stock.
• 2
  MachiningScales linearly with volume and cycle time.
• 3
  FinishingExternal processes add their own queue.

Certificates are a starting filter, not proof. ISO 9001:2015, IATF 16949:2016, ISO 13485:2016, and ISO 27001:2022 each cover a different risk. ISO 9001 is general quality management. IATF 16949 is automotive and adds traceability and change control. ISO 13485 is medical and adds validation and documentation discipline. ISO 27001 covers how your drawings and data are protected. Match the certificate to your industry.

Then look at the inspection plan. A shop that inspects 100% of parts before shipment will produce a report if you ask. Ask what instrument measures your tightest feature and what the room temperature is. A shop that answers quickly has done this before. A shop that promises a report without naming an instrument is likely to send a generic sheet.

Ask for the machine list. Not the total, the breakdown. Sixteen simultaneous 5-axis centers and 27 three-axis machines tell you the shop can route a part to the right envelope rather than forcing it onto one machine. It also tells you whether a 4,000 mm long part is routine or a special arrangement.

Confidentiality deserves a direct question. If your part is a new product, an NDA should be available on request and uploads should be handled as confidential by default. Ask where the files live and who can open them. A shop holding ISO 27001 should be able to describe the access controls without hesitation.

Finally, start with a sample. Our sample center exists so a new buyer can see surface finish, edge quality, and packaging before committing to a production run. A physical part answers more questions than a capability page.

• 1
  Match the certificateAutomotive, medical, and data risk are different.
• 2
  Name the instrumentCMM, micrometer, or profile projector.
• 3
  Start with a sampleCheck finish and packaging first.