CNC Shop Startup Guide: What Actually Sets Your Capability

• 1
  Why a CNC Shop Startup Guide Should Start With Capability
• 2
  Axis Count, Travel and the Jobs Each One Wins
• 3
  What ±0.005 mm Really Demands From a Small Shop
• 4
  Build the Inspection Loop Before the Second Machine
• 5
  Material Choice Changes Your Tooling and Coolant Plan
• 6
  When to Outsource Instead of Buying
• 7
  Cost Structure and the Numbers That Decide Survival
• 8
  Scaling Without Breaking the Process You Already Have
• 9
  Matching Shop Profile to Equipment and Quoting Range

A new shop is not defined by the machines it owns. It is defined by the envelope of parts it can make on time, at the tolerance the drawing asks for, with inspection evidence behind it. That envelope is what customers buy. Everything else, floor space, branding, financing, exists to widen or protect it.

So before you compare spindle speeds, write down the jobs you intend to win. A shop targeting brackets and fixture plates lives in a different world from one targeting impellers or bone plates. The first needs throughput and cheap fixturing. The second needs simultaneous 5-axis work, thermal control and a metrology room.

The gap between those two shops is not just capital. It is the number of process variables someone on your team can hold in their head at once. A startup with two machinists and no inspection plan can run simple 3-axis work at Ra 1.6–3.2 μm all day. Ask the same shop to hold ±0.005 mm across a batch and the failure mode is not the machine, it is the feedback loop.

This is why the useful question is never how much a machine costs. It is what the machine lets you promise. A capability map, drawn before purchase, prevents the most common startup mistake: buying capacity you cannot feed and skipping the equipment that protects the work you already have.

• 1
  Define the part family firstGeometry, material and tolerance band, not a machine brand.
• 2
  Count your feedback loopsWho measures, how often, and what happens when a reading drifts.
• 3
  Write the promise, then buyEvery purchase should widen a specific quoted capability.

Three-axis milling handles prismatic parts with features reachable from one direction: plates, housings, brackets, heat sinks. Setup is simple, fixturing is cheap, and a competent operator can hold ±0.01 mm without heroics. If your target market is industrial machinery and electronics enclosures, a 3-axis base is the correct base.

The fourth axis adds rotation around X or Y, so you machine four faces in one setup. That single change removes most of the positional error that comes from re-fixturing, and it cuts labor per part. Four-axis work is where a lot of automotive and electronics volume actually sits. It is unglamorous and profitable.

Simultaneous 5-axis is a different category. The tool stays normal to a curved surface while two rotary axes move together. That is how you cut impellers, turbine blades, complex medical geometry and deep cavities with short tools. It is also where tool path verification, post-processor quality and operator skill stop being optional.

Travel decides the ceiling. A compact machine with a 500 × 500 × 450 mm envelope cannot quote large weldment machining no matter how accurate it is. If large parts are in your plan, the machine that cuts them is a separate investment from the one that cuts small, high-mix work. Mixing those two goals on one spindle is how shops end up mediocre at both.

• 1
  3-axisPrismatic parts, one dominant direction, lowest setup cost per feature.
• 2
  4-axisMulti-face parts, fewer re-fixtures, strong fit for volume brackets.
• 3
  Simultaneous 5-axisCurved and contoured geometry, short tools, high verification cost.

A tolerance printed on a drawing is a promise about the whole process, not a spec on the machine brochure. Holding ±0.005 mm (±0.0002 in) means the machine, the fixture, the tool, the coolant strategy and the room temperature all stay inside a budget that adds up to less than the tolerance itself.

Start with thermal behavior. A spindle that has run for two hours is not the same spindle that started cold. Aluminum expands roughly 23 μm per meter per degree Celsius, so a 3 °C shop swing across a 300 mm part moves the material about 20 μm. That is four times the tolerance before a single cut is made.

Then look at the fixture. Any part held with uneven clamping force deforms, and it springs back after unclamping. Thin-walled parts are the classic case. A better fixture often buys more accuracy than a better machine.

Finally, consider how you will know. A tolerance you cannot measure is a tolerance you cannot hold repeatably. Optical comparators and micrometers cover simple features. Position, form and runout on complex geometry need a CMM with a stated uncertainty well below the tolerance band. If your measurement uncertainty is half the tolerance, you are guessing.

• 1
  Thermal budgetControl room temperature and let spindles warm up before critical cuts.
• 2
  Fixture stiffnessEven clamping, minimal overhang, support under cutting load.
• 3
  Measurement uncertaintyKeep it at roughly one fifth of the tolerance you promise.

The first equipment purchase most new shops get wrong is the second machine. A second spindle doubles output only if the first one is already producing conforming parts at a known rate. Without inspection data, a second machine doubles the volume of parts you may have to scrap.

A practical sequence: hand tools and a height gauge for setup, a micrometer set for external dimensions, pin gauges for holes, and access to a CMM for first-article and periodic verification. That combination covers most 3-axis work. The cost of a CMM is far below the cost of a returned batch.

Sampling strategy matters more than instrument count. If a feature is produced by the same tool, same program and same setup, its variation is largely systematic. If it depends on a re-fixture, a tool change or an operator decision, treat each occurrence as a new event and measure it.

Document the readings. A short inspection report per lot turns an argument about a rejected part into a five-minute conversation. It also protects you when a customer changes a drawing mid-run and expects the old parts to fit the new design.

• 1
  First articleMeasure every dimension on the drawing before the run continues.
• 2
  In-processCheck the features most likely to drift: bores, thin walls, tight positions.
• 3
  FinalConfirm the drawing, not the setup sheet, before parts ship.

Aluminum is the forgiving start. Grades like 6061 and 7075 cut fast, hold good finishes and tolerate minor coolant mistakes. A shop that begins with aluminum brackets and housings learns programming and fixturing without fighting the material at the same time.

Stainless 304 and 316 work-harden. If the tool rubs instead of cutting, the surface gets harder and the next pass is worse. The fix is aggressive feed per tooth, rigid setups and no dwell. Grade 17-4PH adds heat treatment into the sequence, which changes when you machine and when you measure.

Titanium TC4 (Ti-6Al-4V) has low thermal conductivity, so heat goes into the tool instead of the chip. Tool life drops and the risk of a rubbing fire rises. High-pressure coolant and conservative radial engagement are not optional. Inconel pushes the same problem further, and it is usually not a material for a shop still learning its own spindle.

Plastics bring a different failure mode. POM and PEEK cut cleanly but move with temperature, and ABS and PC can melt or chip depending on sharpness. Vacuum fixturing and light finishing passes matter more than cutting data. If your target market is medical or electronics, plastic work will arrive whether you planned for it or not.

• 1
  Start with aluminumLearn fixturing and programming on a forgiving material.
• 2
  Respect work hardeningStainless rewards feed per tooth and rigidity, not slow rubbing passes.
• 3
  Plan heat treatment into routingMachining before and after hardening are different operations.

A startup cannot own every process, and it should not try. Finishing is the clearest example. Anodizing, plating and powder coating require chemical lines, waste treatment and compliance work that have nothing to do with machining. Sending parts to a qualified finisher keeps your capital in spindles.

5-axis work is the second example. If fewer than one in ten of your quoted jobs needs simultaneous 5-axis motion, buying the machine means paying for idle capacity and for the programming skill to feed it. Partnering on those jobs lets you bid them now and buy later, when the volume justifies it.

Heat treatment, EDM for sharp internal corners, and large-part machining are similar. Each has a utilization threshold. Below it, outsourcing is cheaper. Above it, outsourcing becomes a queue you do not control.

The trap is outsourcing the process that defines your quality. If your customers buy you because of surface finish, keep finishing close and control it. If they buy you because of geometry, keep the geometry in-house and let the chemistry travel.

• 1
  Outsource low-utilization processesFinishing, heat treatment, EDM, oversized machining.
• 2
  Keep what defines youDo not hand your differentiating step to a supplier you cannot audit.
• 3
  Qualify partners earlyA supplier who has never run your part is not a schedule.

Machine time is the number most new owners track, and it is rarely the one that kills them. Setup, programming, fixturing and inspection are where the hours go on low-volume work. A two-hour cut with six hours of preparation has a different cost per part than the same cut in a repeat run.

Quote accordingly. If you price every job by cycle time, you will win high-volume simple work and lose money on the complex prototypes that build your reputation. Track setup hours separately for the first months. The pattern shows up fast.

Financing changes the risk profile. A machine payment is fixed; work is not. Three months of payments in reserve is a reasonable floor before adding a spindle, because the ramp between a signed purchase order and steady repeat business is longer than most plans assume.

Do not compete on price for work you cannot repeat. The jobs that pay for a new shop are the ones where the customer's alternative is a long lead time or a supplier who already failed. Speed and reliability carry a premium that hourly rates do not.

• 1
  Separate setup from cycleTrack them independently and price both.
• 2
  Reserve paymentsKeep enough cash for several months of fixed cost.
• 3
  Sell reliabilityRepeat business comes from parts that fit, not from the lowest rate.

Growth fails in a predictable order. First the quote backlog grows and delivery slips. Then new operators are hired faster than the process is documented, and quality becomes operator-dependent. Then the shop buys a machine to fix throughput and the problem moves to programming.

Document before you hire. A setup sheet with tool numbers, offsets, clamping notes and inspection points lets a new machinist produce the same part on the second day instead of the second month. The document is the product of your process knowledge, and it is what makes the second shift possible.

Add capacity in the narrowest step. If parts wait for the CMM, another mill will not help. If they wait for a deburring bench, another CMM will not help either. Map where parts sit idle and buy for that station.

Keep a partner for overflow. Sending a rush lot to a qualified shop protects a delivery date you would otherwise miss, and it costs less than the customer you would lose. Overflow capacity is insurance, not failure.

• 1
  Document the setupTool list, offsets, clamping, inspection points, revision level.
• 2
  Widen the bottleneckAdd capacity where parts actually queue.
• 3
  Keep overflow partnersA qualified backup protects the delivery promise.