Why process selection is the most expensive decision you'll make

Most parts can be made several ways. Most parts shouldn't be. The wrong method shows up as one of three problems: a unit cost that kills the business case at scale, a tolerance that drifts on the second batch, or a six-week lead time that wasn't in the project plan.

The fix isn't to learn every process in detail. It's to make five decisions in order. Get those right and the method picks itself.

The five decisions that pick the process

In order — earlier decisions constrain later ones:

  1. Volume. How many parts, over what period?
  2. Geometry. What does the part actually look like — prismatic, organic, sheet, hollow?
  3. Tolerance & surface. What's the tightest spec on the drawing?
  4. Material. What grade, and is it process-compatible?
  5. Lead time. When does the first part need to be in someone's hand?

Most engineers start at material or geometry. That's backwards. Volume determines whether tooling cost is amortised; tooling cost is the single biggest variable in unit cost. Start there.

Decision 1: Volume

Volume sets the broad family of processes that are economically viable. It's the only decision that's binary at the extremes:

Volume per yearLikely viable methodsDon't bother with1–10Additive (metal or polymer), CNC machiningInjection moulding, casting (unless one-off investment cast)10–500CNC, additive, sheet metal, urethane castingHard tooling injection moulding500–5,000CNC, sheet metal, investment casting, low-volume injection (aluminium tooling)Hand-finished additive5,000–50,000Injection moulding, die casting, sheet metal stamping, CNC for high-value partsPowder bed additive (cost per part)50,000+Injection moulding, die casting, stamping, forgingAnything without amortised tooling The crossover point that matters most: injection moulding tooling typically pays back somewhere between 1,000 and 10,000 parts depending on cavity count and material. Below that, low-volume processes win on landed cost even if unit price looks higher.

Decision 2: Geometry

Geometry rules out methods volume said were viable.

  • Prismatic with flat faces, holes, pockets: CNC milling. Fast, cheap, accurate.
  • Round, rotational, threads: CNC turning, often combined with milling on a mill-turn.
  • Sheet-derived (bent, welded, formed from flat stock): sheet metal fabrication.
  • Internal channels, undercuts, conformal cooling, lattice: additive — usually metal LPBF for engineering parts, MJF or SLS for polymer.
  • Thin-wall hollow with cosmetic surface: injection moulding (volume permitting), vacuum forming, or rotational moulding.
  • Large monolithic structural: casting (investment for accuracy, sand for size), or DED additive for repair / near-net.

The hard test: if you removed all design freedom and had to make this part with the cheapest tool you own, what would you reach for? That's usually the right answer.

Decision 3: Tolerance and surface finish

Tolerances eliminate methods. They rarely add new ones.

Spec on drawingMethods that deliver as-builtMethods that need post-processing±0.5 mm, Ra 6.3 or worseMost processes—±0.1 mm, Ra 1.6CNC, MJF, SLS, investment castingLPBF (machining critical surfaces)±0.025 mm, Ra 0.8CNC, ground / lapped surfacesAdditive (fully machined), die casting (machined)Ra 0.4 or betterGrinding, polishing, EDMAnything else needs secondary finishing The mistake to avoid: applying the tightest tolerance on the drawing to every feature "just in case." Each tightened tolerance multiplies inspection and rework cost. Tolerances should be loose by default and tight only where the assembly demands it.

Decision 4: Material

Material limits the process pool further:

  • Aluminium 6061 / 7075: CNC, sheet, casting, extrusion. Avoid LPBF unless you specifically need AlSi10Mg properties.
  • Titanium Ti-6Al-4V: CNC (expensive in chip), LPBF (efficient, weight-critical parts), forging (high-volume aerospace).
  • Stainless 316L / 17-4PH: CNC, LPBF, MIM, investment casting.
  • Inconel 625 / 718: CNC (slow, tool wear), LPBF (often the winner), DED for repair / large.
  • Tool steel: CNC + heat treat, EDM for cavities, LPBF for conformal cooling.
  • Engineering polymers (PA12, PA6, PEEK): SLS / MJF for low-volume, injection moulding for high.
  • Elastomers / overmoulds: injection moulding, urethane casting for low-volume.

If your material isn't process-compatible, you have two options: pick a different material that meets the same engineering requirements, or pick a different process. The wrong move is to force-fit the part to a process that the material can't reliably support.

Decision 5: Lead time

Lead time often forces a compromise on the other four:

  • Need it in 24–72 hours: additive (polymer or metal, depending on size), or expedited CNC for simple parts.
  • 1–2 weeks: CNC, sheet metal, MJF, SLS, urethane casting.
  • 3–6 weeks: investment casting, low-volume injection moulding (aluminium tooling), complex multi-axis CNC.
  • 8–16 weeks: hard tooling injection moulding, die casting, forging, large investment castings.

What gets missed: finishing, inspection, and shipping rarely show up in the quoted lead time but routinely add 1–3 weeks. If the part needs heat treat, anodising, or first-article inspection, ask the supplier to include those in the quoted lead time before you commit to a delivery date.

If-X-then-Y: the decision triggers

Once you've made the five decisions, most parts fall into one of these patterns:

  • If volume < 10 and geometry has internal channels → metal additive (LPBF). The tooling for any other method doesn't pay back.
  • If volume < 100 and the part is prismatic with tight tolerances → CNC machining. Faster, cheaper, and more accurate than any alternative.
  • If volume 100–1,000 and the part is sheet-derived → laser cut + bend (sheet metal fab). Tooling-free, fast turnaround.
  • If volume > 5,000 and polymer → injection moulding. Tooling pays back, unit cost collapses.
  • If volume > 5,000 and metal with complex internal features → investment casting + machining where needed.
  • If part is impossible to machine due to internal geometry → additive is usually the only viable route. Validate with a process simulation before committing.
  • If lead time is the binding constraint → additive for prototypes, expedited CNC for short runs, parallel-source production work across multiple suppliers.

The trade-offs no one tells you

Most process comparisons stop at unit cost. The real trade-offs are subtler:

  • CNC machining looks expensive per part but absorbs design changes for free. Additive locks you in early because every change requires a re-print, often a re-orient, sometimes a re-support.
  • Additive looks fast for prototypes but slow for production. A typical LPBF build cycle including post-processing is 3–7 days. CNC can deliver the same part in 24 hours if the machine is free.
  • Injection moulding hides cost in tooling lead time. The first part is 8–16 weeks away. After that, parts come at seconds each.
  • Casting trades unit cost for tolerance. Investment casting is dimensionally good but rarely meets ±0.05 mm without machining. Sand casting needs more.
  • Sheet metal is the most underrated process for low-to-mid volume mechanical assemblies. If your part can be expressed as bent, cut, and welded sheet, almost nothing beats it on cost-to-quality.

Common mistakes that cost real money

  1. Designing for prototype, then re-designing for production. Decide the production process first. Prototype with a process that mirrors production constraints — or accept that the second design will be different.
  2. Assuming 3D printing is always faster. For a single bracket with 3 holes, a CNC quote is often 24 hours. An additive build slot might be 4 days out.
  3. Choosing a process to suit a supplier's preference. Suppliers will quote what they make. The right process might be at a different supplier.
  4. Ignoring tolerance stack-up. A part assembled from five components with ±0.1 mm each can drift 0.5 mm overall. Sometimes one tighter component is cheaper than tightening all five.
  5. Forgetting finishing. A raw LPBF surface is Ra 6–25. A "finished aerospace part" might need machining, blasting, polishing, anodising, coating, and inspection. None of that is in the print quote.
  6. Treating volume as fixed. A "low-volume" run that scales 10× breaks the original process choice. Ask: at what point would I switch processes? Build that crossover into the design.

How to actually make the call

Walk the five decisions in order. After each one, ask: which methods are still viable? By the end you'll typically have one or two left.

If you're left with two — say CNC and additive — the deciding factor is usually which one absorbs the next change. Production parts that won't change again? Pick the one with the lowest unit cost at your volume. Parts likely to evolve? Pick the one with the lowest cost-per-iteration.

If you're left with zero — the constraints are over-tight. Loosen tolerance, change material, split the part, or change the volume assumption. Something has to give.