Small Scale Production Before Prototype: What a Pilot Run Really Tells You

• 1
  What small scale production before prototype actually does
• 2
  Why CNC machining suits the pilot stage
• 3
  What a pilot batch can and cannot prove
• 4
  How to plan the batch without wasting it
• 5
  Cost and schedule logic before tooling
• 6
  Which route fits which pilot goal

A pilot batch is not a smaller version of mass production. It is an information-gathering step. You cut 5 to 50 parts from the intended material so that the design stops arguing with physics. Screens and renders hide the things that only appear on a machine: where a tool cannot reach, where a wall is too thin to hold a thread, where a fixture has nothing to grip.

The word prototype covers a wide range. A CNC prototype can be a one-off shape check or a fully functional part, and those two jobs need different drawings. Small scale production before prototype means you commit to a batch size, a material and a process route early, then watch what breaks. The output is not parts. The output is a list of changes.

This step sits between the concept model and the tooling decision. It is the last cheap moment. Once you cut a die or a mould, a wall thickness change costs weeks. In a pilot batch it costs an edit and a re-cut.

One caution. A pilot batch is only honest if the parts are made the way the production parts will be made. Machining a pilot batch and then switching to die casting hides every casting problem you were trying to find. Match the process, or accept that you are testing the design, not the route.

• 1
  Batch size5–50 parts is usually enough to see repeatability.
• 2
  MaterialUse the intended alloy, not a convenient substitute.
• 3
  ProcessMatch the planned production route where it matters.
• 4
  OutputA change list, not a shipment.

CNC machining needs no tooling. A 5-axis machine can cut an undercut, an angled port or a contoured pocket in one setup, and the same program runs again next month without a new mould. That is the core reason it fits the pilot stage: the cost curve is flat for small quantities and the geometry ceiling is high.

Tolerance matters here too. Holding ±0.005 mm on a bearing bore lets you test the real fit instead of guessing at it. A pilot part that is 0.1 mm off tells you nothing useful about a production fit, because the error is bigger than the feature you are studying.

Surface finish is a second signal. Ra 0.8–1.6 μm is a normal machined finish for a sealing face. If your design only works at Ra 0.2–0.8 μm, the pilot run is the moment to find out what that costs in time and handling.

The limits are real. Machining is slower per part than a moulding process, and unit cost does not fall much as quantity rises. Treat the pilot batch as a test rig. Do not use it to set the production price.

• 1
  Good fitComplex geometry, tight tolerance, low quantity.
• 2
  Poor fitSimple part where the mould already exists.
• 3
  WatchUnit cost will not drop with volume.

It can prove fit, function and assembly order. It can prove that a connector seats, that a gasket seals, that a bracket clears the housing by 2 mm instead of 0.2 mm. It can prove that your inspection method catches a bad part. These are the failures that hurt most later, and they show up fast on real parts.

It cannot prove tool life, cycle time on a moulding machine, or the yield of a process you have not run yet. A machined pilot batch says nothing about die casting porosity, weld distortion in a stamped frame, or the shrink rate of an injection-moulded cover. Different physics.

It also cannot prove demand. Ten parts in the hands of ten users is a usability signal, not a market signal. Do not read volume data into a sample that small.

What it can do is set a baseline. Once you have measured parts, you have a reference for every later change. Without that reference, version two is just another opinion.

• 1
  ProvesFit, function, assembly sequence, inspection.
• 2
  Does not proveTool life, moulding yield, demand size.
• 3
  OutputA measured baseline for later revisions.

Decide the question first. Write one sentence: what must this batch prove? If the answer is everything, the batch will prove nothing, because you will not know which result to act on. One or two questions per batch is a working number.

Freeze a revision number on the drawing before cutting starts. Every part gets stamped or bagged with that revision. When a fit fails, you need to know whether the part or the drawing moved.

Plan the inspection at the same time as the machining. Decide which features are critical, how they will be measured, and what value counts as a failure. A pilot batch without a pass or fail line is just a pile of metal.

Keep the setup notes. Workholding, tool list, program revision and any hand work belong in the same folder as the drawings. When you move to production, that folder is the starting point for the process plan, and it saves a repeat of every decision you already made.

• 1
  One questionPer batch, not per project.
• 2
  Revision controlStamp the drawing revision on every part.
• 3
  Inspection planDefine pass and fail before cutting.
• 4
  Setup notesThey become the production process plan.

The pilot batch is where you spend a little to avoid spending a lot. A machined batch of 20 parts costs more per unit than a production run will, and that is fine. You are buying information, and the information is cheaper now than after tooling.

Schedule risk drops as well. A change in a machined design is an edit and a re-cut. The same change after a die exists means a tool modification, a new sample run and a fresh approval cycle. The earlier the change, the smaller the bill.

There is a second-order saving that people miss. A pilot batch gives you real parts to build a fixture around, a test rig around and a packaging decision around. Those costs are hidden until they arrive.

Set a unit cost expectation for the pilot that is separate from the production target. Mixing the two numbers leads to arguing about the wrong thing at the wrong time. Keep them apart until the design is stable.

• 1
  Buy informationPilot unit cost is not the production price.
• 2
  Change costEdits are cheap now, expensive after tooling.
• 3
  Hidden savingsFixtures, test rigs and packaging get real data.