What Is Design for Manufacturing?

A part can be manufacturable on a five-axis mill and still be a poor candidate for the process you can afford to run in volume. Without a process, a material family, a tool or fixture class, and a quantity, the word "manufacturable" has nothing to be true against. The figures below are hypothetical teaching maths, not BrahmWorks prices, yields, timelines or test results. Fitting that match into a release, rather than into a prototype that happened to get made, is the subject of How to Take a Hardware Prototype to Production.

Start with the sentence "this part will be made by X, in material family Y, at about quantity Z". X might be injection moulding, sheet metal with a brake, die casting, machining from plate, or a printed process you truly intend to ship. Y is a family, not a colour. Z is the quantity you will buy, not the quantity in a pitch.

Each process has shapes it repeats cheaply and shapes it punishes. Moulding repeats a drafted shell and punishes a solid lump and an undercut you did not price. Brake-formed sheet repeats flanges and holes in one flat pattern, and punishes a depth that needs a drawn corner or a weld on a cosmetic face. Machining repeats precise faces from a datum and punishes deep thin pockets and features that need a second setup for a tiny gain. If the sentence naming the process is still blank, you are not doing DFM. You are decorating a CAD model.

Secondary operations are part of the process. Paint, a tapped hole, a heat stake, a weld, a pressed insert: each is a handling and a defect. If the geometry only works after an operation missing from the routing, either add it to the routing and the cost, or change the geometry so the primary process is enough.

A title-block tolerance is a claim that every untoleranced feature will be held that tight, on that process, at the quantity. Teams copy ±0.1 mm from a machining habit onto a moulded enclosure and think they have been rigorous. They have written a claim the moulder may price as inspection, as a steel-safe gamble, or as a polite ignore. None of those is a design.

Put tight numbers only on features that perform a function: a bearing bore, a seal land, a datum that locates a board. Leave the rest at what the process repeats without a special effort. Ask the person who will make the part what they will hold on the features that matter. Then write those numbers. A stack-up that uses a generic title block is a story about a factory you do not have.

DFM also decides what you will not inspect. A scratch spec nobody can see at the line is not a spec. A flash limit with no gauge is a debate. The defects you care about need a place to be seen and a limit the process can meet.

Material mass matters when the material is expensive or the part is large. On a small moulded cover, the tool, the cycle, the side actions, the scrap from a thick section, and the assembly of the cover to its neighbour usually dominate a gram of resin. Chasing mass without naming those drivers produces a thinner part that warps, and a "saving" that shows up as a fixture.

The same shape can have two honest costs. Machined from plate, the cost is mostly time and setups, and it falls only slowly as quantity rises. Moulded, the cost is a tool plus a smaller piece price, and it is absurd at fifty parts and ordinary at several thousand. DFM includes choosing which of those curves you are on. It does not include pretending both curves are available at the prototype quantity.

The teaching part is an L-shaped bracket that holds a sensor in a metal chassis. The prototype was milled. Internal corners are sharp because an end mill left them sharp, and the team then radiused nothing. Hole positions are ±0.05 mm because the mill held them. The proposal for volume is to injection mould the same file in a filled resin, "with DFM".

That proposal has not named a process so much as changed the word. A mould will not love a sharp internal corner, will not hold every hole at the milled band without a reason, and will add draft the mating chassis, still sheet metal, may not have room for. A teaching alternative, not a trial: keep the bracket in sheet metal, with slotted holes that take up the chassis stack, and a single formed flange. The sensor is clamped to a face that is the datum, not to a boss grown in the middle of a moulding. If a moulded bracket is truly required, the holes that locate become the only tight features, the rest open up, and draft is designed with the chassis, not added after the chassis is frozen.

Both options can be DFM. They are DFM for different processes. Calling the milled file "production" because a moulder was willing to quote it is how sharp corners and fantasy tolerances get a tool.

Hypothetical teaching maths. Not a process capability study and not a BrahmWorks result.

Assume a moulded bracket's ordinary hole position, as agreed with a moulder for this exercise, is ±0.3 mm, and that demanding ±0.05 mm adds a post-machining operation at an invented INR 18 per part. Assume 4,000 brackets.

Extra cost = 18 × 4,000 = INR 72,000.

Assume the function is a sensor that still meets its aim if the hole is ±0.3 mm, because the chassis slot already has that much adjustment. Then the INR 72,000 buys a number copied from the mill, not a function. Assume instead that the sensor misses its aim outside ±0.1 mm, and that opening the slot is impossible. Then you either pay the 18 rupees, or you redesign the datum so the mould holds the one feature that matters and nothing else is asked to. The sum does not say ±0.05 mm is always wrong. It says a copied tolerance is a purchase. Write down what function it buys, or do not buy it.

• Process, material family and quantity are written on the review.
• Geometry uses what that process repeats, and prices what it punishes.
• Secondary operations are on the routing, or they are designed out.
• Tight tolerances sit only on functional features, with a datum.
• The rest match what the maker will hold without a special step.
• Cosmetic and flash limits can be seen and judged at the line.
• Tooling, cycle, scrap and assembly are in the cost, not only mass.
• A prototype made by a different process is not treated as proof.
• The maker has seen the features you think are hard, before steel or a hard fixture.

No. DFM makes the cost explicable and removes costs that buy no function. A cheaper resin that fails a chemical requirement is not DFM. A dearer process that is the only one that holds a true functional tolerance can be DFM. The test is whether each rupee is attached to a requirement, not whether the spreadsheet cell went down.

It proves someone could machine that file, or something close to it. It does not prove a mould, a stamping tool or a casting will repeat it. Corners, draft, shrink and setup datums change with the process. Keep the machined unit as evidence of function. Do not keep it as evidence of the production process unless production is machining.

No. A chassis, a cover, a shaft and a seal can have four processes. Forcing one process for the sake of a tidy story creates parts the process is bad at. What should be consistent is the datum scheme between them, so a precise machined bore is not located from a loose moulded wall without that looseness being in the stack.

In the material note, the process note if you specify one, the datums, the few tight tolerances, and the features you explicitly leave to a general tolerance the maker has accepted. A drawing that is silent, plus a slide titled DFM, is not a manufacturing definition. The maker will fill the silence with their defaults. Those defaults may be fine. They are still someone else's design.

Bring the model, the process you intend, and the quantity you will place. BrahmWorks can tell you whether the geometry is aimed at that process, or at the prototype method you happened to use first.

Review the next stage with BrahmWorks