Design for manufacturability is routinely described as a checkpoint — a report someone produces once the CAD is finished. That framing is the reason it so often fails to help. By the time a design is finished, the decisions that determine whether it can be built economically have already been made, and unmaking them is expensive.
It is a sequence of trades, not a gate
Every manufacturing process has a grain. Injection moulding wants uniform wall sections, generous draft and a parting line it can reach. Sheet metal wants bend reliefs and a minimum flange. Machining wants tool access and internal radii that match a real cutter. Designing with the grain costs almost nothing. Designing across it costs cycle time, tool complexity, scrap, or all three.
The useful question at any point in a design is not "can this be made?" — almost anything can be made. It is "what does this geometry cost, and is the thing it buys worth that?" Sometimes it clearly is. A visible surface, a sealing face, a load path: these earn their complexity. Most geometry does not.
Tolerance is where the money hides
Tolerances are the most commonly over-specified property in hardware. A drawing that carries tight tolerances everywhere communicates nothing about what actually matters, and prices every feature as though it were critical. The alternative is deliberate: identify the dimensions that control fit, function and appearance, tighten those, and open everything else.
This requires an actual stack-up analysis rather than an instinct. When several parts mate, the accumulated variation — not the individual tolerance — determines whether a gap stays flush or a button binds. That analysis is cheap while the design is in CAD and impossible to retrofit once tooling exists.
What good DFM produces
- A written record of which features are critical and why
- Process selection confirmed by simulation, not assumption
- A tolerance stack-up across every functional interface
- Material choices validated against lead time and availability
- A costed BOM that reflects the design as it stands today
- A risk register naming what could still go wrong in tooling
None of that is glamorous. All of it is the difference between a programme that tools once and a programme that tools twice.