DFM vs DFA vs DFR

Calling all three "a DFM review" is how a clip is approved because it needs no slide, while nobody owns the force required to seat it or the fact that the clip is the spring. Separate the questions, write the conflict, and give the decision an owner. Figures in this article are hypothetical teaching maths, not BrahmWorks prices, yields, timelines or test results. How the three survive contact with a real build is part of How to Take a Hardware Prototype to Production.

DFM may block a feature the named process cannot form, fill, machine or hold: a side hole with no draw, a wall that will not fill, a tolerance the process will not repeat. It may not block a feature only because another shape would be cheaper, once you have accepted the cost and the process can do the work. Cheap and possible are different words.

DFA may block a stack that only goes together with a private knack: unstable orientation, a fastener used as a fixture, a joint hidden before it is checked. It may not demand the minimum part count when service, a seal or a different material needs a separate piece. Fewer parts is a hypothesis. Repeatable assembly is the requirement.

DFR may block a joint, a material or a thickness that will not survive a duty you will put in writing. It may not demand extreme margins on a quiet indoor product, and it may not treat an unwritten fear as a requirement. With no duty statement, DFR has nothing to pass. "Make it robust" is not a duty.

The conflicts are ordinary, and they should be visible.

A moulded snap can be excellent DFM: one part, no slide, no screw boss. It can be poor DFA if seating force is high and nothing shows that the snap is home. It can be poor DFR if root strain is high, the resin creeps, or a user opens it daily. Replacing it with a screw can help service and hurt piece price. None of those sentences cancels the others.

A thick wall can feel like reliability and still sink, warp and slow the cycle. A thin wall can fill and eject, then buckle under a drop or a clamp. Assembly may demand a local pad that manufacturing must gate and cool. The pad is a second requirement, not a defeat.

Running the three as one meeting, with one person talking, tends to let the loudest constraint win. Run them as three questions on the same feature, even if the same engineers sit in the room. The output is a table, not a mood.

A cell is not a certificate. It is a test of whether you are answering the question in that row.

Write the feature, the three answers, and which requirement may overrule. If a forbidden witness line is the only way to keep a seal without a slide, cosmetics and the seal are in conflict. Someone chooses, and the choice names a consequence: a witness, a dearer tool, a seal you drop, or a shape change. An unnamed conflict returns as steel rework or as a field failure nobody "asked for".

Do not average them. A feature that is slightly bad at all three is not a compromise. It is three open items. Prefer a clean decision you can inspect.

The teaching product is a handheld tool with a user-opened battery cover. It is not a BrahmWorks product. The first proposal is a pair of moulded snaps on a straight-pull tool. DFM is content. The designer seats the cover in one motion. Reliability is argued from five openings.

On paper, DFM can pass: the snaps draft in the draw, the roots miss the weld, the cover ejects. DFA fails while the cover can be rotated and still click, and while a half-seated snap does not stand proud. DFR is an unrun test: weekly opening for two years, plus a solvent wipe, is not answered by five openings in a printed resin.

Teaching disposition, not a trial outcome: keep one moulded snap, add a screw the user can see so a missed snap cannot look closed, and leave the solvent wipe as an open material question. The screw adds a cycle. It buys a visible incomplete state. DFM accepts one boss. DFR no longer rests on an unconfirmed snap as the only lock.

Hypothetical teaching maths. Not a quotation or a BrahmWorks result.

Assume a screw and a brass insert add INR 4.50, plus 15 seconds of assembly. At an invented INR 180 per hour, labour is 15/3600 × 180 = INR 0.75. Unit add is INR 5.25. At an assumed 6,000 units, 5.25 × 6,000 = INR 31,500.

Assume the alternative, two snaps and no screw, avoids that sum and also avoids an insert. Assume a snap redesign and a tool adjustment, if the snaps prove wrong after steel, at INR 70,000. The screw does not "cost" INR 31,500 against a reliability claim you have not measured. It costs INR 31,500 as a known line, set beside an unpriced risk. If your duty is "closed once in the factory, never opened", the screw is a weak buy and the snaps, confirmed seated, may be the better trio of answers. If the duty is weekly opening plus a solvent, INR 31,500 is the price of not betting the product on an untested spring. The arithmetic does not choose. It stops the screw being called free, and it stops the snaps being called reliable.

• The process is named, so DFM has a target.
• The duty is a sentence with a cycle, a load or an exposure, so DFR has a target.
• Assembly is judged by someone other than the designer.
• Each disputed feature has a DFM line, a DFA line and a DFR line.
• A pass on one line is not copied into the other two.
• The overruling requirement has an owner.
• The consequence of the losing requirement is written: cost, witness, service, or a claim you drop.
• Tool money is not spent to close a DFR question a cheaper sample could close.
• "Robust" and "simple" do not appear as the only justification.

Yes. A straight-pull clip with a sharp root can be easy to mould and a poor spring. A generous fillet can be kinder to the resin and harder to fill or to eject. Record both. The process being capable does not mean the duty is met. The duty being met in a machined sample does not mean the mould will reproduce the fillet you tested.

Only for questions that need moulded parts: knit lines, texture, a production-cycle skin. Strain, solvent and a gross clash can often be probed earlier, in the production resin or a named stand-in. Waiting because steel feels official turns every surprise into a tool change.

No. Screws can be the thing that makes assembly checkable: one driver, one torque, a joint you can see. Snaps can add a decision you cannot audit. DFA removes unstable orientations, hidden joints and parts that do no work. It does not worship a fastener count. A screw that clamps a located joint can be better assembly than a clip that only the designer can seat.

The person who owns the product requirements, advised by whoever owns the process and the duty test. A supplier can tell you a feature is expensive or impossible on their process. They cannot tell you which user requirement to drop, unless you have asked them to. Leave the decision implicit and the toolmaker, the assembler and the field will decide it for you, one after another, at rising cost.

Bring the feature, the process you intend, and the duty you are actually willing to claim. BrahmWorks can lay the three answers side by side and name which requirement is being silently overruled.

Review the next stage with BrahmWorks