Design for Manufacture

The first and most powerful DfM move is to simplify the geometry of the part. Simple shapes — flat, rectangular, prismatic, with clean edges and right angles — are faster to make, easier to hold in a jig, more tolerant of small errors, and cheaper in almost every process than complex ones. A complex curve, a tight internal corner, a bespoke cut-out, a changing cross-section: each of these adds tooling, time, scrap and the chance of defects. This is why so much factory-made building fabric is built up from flat planes and straight extrusions — stud-and-board panels, flat floor cassettes, straight steel sections — and why curves and one-offs, where they are wanted, are concentrated in a few special parts rather than spread everywhere. Simplifying is not the same as impoverishing: it means achieving the architectural intent with the least geometric complexity, and spending complexity only where it earns its keep.

The second move is to design for the specific process that will make the part. A factory is not a generic maker; it has particular machines, each with its own grain. A CNC router cuts sheet goods efficiently in two dimensions but cannot easily make a deep undercut. A press-brake folds sheet steel along straight lines to set angles. A welding jig holds members in a fixed geometry. A casting mould wants draft angles so the part releases, and hates re-entrant shapes that lock in. Designing for manufacture means shaping the part to flow with these processes rather than fight them: aligning features to the axes the machine works in, keeping wall thicknesses within what the process likes, allowing the radii a tool actually cuts, and providing the faces a jig needs to grip. You cannot do this well without knowing, early, *which* factory and *which* system will make the part — which is why DfMA pushes the manufacturer relationship to the front of the project.

There is a deeper principle under both moves: a part should be as easy to make as its job allows, and no easier. DfM is not a licence to dumb down architecture; it is a discipline for spending difficulty wisely. The skilled DfMA designer knows which complexity the design genuinely needs and which is accidental — an artefact of habit, of drawing without thinking about making — and removes the accidental kind so the factory's capacity goes to what matters. Every gram of needless complexity you design out of a repeated part is multiplied across the whole production run in time, cost and risk saved.

If simplifying is about the shape of a part, standardising is about the relationship between parts — and it is where DfM meets the dimensional discipline that Module 5 develops in full. The heuristic is to work to a grid and to a limited palette of standard sizes and materials, so that the same stud, the same board, the same bracket, the same panel width recurs across the building rather than a new size every time. Standardisation buys three things at once. It lets the factory tool up once and run long: jigs stay set, machines stay calibrated, operators learn the part. It makes parts interchangeable, so a panel from one run fits anywhere the grid calls for that panel, and a mistake can be replaced from stock. And it simplifies procurement and inventory to a few lines rather than hundreds. The cost of a new, unique size is almost always underestimated, because it is paid not once but every time that part is set up, handled, stored and tracked separately.

Closely tied to standardisation is the drive to reduce part count and part variety — two different reductions, both valuable. Reducing *count* means achieving the design with fewer physical pieces: combining several parts into one (an integrated panel that is structure, insulation and lining together rather than three assembled layers), or eliminating parts whose job another part can do. Reducing *variety* means achieving the design with fewer *different* parts even if the total number stays high: a hundred identical panels is far better for manufacture than a hundred slightly different ones, because variety, not quantity, is what defeats a factory. The classic DfM questions — can this part be eliminated, combined with another, or standardised to match one already in the kit — are asked of every part, and each 'yes' compounds across the run.

For building elements, this plays out as a few repeated panel and module types rather than a bespoke envelope; a small family of standard openings rather than a different window everywhere; common connection details reused throughout; and a short, deliberate materials list. It is worth stressing the Indian dimension here: where abundant site labour weakens the 'save on labour' argument for off-site, standardisation and repetition are the levers that still make the case — large, repetitive housing, institutional and infrastructure programmes, and deep precast-concrete capability, reward the factory precisely because the same well-made part is needed thousands of times. The fewer and more standard the parts, the stronger the off-site case, wherever you build.

Every manufactured part has tolerance — the small, unavoidable range by which its real, made dimension will differ from the nominal dimension on the drawing. No process makes anything perfectly; the question is only *how much* variation a given process and material will produce, and whether the design has allowed for it. Design for Manufacture demands that you specify tolerances the factory can actually achieve with the process and material in question — neither tighter (which drives cost up sharply, as ever-finer tolerance is expensive and sometimes impossible) nor looser than the part's job requires. A tolerance that sounds reassuringly tight on a drawing but that the process cannot hold is not a specification; it is a guarantee of rejected parts, rework, or parts that do not fit when they reach site.

The subtle point — and the one that separates DfM thinking from wishful drawing — is that tolerances accumulate. If a wall is built from several parts each made to its own tolerance, the errors stack up: a row of ten panels each within a few millimetres can, in the worst case, be off by several centimetres end to end. A design that ignores this, assuming every part lands exactly on its nominal size, will produce a building that does not close up — gaps that are too wide or too narrow, modules that foul each other, connections that will not reach. DfM handles this by keeping tolerances realistic, by reducing the number of parts in a dimensional chain (fewer parts, fewer accumulated errors — another reason part-count reduction matters), and by locating from sensible datums rather than letting error compound blindly. The companion discipline, designing the *joints* to absorb this tolerance so imperfect parts still fit, belongs to Design for Assembly and to Module 5; DfM's job is to make sure each part's own tolerance is achievable and declared.

This is also where DfM meets its firm boundary. What tolerance a given factory, process and material can hold, and what tolerance a connection needs to perform for structure, fire and weather, are binding technical questions that belong to the manufacturer's tested system and to qualified engineers, checked against the National Building Code of India and local rules. The designer's job in DfM is to think in tolerance from the outset — to know that parts vary, to design so the variation is affordable and absorbable, and to ask the manufacturer what their system can actually achieve — not to invent tolerance figures. Any number quoted in a lesson is illustrative; the real values are the manufacturer's and the engineer's.

Running through all the heuristics is one practical habit: spot the hard-to-make feature early and design it out while it is still free to change. Some features quietly punish manufacture far more than they reward the design — a tolerance no process can hold, a material combination that will not bond or that traps moisture, a junction where three or four parts must meet at one point, a cut-out that weakens a panel just where it is handled, an access that the assembly later cannot reach. Individually each looks innocent on a drawing; in the factory each becomes a slow station, a source of scrap, or a part that fails inspection. The DfM designer develops an eye for these and, crucially, removes or redesigns them *at concept and scheme*, when the cost of change is a thought, rather than after tooling, when the cost is real money and lost time. A useful discipline is to walk every part through the question: *how, exactly, would the factory make this — and where would it struggle?*

It helps to make this concrete for building elements. A floor cassette made of standard joists at a standard spacing, sheathed in standard boards, is easy: repetitive, jig-friendly, forgiving. The same cassette with a one-off service penetration in a different place on every unit is not — variety has crept in. A facade of repeated standard units with complexity concentrated at a few special corner and parapet pieces is manufacturable; a facade where every unit is subtly unique is a factory's nightmare. A connection bracket used in one standard form throughout is easy; a dozen near-identical brackets that differ only slightly invite the wrong one being fitted. The skill is to see, in your own design, where you have unwittingly created variety, complexity or difficulty, and to consolidate it back toward the standard kit.

Two honesties close the lesson. First, DfM is always in tension with other goods — architectural intent, site constraints, cost, performance — and the answer is not to maximise manufacturability at all costs but to *trade it consciously*: to know what each bespoke or complex part costs the factory, and to spend that cost only where the design genuinely needs it. Naive DfM produces monotony; absent DfM produces unbuildable or ruinous prefab; good DfM is the judged middle. Second, DfM tells you how to *think about making*, but it does not certify anything: whether a part actually performs structurally, resists fire, keeps out weather, and can be made to the needed tolerance are results that belong to the manufacturer's tested, proprietary system and to qualified engineers, under the governing codes. Design for manufacture well, and the factory becomes the ally the hook promised; ignore it, and the copying machine multiplies your mistakes.