Design for cost: what a tolerance really costs

What does a tolerance cost?

A tolerance never appears on the drawing as an amount. It costs money because it changes the route: an extra finishing pass, a slower cut, one more setup, a finer gauge. Trade site DeIndustrie.Online puts it plainly: tight tolerances call for better machines, sharper tools and more operations, so start generous and tighten only where the function demands it.

Every process holds a range. ISO 286 expresses it in IT tolerance grades: for a size over 30 up to 50 mm, IT7 is a band of 0.025 mm, IT9 0.062 mm and IT11 0.16 mm. Ask for a grade the first process cannot hold, and you are asking for a second process behind it.

Cost does not rise in a straight line. Benzaken, Doostan and Evans call the reciprocal power model perhaps the most popular cost model in tolerance research: manufacturing cost climbs ever more steeply as the tolerance is tightened. In its simplest form, halving the tolerance doubles the tolerance-driven part of the cost. A model, not a measurement of your shop, but the shape is familiar: the last hundredth is the expensive one.

Which cost lines does a tight dimension move?

A dimension without its own tolerance falls under the general tolerance, which ISO 2768-1 sets in four classes. A tight dimensional tolerance seldom touches just one line. Take a bore that goes from the general tolerance to IT7.

Machining: where an end mill would have interpolated the bore in one pass, you now pre-drill and ream, or bore it, each with its own tool. Setup: if the bore must also sit exactly relative to a face on the far side, you want both in one setup, and that can overturn the route. What a 0.02 position tolerance does to the route shows how.

Inspection: a band of a few hundredths is no longer checked with a calliper but with a plug gauge, a bore micrometer or a CMM: time per part or per sample. Scrap: the closer the band sits to what the machine holds, the more parts can fall outside it.

‘That IT7 costs you a reaming pass and a gauge check per part’ is a sentence a designer can weigh. ‘Tight tolerances are expensive’ is not.

What do roughness and material do to the price?

Surface roughness works the same way. In milling and turning, a lower Ra often means a separate finishing pass at a lower feed, or grinding as an extra step. So ask for Ra 0.8 on the face that seals or slides, not as a blanket note in the title block.

Material moves two lines at once: the material line and the machining time. A review in Acta Mechanica Slovaca counts austenitic stainless steel among the hardest materials to machine: it work-hardens under the cut and conducts heat poorly. For a range of austenitic grades the authors give a machinability of about 40 per cent of free-machining steel. The same bracket in 1.4404 rather than EN AW-6082 costs more in material and is slower to machine. If the designer is weighing corrosion resistance, put both lines side by side.

What does a feature on the wrong side cost?

A design choice that barely shows on the drawing is where a feature sits. A threaded hole on the underside of a bracket that is otherwise milled from the top means a second setup: reclamp, re-zero, re-measure. Counting setups shows how that adds up. A small internal corner radius in a pocket does the same to machining time: it forces a small cutter, which removes material slowly.

In a small batch that second setup weighs heavily; in a large batch cycle time dominates. Price the job per quantity break and the buyer sees where the redesign pays most; why a small batch costs so much per part does the arithmetic.

Worked example: one bracket, two drawings

A bracket in EN AW-6082, a batch of 25, black anodised by a subcontractor. The first drawing asks for two bores in IT7, Ra 0.8 on every machined face, an R2 internal radius in a pocket and a threaded hole on the underside. The shop's proposal keeps IT7 on the one bore that takes a shaft, limits Ra 0.8 to the sealing face, opens the radius to R6 and moves the threaded hole to the top face.

Anodising, a subcontract line, does not change. Machining falls with a reaming pass and a finishing pass fewer and a larger cutter in the pocket; setup halves with one setup fewer; inspection falls because one bore gets a plug gauge instead of two bores going on the CMM. Together that saves €17 a part; which line moves most depends on your own machines and rates.

Who holds the conversation: designer, buyer or shop?

Three people each hold part of the answer. The designer knows which requirement carries the function. The buyer has a target price or a previous price and compares suppliers. The shop knows how the part is made and what each requirement costs there. Without the third, the other two discuss a total.

Value management has a European standard, EN 12973:2020; one of its stated aims is guidance for value-based decision-making. At a supplier it gets concrete: a cost engineer who sends a second variant with the quote, showing line by line what changes.

Two common routes leave the shop out. A design suite built for engineering departments prices with standard machines and standard rates, not yours. A platform that passes requests to the lowest bidder turns your know-how into a price per part. Both lose what matters: which requirement moves which line.

What does the shop gain by showing the cost line?

A total gets compared; a cost line gets discussed. The shop that can say ‘this second setup costs half your setup time’ gets a seat at the next design, not just the next price round. The conversation moves from discount to redesign, and there you are the party that already knows the route.

That takes a cost build-up in lines: each estimate line with its inputs, time, rate and quantity. A price from one spreadsheet cell, or from the cost engineer's head, cannot be taken apart with the customer at the table.

Never change the drawing yourself: the choice stays with the designer, the evidence comes from you.