Should-Cost Analysis for Machined Parts

A should-cost model is an independent estimate of what a part ought to cost given how it has to be made. Built well, it does one useful thing: it tells you where the money is.

That matters because the two most common purchasing responses to a high quote are both weak. Asking for a discount treats the price as arbitrary. Asking three more suppliers treats the price as a market fact. Neither tells you that seventy percent of the cost is a single deep pocket that needs a long reach tool at a slow feed, which is a design conversation with a much bigger prize than a discount.

The misuse is treating the model output as an entitlement. Your estimate does not know the supplier machine list, its labor market, its current loading or its overhead, and a supplier told that its price is wrong by a spreadsheet will usually decline the argument rather than open its books. Use the model to find the driver, then talk about the driver.

Almost every machined part price decomposes into the same five. Getting each within a reasonable band is enough; precision in the model is not the point.

Two mistakes account for most bad material estimates.

The first is costing the finished weight. Shops buy stock, not parts. A pocketed housing that finishes at 300 grams may start from a 1.4 kilogram block, and you pay for the block. Estimate the smallest standard stock size the part fits in, with machining allowance on the faces, and cost that.

The second is ignoring the stock size step. Material comes in discrete sizes. A part 52 millimeters deep does not buy 52 millimeters of plate; it buys the next size up, and the difference is scrap. This is why a few millimeters of design change sometimes moves the material cost more than a material substitution would.

Credit for scrap value exists but is usually small and inconsistently applied. Aluminum and copper alloy turnings have real value; mixed or contaminated scrap often does not. Leave it out of a first pass model rather than arguing about it.

You are not trying to reproduce the shop estimate. You are trying to get within a band that tells you whether the quote is in a sensible place.

Work feature by feature. Count the setups first, because each one costs both setup time and cycle time in load, unload and touch off. A part that needs three orientations is a different part from one that needs two, and reducing that is usually the single largest lever available.

Then sum the obvious work. Roughing volume divided by a plausible material removal rate. Finishing area divided by a plausible surface rate. Holes counted by type and depth, since a drilled hole, a reamed hole and a deep hole are three different costs. Threads counted separately, especially blind and fine ones.

Then add the things that cost more than they look. Deep pockets needing long reach tooling run slow because the tool cannot take the load. Thin walls run slow because they chatter. Tight tolerances add a finishing pass and an in process measurement. Fine surface finishes add passes. Small internal radii force a small cutter, and a small cutter removes material slowly no matter how fast it spins.

Apply a machine rate at the end. Rates vary widely by region, machine class and shop, so treat these as a band rather than a number.

A 6061-T6 bracket, roughly 120 by 80 by 40 millimeters, two setups, one pocket, eight holes, four tapped, deburred and clear anodized.

Material: a 130 by 90 by 45 block of 6061 plate, about 1.4 kilograms bought, roughly 9 dollars at a plate price of about 6.50 per kilogram.

Setup: 2.5 hours of programming, fixturing and first article at about 85 per hour, so about 210 dollars for the lot.

Cycle: about 22 minutes across both setups at 85 per hour, so about 31 dollars per part.

Secondary: deburr about 4 dollars, anodize about 6 dollars per part in small lots and less in larger ones.

The interesting part is how the total moves.

Three things fall out of that table, and they are the reason to build the model at all.

Setup dominates at low quantity and disappears by a few hundred pieces. So at quantity 10, asking for a better price is asking the shop to eat setup, and combining two parts into one setup or ordering 25 instead of 10 will do more than any negotiation.

Cycle time barely moves with quantity. Past the first efficiency gains it is a fixed property of the geometry, which means the only real lever on it is the design. If cycle is your biggest bucket and you need a lower price, the conversation is about the pocket depth, the internal radii, the number of setups and the tolerances, not about the rate.

And the curve flattens. If quantity 1000 is only twenty percent below quantity 100, ordering a year of inventory to chase price is buying obsolescence risk with cash. That is worth knowing before you commit to a build quantity.

Lead with the question, not the number. "Our estimate says this part is cycle time dominated, is that consistent with how you would run it?" invites an answer. "We think this should cost 51 dollars" invites a defense.

Ask what would have to change to move the price meaningfully. Good shops answer this well and specifically, because they see it every day: a radius that lets them use a bigger cutter, a tolerance that removes a finishing pass, a feature moved so the part runs in two setups instead of three.

Then check your own model against the answer. If a supplier tells you the driver is something your estimate did not include, your estimate was wrong, and the useful outcome of the exercise was learning that. A should-cost model that never gets corrected is not a model, it is a position.