Should-Cost Model Calculator

What is a should-cost model and what does it actually tell you?

A should-cost model is your own independent estimate of what a competent supplier should charge to make a part and still earn a sustainable return. You build it from the drawing up: material, process time, overhead, tooling, logistics, SG&A, profit. Then you put it next to the quote.

The structure is standard — materials, then labour, then machine and overhead, then tooling and capital, then logistics, then SG&A, then profit — with each line calculated as quantity times rate times yield.

The calculator's defaults model a machined aluminium bracket: 0.42 kg finished, four operations, 250-piece batches, 12,000 a year. It builds to a should-cost of $20.50 against a $24.80 quote — a $4.31 gap per part, 21%, worth $51,660 a year.

What the model is not is a prediction of the supplier's price. It should be close enough to isolate the major drivers behind the gap — cycle time, tooling, yield, routing, material, overhead, profit — and its real value is exposing which assumptions need testing.

That reframes the conversation. Instead of "your price is too high," it becomes "our model has 4.6 minutes of milling cycle time and an 18% factory overhead — where do yours differ?" The gap between estimate and price is the starting point of the discussion, not the conclusion.

Why does batch size change the price so much?

Because setup is a batch cost and everything else is a part cost, and dividing one by the other is where price non-linearity comes from.

The illustration used across the field: a part with a two-minute cycle time sounds cheap until you learn the setup takes 45 minutes. On a run of ten, that setup amortises at 4.5 minutes per part — more than doubling the effective labour cost. On a run of a thousand, it's negligible.

The calculator's defaults carry 90 minutes of total setup across four operations. At a 250-piece batch that's 0.36 minutes per part against 7.10 minutes of cycle time. Change the batch and the whole price moves:

A $6.12 range on the same drawing, driven by nothing but run length. And notice the shape — the curve is steep at the small end and flat at the large end. Going from 25 to 250 saves $5.64; going from 250 to 1,000 saves another $0.48.

Two practical consequences. First, quoting an annual volume without a batch size gives the supplier room to assume the worst and price accordingly. Quote both. Second, if you're chasing cost down, the first question isn't the rate — it's whether you can consolidate releases into fewer, larger runs. That's usually cheaper than renegotiating.

How should scrap and yield be handled?

Compounded across operations, not applied once at the end — and the difference is bigger than it looks.

Each operation has its own yield. Multiply them and you get rolled throughput yield. The calculator's defaults run 99.5%, 97%, 99% and 98%, which multiply to 93.64% — a 6.36% loss, not the 5% you'd get by adding them up wrong or the 3% a single end-of-line assumption might use.

The consequence is that you must start 1.0679 parts for every good one, and every scrapped part has already consumed material and machine time at every operation it passed through before failing. A part scrapped at anodising burned the sawing, the milling and the deburring too.

The calculator prices what the shortcut misses. Modelling a single 3% end-of-line scrap rate produces a should-cost of $19.96 against the correct $20.50 — $0.54 per part understated.

That error runs in the dangerous direction. Understating your should-cost inflates the apparent gap, so you walk in asking for a reduction the supplier can demonstrate is unjustified, and you lose credibility on the numbers that were right.

On material specifically: account for grade, scrap allowance and yield. Machining aluminium commonly needs a 5–10% scrap allowance on top of the geometric loss, and the calculator handles the geometry through material utilisation — 0.62 here, meaning 38% of purchased weight becomes chips. It credits those chips back at recovery value, which most quick models forget entirely.

What overhead and margin should I assume?

Carefully, because this is where double-counting happens and where a model loses credibility fastest.

Overhead is the hardest component to estimate because it varies enormously between shops — a newer facility with CNC machines still under finance carries far higher depreciation than an older shop with paid-off equipment. The common rule of thumb applies overhead as 150–200% of labour.

But that rule assumes a bare labour rate. The calculator uses fully burdened chargeout rates — $85 an hour for CNC milling already contains the machine, its depreciation, energy and maintenance. Applying a 150% labour burden on top would count the machine twice. So factory overhead here is a modest 18% addition covering plant costs not already in the rates.

Check which convention your rates follow before you trust anyone's overhead percentage, including your own.

Supplier margin is more tractable. Typical ranges run 8% to 20% depending on complexity and competition; the defaults use 11%. Resist the temptation to model a supplier at zero margin — a model that leaves your supplier no return isn't a negotiating position, it's a fantasy, and the supplier will tell you so.

At the defaults the structure comes out 23.7% material, 54.5% conversion and overhead, 17.1% SG&A and margin. That split is itself diagnostic. A part that's 70% material is a commodity negotiation about metal prices; a part that's 70% conversion is a conversation about cycle times and routing.

My model says $20 and the quote is $25 — is the supplier overcharging?

Probably not. Small gaps of 5–10% are normal; large gaps usually signal incorrect assumptions rather than bad faith — a wrong material grade, an underestimated process complexity, or a missed operation.

The calculator's defaults produce a 21% gap and flag it as "Large — drill into assumptions before alleging overpricing." That flag is deliberate. A 21% gap is a prompt to check your own work first.

Go through the likely causes in order:

• Did you miss an operation — heat treat, secondary machining, a plating step, final inspection?
• Is your cycle time from a similar part rather than this one?
• Is the batch size in your model the batch the supplier is actually planning to run?
• Is your material grade the one on the drawing?
• Is the supplier carrying tooling cost you've amortised over a different volume?

Then take specifics into the room. Lead with a line item — "our model estimates material at $4.86 a part on 0.68 kg of gross weight at $6.80 a kilo" — rather than a target price. That invites correction on a fact rather than a fight about a number.

Two scope limits worth stating. This is an ex-works manufacturing cost; duties, tariffs, inbound freight and inventory carrying are a separate total-cost-of-ownership question. And it models a single routing — if you want to compare nearshore against offshore, or a machined part against a casting, run it twice and compare the outputs rather than trying to blend them.