Costing a machined part when the shop has no data yet
You do not need an ERP system or a month of study to produce a defensible cost. Pick one part, measure three numbers over a week, and take the rest from records the shop already keeps.
Most shops that have never costed a part properly assume the barrier is a system. It is not. The barrier is that costing *everything* is a large job, so it never starts.
So do not cost everything. Cost **one part**, well. A single accurate costing is worth more than a spreadsheet of estimates, because you can defend it and because it becomes the baseline every improvement is measured against.
Step 1 — Pick the right part
Not the most interesting part. The one that meets all three of these:
- You run it regularly, so you can observe it rather than remember it.
- It goes through several operations, so the exercise teaches you the routing.
- Someone already suspects it is priced wrong. That gives the work a reason to finish.
Step 2 — Measure three numbers over one week
Only three numbers require you to go and look. Everything else already exists on paper somewhere.
| Number | How to get it | Why this one |
|---|---|---|
| Cycle time per part | Time several consecutive parts and take the average, in normal conditions — not a demonstration | It drives machine cost and it is the number people misremember most |
| Setup time | Time one changeover from last good part to next good part | Usually larger than anyone expects, and it is what makes small lots expensive |
| Parts per tool change | Ask the operator to mark the count between two changes | Turns tooling from a guess into a measured cost |
Step 3 — Take the rest from records you already have
None of the following needs new measurement:
- Material price and stock size — purchase orders.
- Labor rate — payroll, loaded with the on-costs your accountant already applies.
- Outside processing — vendor invoices for that part.
- Scrap rate — inspection records, or the difference between parts started and parts shipped.
- Overhead — total shop costs for a period divided by total machine hours in the same period.
The last one is the only calculation, and it is deliberately crude. A crude overhead rate applied consistently is far more useful than a precise one you cannot reproduce next quarter.
Step 4 — Enter it and read the warnings
Enter the part and let the system check it. The warnings are not decoration — they compare your figures against each other and flag combinations that are usually a data entry error rather than an unusual part.
Then leave the fields you genuinely do not know **empty rather than zero**. Empty and zero mean different things: empty says not yet established, zero says measured and there is none. A blank field treated as zero produces a cost that is wrong and looks exactly like a cost that is right.
What you now have
A baseline. Every improvement from here — a faster cycle, a longer tool life, a cheaper supplier — is measured as a change against a number you can defend, instead of an opinion about whether things feel better.
And the second part is much faster than the first, because the shop rate, the overhead basis and the labor rate are already established. The work you just did was mostly not about this part at all.
Related guides
- CNC machining cost breakdown: all ten cost categories, in process order
How to cost a machined part properly — material, machine hours, labor, tooling, scrap, rework, outside processing, packaging, freight and overhead — and why two engineers costing the same part get two different numbers.
- Six fields almost nobody fills in, and which way each one bends the cost
Material yield, personal and machine allowances, energy, units per pack and fixed cost per lot: six inputs that are usually left empty, and why five of the six make the cost look cheaper than it is.