"Is this quote fair?" Most buyers answer by comparing three quotes. A better answer is to know what it should cost before the quotes arrive.
What it is
Should-cost analysis builds a cost from the bottom up:
- Material. How much, what grade, at what price, plus scrap.
- Process. Each operation, its cycle time, and the machine and labor rate.
- Setup. Spread over the batch size.
- Overhead. Facility, tooling, quality, logistics.
- Margin. A reasonable profit for the supplier.
It's used heavily in automotive, aerospace and government buying. The point isn't to squeeze suppliers. It's to understand where cost comes from, so you can talk about it, and so design can avoid expensive features.
Why people skipped it
Building one takes a cost engineer who knows the processes and the rates. Most companies have few or none. So should-cost gets done only for the biggest parts.
How a bot does it
- Give it the drawing and the spec. Material, dimensions, tolerances, finishes.
- Give it your rate table. Machine rates, labor rates, material prices, overhead factors. These must come from you or a trusted source.
- Have it propose a routing: the likely operations, in order.
- Have it estimate each line and show the math.
- Have it point out the cost drivers: the features that add the most cost.
- A person who knows the processes checks the routing and the big lines.
The bot brief
Build a should-cost estimate for the attached part.
Inputs: drawing [file], spec [file], rate table [file], annual volume [n], batch size [n].
Use ONLY rates from the rate table. If a rate is missing, stop and list it.
1. Propose a process routing, step by step. Mark it as PROPOSED.
2. For each step: cycle time estimate (with reasoning), rate used, cost.
3. Material: gross weight, scrap allowance, price used, cost.
4. Setup, overhead and margin per the rate table rules.
5. Total per piece. Then the top 3 cost drivers and, for each, a design or process
change that might reduce it.
Mark every cycle time as ESTIMATE and show the reasoning.Starting point only. Test it on your own data and model, and adjust as needed before you rely on it.
An example
Steel mounting bracket, 5,000 per year, batches of 500. All rates from the shop's rate table.
| Line | Basis | Cost per piece |
|---|---|---|
| Material | 0.8 lb steel + 10% scrap at $0.90/lb | $0.79 |
| Laser cut | 40 sec at $120/hr | $1.33 |
| Bend (3 bends) | 30 sec at $90/hr | $0.75 |
| Setup | 1 hr laser + 1 hr brake, over 500 pieces | $0.42 |
| Powder coat | Outside coater, quoted rate | $1.10 |
| Overhead and margin | 25% of the above | $1.10 |
| Should-cost | $5.49 |
If quotes come in at $9, there's something to talk about. And the bot's driver list says the powder coat and laser time matter most, which tells design where to look.
Where bots differ
The structure is easy for a bot. The numbers are the risk. Cycle times are judgment, and a bot that has never seen your press brake is estimating. Rates must come from your own table, not from the bot's memory. See Don't make it up.
What this means for you
- Keep a current rate table a bot can read.
- Have the bot draft a should-cost before quotes come in.
- Have a process expert check the routing and cycle times.
- Use the cost drivers in design reviews.