Automating quotes in a machine shop with Python and spreadsheets

In a job shop, every RFQ repeats the same steps: read the drawing, pick stock, estimate setups and cycle time, add heat treatment or coating, apply margin, write the quote. The judgment is in the estimates. The time goes into the arithmetic and the document. That is the part worth automating.

What the estimator keeps, what the tool does

Step	Estimator	Tool
Read the drawing, spot risks	✓	Can pre-fill title-block data
Choose process and operations	✓	Suggests from part family
Stock size and cost	Confirms	Picks from stock list and price table
Setups and cycle times	Adjusts	Applies rules per operation
Outside processes	Confirms	Uses last supplier prices
Margin and quantity breaks	Decides	Calculates
Quote PDF	Reviews	Generates from template

The data tables

Put your rules in tables, not in code:

• Machines: hourly rate, setup rate.
• Stock: material, shape, size, price per kg or per metre, minimum cut length.
• Operations: setup time and time rules per operation type (face, turn, drill, mill pocket), by material group.
• Outside processes: supplier, price per piece or per kg, minimum charge, lead time.
• Margins: by customer type and quantity.

The estimator can change any value without touching code.

A small engine

The calculation itself is short. A simplified version:

from dataclasses import dataclass

@dataclass
class Operation:
    machine_rate: float      # per hour
    setup_min: float         # minutes per batch
    cycle_min: float         # minutes per part

def quote(qty, stock_cost_per_part, ops, outside_per_part, margin):
    machining = sum(
        op.machine_rate * (op.setup_min / qty + op.cycle_min) / 60 for op in ops
    )
    cost = stock_cost_per_part + machining + outside_per_part
    return round(cost * (1 + margin), 2)

ops = [Operation(90, 45, 6.5), Operation(110, 60, 4.0)]
print(quote(qty=50, stock_cost_per_part=3.20, ops=ops, outside_per_part=1.10, margin=0.25))

The numbers above are placeholders. The value is that setups are spread over the batch automatically, and every term is visible.

Rolling it out without breaking quoting

1. Pick one part family you quote often: turned shafts, flat plates, simple brackets.
2. Collect twenty past quotes for that family, with what was actually charged.
3. Calibrate the rules until the tool reproduces them within a range the estimator accepts.
4. Run in parallel for a few weeks: the estimator quotes as usual and compares.
5. Switch when they trust it, and keep the old way for parts outside the family.

Where AI helps

Language models are useful at the edges: pulling part number, material, quantity and due date out of an RFQ email or a PDF title block, or drafting the cover text of the quote. Keep them out of the price calculation, and always show the estimator what was extracted before it is used.

How do you estimate cycle time without a CAM program?

Most shops do not program CAM before quoting, and they should not. Cycle time at quote stage comes from rules the estimator already uses in their head, made explicit:

Operation	Typical rule input	Example rule form
Turning	Length and diameter removed, material group	minutes = k × volume removed
Drilling	Number of holes, diameter, depth	minutes per hole by size band
Milling pockets	Volume removed, material group	minutes = k × volume / material factor
Deburr and inspect	Part size and tolerance class	fixed minutes per part band

The constants come from your own history. Take finished jobs where you know the real times, fit the constants, and accept that the rule is an estimate the estimator can adjust. The goal is a consistent starting point, not a CAM-accurate number.

Handling quantity breaks

Customers ask for 10, 50 and 200 pieces in the same RFQ. The setup spreads differently in each case, stock may switch from cut pieces to bar feeding, and outside processes often have minimum charges. Compute each quantity independently with the same rules, instead of applying a percentage discount. The quote then shows why the 200-piece price is lower, and the estimator can see when a minimum charge from the plating shop makes the 10-piece price jump.

Generating the quote document

The last step is the one that saves the most typing. Keep your existing quote template (Word, Excel or a PDF form), add placeholders for customer, part number, revision, quantities, prices, lead time and validity, and fill them from the engine. Two details make the document trustworthy:

• Revision and drawing reference printed on the quote, so a later revision triggers a requote instead of an argument.
• Assumptions block listing material, tolerance class, finish and what is excluded. Most disputes after a quote are about assumptions nobody wrote down.

Keeping it maintainable

A quoting tool that only its author understands becomes a risk. Keep it boring: rules in tables, a short README, one function per cost element, and a test file with five or six past quotes that must still produce the same prices after any change. When someone updates machine rates, the tests show which quotes moved and by how much.

What not to automate

Leave manual the parts that need judgment every time: unusual materials, five-axis work, assemblies, anything with a new customer requirement. Mark them in the tool as “manual quote” instead of forcing a formula. A tool that is right for most routine RFQs and honest about the rest is more useful than one that pretends to cover everything.

What it is worth

The gain is consistency as much as speed: two estimators quoting the same part get the same base price, and every quote can be explained line by line.