How to Build a Should-Cost Model with AI Before You Send an RFQ
One supplier quotes ₹360 for a machined mounting block. Another quotes ₹290 for the same drawing. The second quote appears more attractive at first glance, but the order assumptions reveal important differences. The first supplier has priced batches of 50. The second has priced batches of 200. If your production schedule calls for 50 parts at a time, the ₹70 gap tells you very little about which supplier would be cheaper on your terms.
A should-cost model gives you a way to catch that mismatch before you send a request for quotation (RFQ). This guide shows direct-materials buyers and cost engineers at Indian automotive component and industrial machinery manufacturers how to use artificial intelligence (AI) to organise inputs, calculate a cost for a machined part, and ask for quotes they can compare. Buyers sourcing internationally can apply the same method with local rates and consistent delivery terms.
What a Should-Cost Model Estimates
A should-cost model estimates what a part would cost to make under an explicit set of assumptions. It builds the estimate from the materials, production steps, and resources needed. If you want to compare the estimate with a bid, you can then add an assumed supplier margin and specify what the comparison includes.
Set that scope first. Does the price include inspection, packaging, and freight? Is a fixture paid for up front or recovered in the piece price? Is the volume the annual forecast, the quantity per order release, or the number of parts the shop can run in one setup? Each can be different. An annual forecast helps a supplier plan capacity, but it only spreads a setup charge if the supplier can actually run more parts per setup.
What Goes Into the Cost of a Machined Part
Start with the stock the shop buys. A finished part’s weight omits material removed as chips or left over from cutting blanks. Then map the likely machining sequence. Separate setup, such as locating the fixture and proving out the first part, from the cycle repeated for each piece. Specify the inspection and packaging obligations, and put one-off and recurring tooling on separate lines where both apply.
Ask what the machine-hour rate includes. Labour, power, depreciation, and shop overhead may already be built into it. Adding those costs again would inflate your estimate. If you add overhead separately, specify which costs it applies to. Use the same care with profit. A 10% margin on the selling price and a 10% markup on the cost yield different prices.
Where AI Helps and Why Models Go Stale
Use AI to create a first-pass input sheet from the current drawing and specification. Ask it to capture the revision, units, material grade, finish, and inspection notes, with a page or callout reference for each entry. Mark any unclear or missing fields for further review. A buyer can then bring a short list of questions to engineering, including which features need a separate operation and whether a dimension on the drawing has been read correctly.
An approved tool connected to your price history or a market data feed can also retrieve candidate benchmarks. Record the source, date, units, and location before using one. India’s official Wholesale Price Index release, for example, can show the direction of movement in a relevant category. Confirm the actual price for the specific grade and stock form you intend to purchase.
Keep the raw extraction linked to the drawing. If a title block and a purchase specification give different material grades, flag the conflict and ask which document governs. Even a correctly extracted tolerance does not establish how the supplier will achieve or inspect it. Those are process decisions with their own cost implications.
Price updates cover only one kind of change. A new drawing revision, a different inspection requirement, or a smaller release can change the model even when rates stay steady. Have the buyer confirm the commercial inputs, and a manufacturing engineer review the proposed route and times. Keep the assumptions in editable cells and let a spreadsheet or costing engine calculate the totals.
Build a Should-Cost Model Before the RFQ
Consider a hypothetical steel mounting block made by computer numerical control (CNC) machining. The buyer expects 1,200 parts a year and is weighing releases of 50 and 200. Assume one setup per release, a nine-minute machining cycle, and the same material, tooling, and inspection requirements in both cases. All rates and times below are illustrative. The estimate includes packaging and excludes freight and taxes.
| Cost element | Calculation basis | 50 parts per setup | 200 parts per setup |
| Steel blank | 0.8 kg × ₹120/kg | ₹96 | ₹96 |
| Machining | 9 min × ₹600/hour ÷ 60 | ₹90 | ₹90 |
| Batch setup | ₹3,600 ÷ parts per setup | ₹72 | ₹18 |
| Tool wear | Allowance per part | ₹12 | ₹12 |
| Inspection and packing | Allowance per part | ₹22 | ₹22 |
| Additional overhead | Admin allowance outside the machine rate | ₹18 | ₹18 |
| Estimated cost per part | Sum of cost elements | ₹310 | ₹256 |
| Indicative selling price | Cost ÷ 0.90 for a 10% margin | ₹344.44 | ₹284.44 |
Illustrative costs per part. Both scenarios use the same annual demand and assume each release is machined in a single setup.
Even with the current steel price, the should-cost model can be significantly inaccurate if it assumes an incorrect number of parts per setup. Here, the change in batch size moves the indicative price by ₹60 per part. A 10% increase in the assumed steel price would move it by about ₹10.67, with the other inputs held constant. The buyer should spend more time testing the setup assumption in this example because it has more effect on the comparison.
Carry the assumptions into the RFQ. Send the drawing revision and annual forecast, specify the release quantity, and request suppliers to provide the basis for any setup charge. If they can run 200 at once and deliver 50 at a time, ask who holds the remaining stock and what that arrangement costs. If a supplier needs multiple setups to complete a release, ask how it has allowed for them.
Use the Model in Supplier Negotiations
Return to the ₹360 and ₹290 quotes. If both use the same price scope as the model, the first is about ₹16 above its 50-part benchmark, and the second about ₹6 above its 200-part benchmark. Those checks tell you where to probe in each bid. Ask both suppliers to price the same release schedule before comparing their offers. Confirm that both bids cover the same drawing revision, inspection scope, packaging, and delivery terms. If one charges separately for a fixture, put both bids on the same basis using the quantity you can reasonably commit to. For cross-border bids, also align currency, freight, and duties where applicable.
Keep the comparison at the quantity you can actually commit to. A lower unit price at 200 parts may be attractive, but buying four releases in advance changes how much cash and inventory are tied up. If the drawing is likely to change before all 200 parts are used, include that exposure in the decision.
Use the cost breakdown to ask specific questions. A supplier might allow more time because it plans to inspect every part, while the model assumes a sample. Ask engineering which inspection is required and request the same basis across bids. If a supplier can explain the gap with a credible process difference, update the model. If the gap is driven by the buyer’s release schedule, see whether production planning can change it before asking for a concession.
How to Evaluate Should-Cost Software
A spreadsheet is a sensible starting point when a few buyers and engineers can maintain the rates and review each part. Add an AI-assisted document tool if drawing extraction is the bottleneck. A dedicated costing platform is worth evaluating if the team repeatedly models similar parts across suppliers or regions and needs shared rate libraries, version history, and consistent review.
Test each option on the same approved drawing. Can the reviewer trace an extracted figure back to its source, make corrections, and observe how the price changes? Can the tool show which revision and rate date were used? Verify who can access drawings, how long the tool retains them, and whether you can export the breakdown. Test a revised drawing as well. For example, change the finish requirement and check whether the reviewer can see which inputs were carried forward and which ones need fresh validation. A fast estimate is useful only if another person can reproduce it.
What to Confirm Before You Send the RFQ
Confirm the drawing revision, expected release quantities, and which costs suppliers should show separately. A cost estimate becomes a better benchmark when its production assumptions match the order you intend to place. Use each unresolved assumption to write a question suppliers can answer in their bid.
Ready to see how your machining requirements translate into a supplier quote? Request a quote for precision-machined components from ZETWERK. Share your part drawings, 3D files, and production requirements, including how many parts you need per delivery.
Frequently Asked Questions (FAQs)
Does a small procurement team need dedicated should-cost software?
No. A controlled spreadsheet can be enough for occasional estimates if the team records rate sources, versions, and assumptions. Dedicated software is worth testing when repeated data entry or inconsistent models across buyers become costly.
Can you build a should-cost estimate before you have a drawing?
Yes. Use a comparable part or preliminary specification to create a broad range. List the design decisions that could materially change it, such as stock form, tolerances, or the need for a secondary operation.
Can you upload confidential drawings to an AI tool?
Only if your organisation authorises the tool for those drawings and the applicable confidentiality terms permit it. Check retention, access, and whether the provider uses uploads for training before submitting a file.
How do you estimate cost when a supplier’s machine rate is unknown?
Use a documented range from comparable equipment and locations, and test how the estimate changes across it. Check what the rate covers so that labour or overhead is counted once.
How does buyer-owned tooling affect the model?
If the buyer has already paid for the tooling, exclude its purchase cost from the recurring piece price. Include any maintenance, consumable, or replacement costs the supplier will actually bear, and confirm how those charges appear in the quote.
When should you refresh the estimate between RFQs?
Refresh it when a material input changes, such as the drawing revision, expected release size, exchange rate, or stock price. Review any input past its stated validity before using the model for a new sourcing decision.
What if the only qualified supplier quotes above the modelled range?
Examine the model’s scope and ask the supplier which assumptions explain the gap. If the higher price reflects a real constraint, update the estimate and assess whether a design or schedule change could reduce the cost without compromising qualification.




