Deep diveOperations

Should-cost analysis: how to build and use a cleansheet cost model

A should-cost model estimates what a part or service ought to cost by rebuilding it from materials, process time, labor, overhead, logistics and a fair margin. Built well, it moves a negotiation from percentages to specifics: which assumption explains the gap between quote and model. This deep dive covers when the effort pays, the cost layers, inputs, validation, sensitivity and index clauses, and where AI helps.

Reviewed 7 min read

On this page
  1. When a should-cost model is worth building
  2. The layers of a cleansheet cost
  3. Terms a should-cost model is built from
  4. Should-cost inputs and where to find them
  5. Building and validating the model
  6. Negotiating with a cost model without damaging the relationship
  7. Where AI helps a cost model and where it misleads
  8. A hypothetical machined aluminum housing
  9. Questions and answers
  10. Sources

When a should-cost model is worth building

Cleansheet costing takes engineering time. Spend it where the market cannot tell you the price.

  • If

    The part is engineered to your drawing, spend is high and only a few suppliers can make it.

    Then

    Build a full should-cost model.

    Competition alone will not reveal a fair price, and the stakes justify the effort.

  • If

    The item is a commodity or catalog part with many sellers.

    Then

    Rely on competitive quotes and market prices.

    The market already prices it better than a model would.

  • If

    Most of the price is one raw material that trades on a published index.

    Then

    Model conversion and margin, and link the material portion to the index.

    Negotiating a material cost that moves with the market wastes effort on both sides.

  • If

    The part is still in design.

    Then

    Use a simplified model to compare design options before suppliers quote.

    Cost is easiest to change before tooling and specifications are fixed.

The layers of a cleansheet cost

Supplier margin01Logistics and packaging02Overhead and tooling03Direct labor04Conversion05Material06
  1. Supplier margin

    Profit on full cost, judged against the supplier's risk, volume and investment.

  2. Logistics and packaging

    Freight, packaging, duties and handling to the delivery point.

  3. Overhead and tooling

    Plant and administrative costs allocated to the part, plus tooling amortization.

  4. Direct labor

    Operator time per part times a regional, fully loaded labor rate.

  5. Conversion

    Machine time per operation times a machine-hour rate.

  6. Material

    Gross material per part, including scrap and yield loss, at index or contract prices.

Conceptual stack of a should-cost model, from material at the base to margin at the top. The relative size of each layer varies widely by part and process.

Terms a should-cost model is built from

Process routing
The sequence of operations that turns raw material into the finished part, such as sawing, milling, deburring, anodizing and inspection, each with its own machine and time.
Cycle time
Time per part on one operation, including loading and unloading. Usually the most contested input in the model.
Machine-hour rate
The cost of running a machine for an hour: depreciation, financing, floor space, energy and maintenance, divided by the hours it is realistically used.
Scrap and yield
Scrap is material removed in machining or lost to rejects; yield is the share of good parts. Both raise cost per good part.
Overhead absorption
How indirect costs such as supervision, quality and administration are spread across parts, commonly as a rate per machine or labor hour.
Index-linked pricing clause
A contract term that adjusts the material portion of the price when a named published index moves, within agreed bands and review dates.

Should-cost inputs and where to find them

InputTypical sourceWatch-out
Drawings and specificationsEngineering, the product lifecycle system, supplier quotesTolerances and finishes drive process steps; an outdated revision invalidates the routing
Material pricesPublished commodity indices such as metal exchange prices, plus regional premiums and contract pricesIndex prices exclude premiums, alloy surcharges and the supplier's own buying terms
Labor ratesOfficial statistics such as US Bureau of Labor Statistics occupational wage data2 and Eurostat hourly labor cost statistics3Statistics give wages or average costs, not fully loaded shop rates; adjust for benefits, shifts and productivity
Machine ratesEquipment prices, depreciation policy, energy tariffs, utilization assumptionsA low assumed utilization inflates the rate; agree a realistic figure with engineering
Overhead and marginSupplier financial statements, industry norms, past negotiationsMargins differ legitimately with volume, risk and engineering support

Building and validating the model

  1. Lay out the routing

    With a manufacturing engineer, list every operation from raw stock to packed part, the machine type for each and any outside processing such as heat treatment or coating.

    Output
    Process routing

  2. Estimate times and material

    Estimate cycle times from feature counts, material removal rates or time studies, and gross material from the stock size before machining. Apply scrap and yield assumptions per operation.

    Output
    Time and material sheet
    Owner
    Cost engineer

  3. Price each layer

    Apply machine-hour and labor rates for the supplier's region, add overhead, tooling amortization, packaging and freight, then a margin consistent with the supplier's risk.

    Output
    Layered cost estimate

  4. Validate internally

    Review the model with engineers who know the process and buyers who know the supplier. Challenge every assumption that carries a large share of cost before anyone outside sees it.

    Output
    Reviewed model with assumptions log

  5. Run sensitivities

    Vary the assumptions that matter most, typically cycle time, utilization, material price and yield, and record how the total moves. This shows which questions to ask first and which differences are noise.

    Output
    Sensitivity table

  6. Test with the supplier

    Share the structure and the assumptions in question, and ask where the supplier's process differs. Update the model when they reveal a real constraint, such as an operation your drawing forces.

    Output
    Agreed or explained gaps
    Owner
    Category manager

Negotiating with a cost model without damaging the relationship

A should-cost model is a tool for asking better questions. Opening with a demand to match the model's total invites a defensive answer. Opening with a specific difference, such as 'our routing has three operations and your quote implies four; what are we missing?', invites the supplier to explain or reconsider.

Sometimes the explanation is legitimate: a tolerance that forces a slower process, a certification cost, low volumes. Those findings are valuable too, because they point to design or specification changes that cut cost for both sides. Where material dominates, an index-linked clause takes the most volatile element out of the annual negotiation. Contract optimization is part of ColdAI's procurement transformation offering1, and should-cost work usually feeds it.

Where AI helps a cost model and where it misleads

Misread drawings

Early signalFeatures and tolerances extracted automatically from drawings flow into the model unchecked.

MitigationUse AI extraction for a fast first pass, then have an engineer confirm the features that drive the routing.

Mismatched index data

Early signalMaterial prices pulled from feeds without checking grade, region or premium.

MitigationLet automation track indices and flag movements, but map each part to the right grade and premium once, by hand.

Plausible but invented inputs

Early signalA language model proposes cycle times or rates with no stated basis.

MitigationAccept no time or rate without a traceable basis: a time study, a removal-rate calculation or a confirmed supplier figure.

An opaque model

Early signalThe team cannot explain where a number came from when the supplier asks.

MitigationKeep an assumptions log with a source for every input, so each figure can be defended or corrected.

A hypothetical machined aluminum housing

Questions and answers

How long does it take to build a should-cost model?

It depends on part complexity and data availability more than on tools. A simple part with a short routing can be modeled quickly once drawings and rates are at hand; a complex assembly with many purchased components takes much longer. Start with a few high-spend parts and build a reusable library of machine and labor rates, and later models come faster.

Should we share the should-cost model with the supplier?

Share the structure and the specific assumptions you want to discuss rather than handing over the whole model at the outset. Suppliers respond better to targeted questions than to a total they are told to match. As trust builds, some organizations work through models openly with strategic suppliers, which can surface design changes that reduce cost for both parties.

What margin should a should-cost model assume?

One consistent with the supplier's risk, volume, investment and market conditions, informed by published financial statements where available and by past negotiations. Margin is rarely the best place to negotiate: a supplier that cannot earn a fair return will underinvest or walk away. Gaps in conversion cost, scrap or overhead allocation are usually more productive to discuss.

Does should-cost modeling work for services as well as parts?

Yes, with a different structure. A service model rebuilds cost from roles, hours, loaded rates by location, tools and overheads, plus margin. It helps where pricing is opaque, such as facilities, engineering or outsourced back-office services. The same disciplines apply: a traceable basis for every input, internal validation first and conversations focused on specific differences.

Sources

  1. Operations capability: procurement transformation and contract optimization — ColdAI
  2. Occupational Employment and Wage Statistics — US Bureau of Labor Statistics · checked 10 October 2026
  3. Hourly labour costs (Statistics Explained) — Eurostat · checked 10 October 2026

More in Operations

Back to Operations

Next step

Pick one high-spend part and test the method on it

Send a description of the part, its annual volume and the quotes you hold. We will reply with whether a should-cost model is worth building for it and which inputs your team would need to gather.

Discuss a should-cost model