Should-cost analysis: building a cost model before you negotiate

Sourcing & negotiation · By the Buyer's Ledger editorial team · Published 2026-10-31

Most negotiations start from the supplier's price. A should-cost analysis starts somewhere else: from an independent estimate of what the product or service ought to cost to make and deliver, built up from its materials, labour, overheads and a reasonable margin. The gap between that estimate and the quoted price is where the conversation becomes specific. This guide explains when the effort pays off, how to build a model, and the mistakes that turn a useful tool into a source of false confidence.

What a should-cost model is, and what it is not

A should-cost model is a structured estimate of the cost of a requirement, assembled bottom-up from cost drivers rather than taken from past prices. For a machined part, that means material weight and grade, machine time, set-up, scrap, labour rates, energy, packaging, freight, overhead and profit. For an outsourced service, it means staffing levels and grades, wage rates and on-costs, management layers, equipment, facilities, mobilisation and margin.

It is related to, but different from, two other tools on this site:

Price analysis compares a quote with other quotes, past prices or market indices. It tells you whether a price is in line with the market, not why the market is where it is.
Total cost of ownership adds up everything the buyer pays over the life of an asset: purchase price, installation, running costs, maintenance and disposal. Should-cost looks inside the purchase price itself, at the supplier's cost structure.

When it is worth the effort

A good model takes time, data and often engineering or finance input. It is worth building when several of these apply:

the spend is large or recurring, so a small percentage matters;
competition is weak: a single or sole source, a framework call-off with few suppliers, or bids that cluster suspiciously close together;
the cost structure is knowable, with identifiable materials, processes or staffing that can be priced from public or internal data;
you are deciding whether to make or buy, or to insource or outsource, and need a fair comparison of the two;
a supplier is asking for a price increase and you need to test the claim against what actually moved.

It is rarely worth it for low-value commodity purchases, where a competitive request for quotation already reveals the market price.

Building the model, step by step

1. Define the requirement precisely

The model can only be as specific as the specification. Fix the volumes, quality levels, service hours, locations and contract length you are modelling, and write them down. A should-cost estimate that silently assumes a different scope from the supplier's quote will produce a gap that is pure misunderstanding.

2. Break the cost into its drivers

Draw the cost tree. For goods, a common structure is: direct materials, direct labour, manufacturing overhead (machines, energy, maintenance, depreciation), scrap and yield losses, packaging and logistics, selling and general overhead, and profit. For services: productive staff hours, wage rates, statutory and contractual on-costs, supervision and management, consumables and equipment, premises, IT, mobilisation and transition, risk allowance, and profit. Each branch should end in a quantity multiplied by a rate.

3. Find the data

Quantities usually come from the specification, from engineering estimates (weights, cycle times, staffing rosters) or from teardown of an existing product. Rates come from a mix of sources: published commodity prices and price indices, official statistics on wages and labour costs for the relevant country and sector, energy tariffs, freight quotations, your own historical data, and cost breakdowns that suppliers provided in earlier tenders. Record the source and date of every input in a log; an undocumented model cannot be defended in a negotiation or an audit.

4. Make the assumptions explicit

Every model rests on choices: the plant's utilisation, the scrap rate, the productivity of a cleaning team, the overhead rate, a fair margin for the sector. Put them on a single assumptions sheet where they can be challenged and changed, rather than burying them in formulas.

5. Test the sensitivity

Change one input at a time and see which ones move the total most. Usually a handful of drivers (material price, labour hours, utilisation) explain most of the cost. Those are the ones to verify carefully and to discuss with suppliers. Sensitivity analysis also gives you a range rather than a single figure, which is more honest and more useful at the table.

6. Review and update

Have someone else check the logic and arithmetic. After award, the model can be updated with contracted figures and used to assess change requests and index-linked price reviews during contract management.

Using the model in a tender and a negotiation

The point of a should-cost model is not to prove a supplier wrong. It is to ask better questions. When a quote is well above the estimate, ask the supplier to break its price down along the same structure and compare line by line. The gap often turns out to be a legitimate difference in scope or risk you had not modelled, a cost driver you misjudged, or a genuine inefficiency or margin you can negotiate. When a quote is well below the estimate, that is a warning too: the supplier may have misunderstood the requirement, priced to win and recover later through change orders, or be unable to deliver at that price.

A few rules keep the exercise productive:

Lead with questions, not your number. Revealing your estimate early anchors the discussion on your assumptions rather than the supplier's facts. As our guide to negotiation basics explains, information is a concession; trade it deliberately.
Negotiate the drivers, not the total. Agreeing that machine time is overstated, or that a management layer is unnecessary, produces savings that survive; a bare demand for a lower price invites cuts in quality.
Accept that the supplier knows its costs better than you. Your model is an informed estimate. If the supplier shows evidence that an input is wrong, update the model.
Look for joint improvements. Should-cost work often reveals specification choices, such as needlessly tight tolerances or unused service hours, whose removal saves more than any margin negotiation.

How public buyers use the method

Two well-documented public frameworks show the method at scale.

In the United Kingdom, the Cabinet Office's Should Cost Modelling guidance note (May 2021), written for central government under the Sourcing Playbook and the Construction Playbook, describes a model as a forecast of what a service, project or programme should cost over its whole life. It treats producing one as good practice for all procurements, to be agreed during planning before the contract is advertised, and sets out five stages: plan, design, develop, test and use. Two points in it are worth borrowing in any organisation. First, the model is normally not shared with bidders during negotiation; it informs the buyer's position and its view of whether bids are deliverable. Second, it can only be used as a formal evaluation criterion if it has been disclosed to bidders, with the scoring method set out in the procurement documents. The note also links the model to the referral of abnormally low bids in complex procurements, using a comparison with both the other bids and the should-cost estimate.

In the United States, the Federal Acquisition Regulation describes should-cost reviews at FAR 15.407-4 as a specialised form of cost analysis that does not assume a contractor's historical costs reflect efficient and economical operation. Instead, a multi-functional government team of contracting, pricing, audit and engineering staff evaluates the contractor's workforce, methods, materials, equipment and management, and quantifies recommended improvements so that realistic negotiation objectives can be set. The regulation distinguishes a program review, focused on a particular contract, from an overhead review of the contractor's indirect costs. The core idea transfers to any buyer: an efficient cost, not a historical one, is the benchmark.

Common mistakes

False precision. A model that shows a cost to the cent from rough inputs invites misplaced confidence. Show ranges and round sensibly.
Ignoring the supplier's real risks. Warranty exposure, demand uncertainty, payment terms and financing costs are real costs. Leaving them out makes every quote look inflated.
Using a zero or token margin. Suppliers need a profit to stay in business and invest. A should-cost estimate that leaves none is not a target; it is a recipe for a failing supplier.
Stale data. Commodity prices and wage rates move. Date every input and refresh the key drivers before each negotiation round.

A minimal checklist

Scope, volumes and contract term written down and matching the tender documents.
Cost tree agreed with engineering or operations colleagues.
Every input linked to a dated source in a data log.
Assumptions on one sheet, reviewed by someone else.
Sensitivity analysis identifying the top cost drivers.
A decision on whether the model will be disclosed to bidders, made before the tender is published.
A plan for updating the model after award.