How to Build a Should-Cost Model for a Food Ingredient (Without a McKinsey Engagement)

You don't need a McKinsey engagement or a 400-row cleansheet to build a should-cost model that wins a negotiation. You need a coarse structure that isolates the five cost levers that actually move the price, plus a defensible logic for each one. Build that in an afternoon and you'll walk into the supplier conversation knowing which numbers are real and which are padding.

This is the version senior buyers actually use — deliberately rough, built to be challenged out loud across a table, not admired in a spreadsheet. Here's how to build one for an ingredient you've never purchased before.

Key takeaways

• A should-cost model is a structure that tells you which lever moves the price — not a precise prediction of the supplier's cost.
• The usable version maps five levers: raw material, conversion, yield, freight/logistics, and margin. Everything else is noise.
• Precision is the enemy. A 40-line cleansheet built to two decimals never survives contact with a real negotiation. Coarse ranges you can defend out loud do.
• You can populate every input from free and observable sources — commodity indices, public freight benchmarks, industry conversion logic — without a paid subscription.
• The model's job is to convert into a negotiation position: index-linking, pass-through clauses, and a target you can anchor to.
• The most common failure is treating the output as a single "right number" instead of a range with a challenge attached to each line.

What is a should-cost model? (definition)

A should-cost model is a bottom-up estimate of what an ingredient should cost to produce and deliver, built from its underlying cost drivers rather than from the supplier's quoted price. Instead of accepting "the price is €2.40/kg," you reconstruct that €2.40 from the ground up: so much for the raw commodity, so much to convert it, so much lost to yield, so much to move it — and what's left is margin.

The point is not to arrive at the supplier's exact internal cost. You'll never have that, and you don't need it. The point is to know where the price comes from, so that when the supplier says "raw material went up 12%," you know whether raw material is the bulk of their cost (in which case a 12% move matters) or a quarter of it (in which case they're using a small input to justify a big increase).

A should-cost model is also called a cleansheet or should-cost analysis. Whatever the label, the function is the same: turn an opaque price into a structure of challengeable lines. For the formal definition and how it sits alongside related tools, see our should-cost model glossary entry.

Why precision is the enemy of a usable model

Here's what most buyers get wrong, and it's the single biggest reason should-cost work gets abandoned.

They build the model to be correct. They chase every input to the third decimal, model three energy scenarios, source utility rates by region, and add overhead-allocation lines no supplier will ever discuss. Three weeks later they have a beautiful 40-tab workbook — and they never open it in a negotiation, because the moment a supplier challenges one assumption the whole edifice wobbles and the buyer loses confidence in front of the room.

A should-cost model is a negotiation instrument, not an accounting document. Its accuracy requirement is directional, not absolute. You don't need to know that conversion is €0.31/kg. You need to know it's "roughly a quarter to a third of the raw-material cost, and it doesn't move when the commodity moves." That's a statement you can defend — and, more importantly, one that forces the supplier to argue on your terms.

The insider rule: build the model to the level of detail you can say out loud and defend. If you can't explain a line in one sentence to a CFO, delete it. The coarse model wins because it survives the conversation. The precise model loses because it invites a debate about your assumptions instead of theirs.

This is the same discipline you'd apply when you take over a new procurement category in your first 90 days — coarse and decision-grade beats precise and paralyzed, every time.

The 5-lever cost-driver map applied to an ingredient

Every food-ingredient price, no matter how complex the supply chain, decomposes into five levers. Map these and you've mapped the negotiation.

1. Raw material

The dominant lever for almost every commodity ingredient. This is the underlying agricultural or feedstock input — the crude oil before refining, the milk before it becomes whey, the cocoa bean before the press.

Your job is to (a) identify the reference commodity the price actually tracks, and (b) estimate what share of the finished price it represents. For most single-origin commodity ingredients, raw material is the largest single line — often the majority of total cost. When raw material dominates, your negotiation lever is index-linking, not haggling over a fixed price (more on that below).

2. Conversion

The cost to transform the raw commodity into the spec you're buying: pressing, refining, milling, drying, filtration, fractionation. Conversion is driven by energy, labor, and plant utilization — and, critically, it is largely independent of the commodity price. This is the lever suppliers most often hide inside a "market increase." If the bean went up but the processing cost didn't, a blanket price hike is padding the conversion line.

3. Yield

How much sellable product you get from a unit of raw input — and the cost of what's lost. A refining or extraction step at 85% yield carries a very different cost structure than one at 97%. Yield losses are sometimes recovered as by-product credits (the meal left after oil extraction, for example), which can quietly reduce the true raw-material burden. "What's the yield, and is the by-product sold?" is one of the sharpest questions a buyer can put to a supplier — because most don't expect the buyer to know to ask.

4. Freight and logistics

Moving the ingredient from plant to your dock: ocean or road freight, storage, handling, duties, and any temperature or hazmat requirements. This lever has grown far more volatile in recent years and is now a real negotiation surface in its own right. Two suppliers with identical conversion costs can land at very different delivered prices purely on logistics — which is also why this is the lever multi-sourcing exploits. (See when to dual-source and when not to.)

5. Margin

What's left after the four cost lines: the supplier's gross margin, overhead recovery, and profit. You'll never get this from the supplier directly, and you don't compute it — you back into it. Once you've estimated the other four levers, margin is the residual. If it looks fat relative to the value the supplier adds, you've found your opening. If it's thin, you know not to push on price and to look for value elsewhere (terms, volume commitments, spec flexibility).

The discipline of these five lines is what separates real cost-driver analysis for a commodity ingredient from a generic benchmark. A benchmark tells you the price is high. The lever map tells you why — and therefore what to ask for.

Where to find each input without paid data (illustrative ranges only)

You can populate a working model from observable, free sources. None of the figures below are specific to any one ingredient — they're the method for sourcing each line.

• Raw material: Public commodity indices and exchange-traded references (vegetable oils, cocoa, dairy, grains) give you the trend and the level for the underlying input. Even where the exact contract grade isn't quoted, a closely correlated reference is enough to track direction. Illustrative only: for many commodity ingredients, raw material lands somewhere in the broad range of half to three-quarters of delivered cost — but verify per ingredient; never assume.
• Conversion: Industry process descriptions, equipment-vendor literature, and trade-association material describe the processing steps and their relative intensity. You're estimating relative magnitude, not absolute cost — "refining is a low-single-digit fraction of raw material" is a usable, defensible statement.
• Yield: Technical and academic literature on extraction, refining, and milling yields is largely public. By-product markets (oilseed meal, whey permeate) are often quoted in the same commodity reports you're already reading.
• Freight/logistics: Public freight benchmarks and container-rate indices give you direction; lane-level estimates come from forwarder quotes, which cost nothing to request.
• Margin: Residual — never sourced, always derived.

A word of caution: the gap between "directionally right from free sources" and "decision-grade for a large contract" is exactly where a structured intelligence report earns its keep. The olive oil cost-drivers and hedging breakdown and the sesame oil North America procurement report exist because reconstructing the full lever map for a specific oil from scratch takes longer than most buyers have.

Arm yourself with the real cost-driver breakdown. Every Intel Report ($349) includes a structured cost-drivers tab that maps all five levers for a specific ingredient — reference commodity, conversion logic, yield, freight benchmarks, and where the margin really sits. It's the should-cost model already built, so you walk into the negotiation with the structure instead of spending a week reconstructing it. Browse the report catalog →

How to turn the model into a negotiation position

A model that stays in the spreadsheet is worthless. The conversion from analysis to leverage happens in three moves.

First, anchor on the dominant lever. Whichever line is largest is where you negotiate. If raw material is the bulk of cost, you don't argue about the supplier's overhead — you tie the price to the commodity and stop debating the rest.

Second, isolate the fixed lines. Conversion, much of logistics, and margin don't move with the commodity. When a supplier passes through a "market increase," your model lets you say: "Raw material moved, I accept that on the raw-material share. Conversion didn't move — why is the whole price up?" That single sentence, backed by structure, recovers more margin than a dozen rounds of generic pushback. This is the heart of how to negotiate price with should-cost data: you're not disputing the number, you're disputing which line it belongs to.

Third, convert the structure into contract mechanics (below).

Index-linking and pass-through clauses

When raw material dominates, the cleanest outcome isn't a lower fixed price — it's an index-linked price that moves the commodity portion (and only that portion) with a published reference, while conversion, logistics, and margin stay fixed. This protects you on the way up and guarantees you the benefit on the way down, which a fixed price never does.

A well-built pass-through clause specifies: the reference index, the exact share of price it applies to (your should-cost model gives you this number), the reset frequency, and a cap or collar if you want to limit volatility. The reason buyers can't write a tight pass-through clause is almost always that they never built the lever map — they don't know what share of the price is commodity, so they let the supplier index the whole price, including the fixed lines. That's a gift to the supplier. Your model closes it.

Common mistakes that make a should-cost model backfire

• Presenting it as "your real cost." Never tell a supplier you've calculated their cost — you'll be wrong on a line, they'll seize on it, and your credibility evaporates. Present it as your understanding of the cost structure and invite them to correct specific lines. The corrections are themselves intelligence.
• Over-building. The 40-line cleansheet you never open. If you spent more than a day on the first version, you modeled for accuracy instead of leverage.
• Indexing the whole price. Letting the commodity index move the fixed lines too. This single error can cost more than the entire negotiation was worth.
• Ignoring by-product credits. Treating gross raw-material cost as net when the supplier sells the by-product. You're negotiating against an inflated raw-material line.
• Single-number thinking. Treating the output as one "should-cost price" rather than a range with a challenge per line. The range is the model; the single number is a trap.
• Skipping the freight lever because "it's small." It often isn't anymore, and it's frequently the most negotiable line because it's the least sticky.

Worked mini-example: structuring an oil-based spec

Suppose you're handed a refined vegetable oil you've never bought. Here's the coarse model, built in an afternoon, no paid data.

1. Raw material. Identify the reference: crude oil of that type tracks a published vegetable-oil index. Assume — illustratively — it's the dominant lever, the clear majority of delivered cost. Move: index-link this share to the published reference.
2. Conversion (refining). Bleaching, deodorizing, filtration. Energy- and capital-driven, independent of the oil price; a modest fraction of raw material. Move: this line should not move when the commodity moves — hold it fixed.
3. Yield. Refining losses plus any recovered fractions. Ask whether the refining by-product is sold; if so, it offsets raw-material cost. Move: confirm net vs. gross raw-material basis.
4. Freight/logistics. Bulk liquid transport, tank storage, possible heated handling. Benchmark against public freight indices and one forwarder quote. Move: negotiate separately; consider a second source on a different lane.
5. Margin. The residual once the four lines are estimated. Move: if it looks generous for a commoditized refined oil, that's your opening; if thin, pivot to terms and volume.

You now have a one-page structure: one line to index, three to hold fixed, one residual to test. That is a should-cost model. It will not match the supplier's books to the cent — and it doesn't need to. It tells you which lever moves the price, and that is the entire job. For a fully reconstructed version of exactly this kind of oil model, see our olive oil cost-drivers and hedging deep dive.

FAQ

How accurate does a should-cost model need to be? Directionally accurate, not precise. You need to know which lever is largest and which lines move with the commodity versus stay fixed. A coarse model you can defend out loud beats a two-decimal model that collapses under one challenge.

Can I build a should-cost model without buying market data? Yes, for a first working version. Commodity indices, public freight benchmarks, and industry process literature cover the five levers well enough to negotiate. Paid intelligence earns its place when the contract is large enough that the gap between "directionally right" and "decision-grade" is worth real money.

What's the difference between a should-cost model and a cleansheet? They describe the same thing — a bottom-up reconstruction of cost from its drivers. In practice "cleansheet" often implies the over-detailed version. The usable should-cost model is deliberately coarser, built for the negotiation rather than the audit.

How do I use should-cost data in an actual negotiation? Anchor on the largest lever, isolate the lines that shouldn't move with the commodity, and convert the structure into an index-linked price with a pass-through clause that touches only the commodity share. Present it as your understanding of the structure, not as the supplier's confidential cost.

Which cost lever should I negotiate hardest on? Whichever is largest — usually raw material for a commodity ingredient, where the win is index-linking rather than a lower fixed price. Then attack the fixed lines (conversion, margin) whenever a supplier tries to pass a commodity move across the entire price.

Written by Amin Dabbech, founder of ProCure Navigators — 18 years in food-ingredient and packaging procurement. ProCure Navigators delivers decision-grade procurement intelligence reports for buyers and investors who need insider cost-driver detail without a six-figure consulting engagement.