Should-cost modeling is the practice of building a bottom-up estimate of what a product ought to cost to manufacture — from component market pricing, conversion rates, tooling amortization, and overhead — before you ever receive a supplier's price. Building it first is what turns a quote from a verdict into a conversation.
A founder I'll call Priya runs a connected-lock company in Seattle. In November she sent three suppliers the same package: a 42-line BOM, a 3D model, and a request for 20,000 units. The quotes came back in the first week of December, and they clustered beautifully — $30.40, $31.80, and $32.15. Tight clustering felt like validation. She called me, half-celebrating, and said the line I've learned to distrust: "The market has spoken."
The market had not spoken. Three suppliers had read the same BOM, applied the same default assumptions, and quoted the same comfortable margin. When we rebuilt her cost model from the bottom up, the number came out at $24.10. That is a 27% gap, and on 20,000 units it is roughly $150,000 — sitting in her margin, invisible, because nobody had ever asked what the product should cost.
The panic came later, when she realised she'd almost signed. Not because the suppliers were dishonest — none of them were. They had quoted honestly against a package that invited conservative assumptions: a BOM with eight unspecified parts, a housing with cosmetic requirements written as "premium finish," and no stated test flow. They priced the ambiguity. She had handed it to them.
At SciPure, our founder came out of the Foxconn-Apple joint lab — two decades of taking consumer electronics and home products from a concept on a whiteboard to tens of millions of identical units. Between us we hold 29 granted patents, and our founder was named a Shenzhen science-and-technology leadership award winner in 2020. We are a Shenzhen-based product engineering advisor: we help consumer-electronics and home brands manage the full journey from concept design to mass-production management, including DFM optimization, injection molding, electronics assembly, QC systems, and supply chain development — backed by our own 2,000 m² production facility in Shenzhen's Guangming District (ISO 9001, AQL 1.0). We have run 150+ product programs since 2024, and should-cost modeling is the first thing we build, because it is the cheapest lever a hardware team has.
Why a Black-Box Quote Is a Negotiation You Already Lost
The reason to model first is simple: a single all-inclusive unit price gives you nothing to argue with. The conclusion to hold onto is that you cannot negotiate a number you cannot decompose.
- A quote that lists only a per-unit price hides every cost driver at once — you cannot tell whether an expensive component, an inflated labor estimate, or amortized NRE is what's inflating it.
- When NRE is quietly folded into unit price, your cost of goods is artificially inflated and your true gross margin is invisible, which makes financial modeling nearly impossible.
- Tight clustering across three suppliers is usually not market validation. All three may simply be pricing the same ambiguity with the same default margin.
- A slightly higher but fully itemized quote is often the better long-term deal, because it gives you a shared basis for cost reduction later.
The Four Buckets Every Should-Cost Model Needs
A usable model has four buckets and nothing else. The conclusion: if you can name all four and defend each assumption, you know more about your product's cost than most first-time founders ever do.
- Landed BOM cost — the sum of component pricing at your actual volume tier, plus a scrap allowance (around 2%) and a logistics buffer (around 8%).
- Conversion cost — assembly, test, and packing labor plus the overhead absorbed against it. Practitioner benchmarks put this at roughly 25–35% of landed BOM for many consumer-electronics builds; a 30% heuristic is a reasonable starting point.
- CM markup — the contract manufacturer's overhead and profit, typically in the 15–30% range depending on product complexity and volume.
- NRE and tooling — kept strictly separate as one-time capital, then amortized. If your NRE is $50,000 and your unit cost is $68.64, your break-even on that investment is 729 units.
Worked example, using the standard flow: a $40.00 raw BOM, plus 2% scrap ($0.80) and 8% logistics ($3.20) gives a landed BOM of $44.00; conversion at 30% adds $13.20 for a subtotal of $57.20; a 20% CM markup adds $11.44, for an estimated should-cost of $68.64 per unit.
How to Benchmark Components Without a Purchasing Department
Components are where the model earns its keep, and where the data is most accessible. The conclusion: electronic component pricing is far more transparent than most founders assume, and you do not need a procurement team to access it.
- Electronic components typically represent 50–70% of total PCBA cost and 40–60% of finished product cost, so this bucket dominates the outcome.
- Pull distributor pricing from DigiKey, Mouser, Arrow or Avnet at your expected volume tier — then remember that online prices sit above production-volume pricing, and a common correction is to reduce distributor pricing by roughly 20% when estimating volume cost.
- Aggregators matter: Octopart searches across 600+ distributors and returns consolidated pricing, stock levels, and lifecycle status, with daily data updates.
- Volume break structure is the single biggest variable. Component costs can drop 30–50% between 1,000-unit and 100,000-unit quantities, so always match the model to real production volume.
- For a fast sanity check on labor: count SMT placements from the BOM and multiply by roughly $0.10 at mid-volume — the result typically lands at 8–10% of material cost. Add about $0.05 per through-hole component when there are 1–15 of them.
- Not every BOM line can be modeled the same way. Commodity passives have clean market prices; custom mechanicals need material + cycle time + tooling amortization; regulated items like battery packs and RF modules need certification cost and approved-vendor status baked in.
The Sensitivity Layer: Where the Real Decisions Happen
A single unit-cost number is less useful than knowing which variables move it. The conclusion: the value of a model is not the number it produces, but the ranked list of risks it hands you.
- Tariff exposure is usually the largest single swing factor. Moving a Section 301 rate from 0% to 25% on a $19.47 base-case product adds roughly $4.87 per unit — more than yield, test time, and freight combined.
- PCBA yield sliding from 96% to 90% adds about $0.42 per unit, and test time climbing from 60 to 150 seconds adds about $0.55.
- Air-freight share moving from 10% to 60% adds roughly $0.85 per unit — which is why freight belongs in the model, not in the shipping department.
- EMS markup widening from 6% to 10% adds about $0.48 per unit, which is the number to raise if you suspect your CM is marking up components above market.
- EMS component markup typically runs 3–12%, depending on provider scale and your volume. Tier-1 providers can often buy 5–15% below what a mid-size OEM accesses directly and retain part of that spread; smaller contract manufacturers tend to mark up 5–12%.
- Memory pricing is cyclical and can swing dramatically in a single quarter — the memory market entering 2026 is in an upswing driven by AI-related demand. Timestamp your model and refresh it at least quarterly.
Use this decision tree to decide how deep to go:
- If you are pre-tooling with one product and no sourcing hire → model the BOM and conversion roughly, get the sensitivity table right, and use it to ask better questions. Cost of skipping: a 25% margin leak you never see.
- If you are comparing two suppliers or two countries → force every quote onto identical assumptions first (same Incoterms, same volume, same test flow, NRE separated). Mismatched assumptions produce arguments, not decisions.
- If you are past first production and scaling → maintain the model as a living artifact tied to BOM revisions, and track purchase price variance per line so every variance maps to a specific part, quote, and revision.
Reading the Quote: Green Flags and Red Flags
How a supplier presents a number tells you how they will behave as a partner. The conclusion: you are looking for a collaborator who is confident enough to be transparent, not the lowest number in the stack.
- Green flag: an itemized quote separating BOM, conversion (often broken out by assembly, test, and packing), and markup — with NRE and tooling listed as distinct one-time charges on a separate schedule.
- Red flag: a single all-inclusive per-unit price, with no visibility into drivers and no separation of NRE.
- Red flag: resistance to discussing any line item. A partner who won't explain a number is a partner who intends to keep moving it.
- Green flag: willingness to requote specific lines when you bring data. The best suppliers, in our experience, treat a competent customer's cost model as a relief rather than a threat.
What the Teams That Negotiate Well Do Differently
The teams that consistently land good pricing are not the most aggressive. They are the best prepared. At Tesla, should-cost models were used not as weapons but as shared tools — when a supplier understood that the cost structure was known, the conversation shifted from price negotiation to joint cost reduction, and the savings were larger and more durable. Apple's teardown-driven component benchmarking is the industry's most disciplined version of the same instinct. Anker's margin discipline comes from treating unit cost as a design variable, not a procurement outcome.
The habit they share is cheap to copy: they know their number before they ask for anyone else's. In 2026 that is easier than it has ever been — distributors publish tiered pricing publicly, aggregators price an entire BOM in seconds, and teardown databases let you benchmark against what a competitor's comparable display actually costs. But no tool will build the model for you, and no tool can tell you which assumption in it is wrong. That is engineering judgment, and it is worth roughly 27% of your margin.
"Three suppliers quoted within two dollars of each other and we took it as proof we were priced fairly," a founder at a Seattle connected-hardware company told us. "SciPure's model said we were 27% high. Walking that back took one afternoon of questions — and it paid for the entire tooling program."
The Principle
The anti-intuitive truth most founders resist: the supplier's first number is not a price, it is an opening position — and the only thing that makes it negotiable is a number of your own.
The engineering proverb we work by: if you can't decompose it, you can't negotiate it. Every dollar you can't name a driver for is a dollar you've agreed to pay twice.
If you have a product concept and want to know if it's manufacturable, send a written brief at www.scipure.net/inquiry.html. We'll review your design and tell you exactly where the risks are.
Frequently Asked Questions
What is should-cost modeling in consumer electronics?
Should-cost modeling is a bottom-up estimate of what a product should cost to manufacture, built from component market pricing, assembly process rates, tooling amortization, and overhead estimates. It differs from traditional manufacturing costing because consumer electronics relies heavily on distributor benchmarks and market pricing for electronic components, which typically represent 40–60% of finished product cost. The output is not just a unit number but a ranked list of the variables that move it.
How much markup do contract manufacturers typically apply?
CM markup for overhead and profit typically falls in the 15–30% range, depending on product complexity and production volume. Separately, EMS providers mark up components by roughly 3–12% — tier-1 providers can often source 5–15% below what a mid-size OEM accesses directly and retain part of that spread, while smaller contract manufacturers tend to mark up 5–12%. Quoting the same BOM as both turnkey and consigned reveals whether your partner is competitive on buying, conversion, or both.
Why do three suppliers sometimes quote nearly the same price?
Tight clustering is often mistaken for market validation, but it usually means all three read the same ambiguous package and applied the same conservative default assumptions. If your BOM has unspecified parts, undefined cosmetic standards, and no stated test flow, suppliers will price that ambiguity. Tight quotes are a signal to build your own model, not to stop asking questions.
What is the biggest single variable in a component cost model?
Volume break structure. Component costs can drop 30–50% between 1,000-unit and 100,000-unit quantities, so a model priced at the wrong tier will be badly wrong overall. The second-biggest is memory pricing, which is cyclical and can swing sharply in a single quarter — the memory market entering 2026 is in an upswing driven by AI-related demand, so models should be timestamped and refreshed at least quarterly.
Should NRE and tooling be included in unit cost?
No — keep them separate as one-time capital and amortize them deliberately. Folding NRE into unit price artificially inflates your cost of goods and hides your true gross margin. As a planning figure, if NRE is $50,000 and unit cost is $68.64, the break-even on that investment is 729 units. That break-even number is what belongs in your financial projections, not the amortized-per-unit figure.