How we make money
Open bookMost manufacturing margin lives in the dark: a markup inside the materials, a markup inside the freight, an office full of quoting and account management loaded invisibly into every unit. Moduloa's model removes all three hiding places — the OS does the administration, the BOM is costed line by line in the open, and logistics passes through at documented cost. What remains is margin you can point at. This page says exactly where it sits, in three published models — because a customer who can see how we earn is a customer who never has to wonder what they are really paying for.
Where the industry hides its margin — and what we do instead
Half the economics is transactions — and transactions are software
Contract manufacturing's admin layer looks small on the income statement and enormous where it matters. Read against the industry's thin gross margins, selling and administration consume 40–50 cents of every gross-profit dollar at the incumbents. Here, that layer is the OS.
The claim you sometimes hear — that contract manufacturers carry an admin layer around 20% — does not survive contact with their filings as a share of revenue: the real number is 2–6%. But that reading buries the story twice. Gross margins in this industry run 9–12%, so a 4%-of-revenue admin layer consumes 40–50% of every gross-profit dollar. And in the quotes themselves, where low- and mid-volume customers actually meet the number, the loading above direct cost runs 25–33%, with 10–15% markups stacked on materials before assembly is priced at all. The 20% is real — it just lives in the quote, not the annual report. Either way, it is the part of the price that pays for humans doing transactions.
| Company | Fiscal year | SG&A / revenue | SG&A / gross profit |
|---|---|---|---|
| Benchmark Electronics | FY2025 | 5.8% | 59.1% |
| Plexus | FY2025 | 4.9% | 49.1% |
| Jabil | FY2025 | 3.8% | 42.4% |
| Flex | FY2026 | 3.8% | 41.0% |
| Sanmina | FY2025 | 3.6% | 40.5% |
| NOTE AB | FY2025 | 3.9% | 27.8% |
| Fabrinet | FY2025 | 2.6% | 21.2% |
| Celestica | FY2025 | 2.1% | 17.4% |
| Median of the eight | — | 3.8% | 40.7% |
Every row comes from the companies' own audited statements — SEC filings for the US and Canadian reporters, the published year-end report for NOTE. Hon Hai/Foxconn, at roughly $250B of revenue, reports total operating expenses including R&D under 3% — scale makes the line vanish. Nordic peers Kitron, Scanfil, and GPV report by nature under IFRS and publish no SG&A line at all, so no number is cited for them here.
In a published should-cost teardown of an EMS program, direct labor — the humans actually assembling the product — was just under 2% of program revenue, while materials ran 75–85%. Nearly everything between the parts and the price is overhead, indirect labor, and margin: quoting, program management, procurement transactions, expediting, order administration. Manufacturing research named this forty years ago — the hidden factory, where overhead grows with the number of transactions, not the number of units. Transactions are exactly what software deletes: intake validates instead of a quoting team, the quote is computed instead of negotiated, the unit record reports instead of an account manager. What software cannot delete, we keep and pay for gladly: the quality system, compliance and audits, the machines and the people who fix them.
Because the unit price here is assembled in the open — attested BOM at supplier cost, published assembly rate, documented freight, one published fee — the classic loadings have no line to hide in. The materials markup is refused, the freight markup is passed through, and the admin loading that consumes half the incumbents' gross-profit dollar becomes a compute bill. The saving is structural, not a discount, and the landed-cost view makes it visible per unit: quoted at RFQ, reconciled at shipment. Two incumbents prove the direction is real — Celestica and Fabrinet run the leanest admin in the table and post operating margins near the top of it. They are the floor we benchmark against, not the ceiling we hide behind.
What a unit actually costs you
One landed number, built from parts you can audit — quoted at RFQ with its confidence attached, reconciled at shipment against real invoices, variance on the record.
Two honesty mechanisms hold this together. The quote's confidence number covers the whole stack — attested lines versus estimated ones, the labor model's maturity, the freight snapshot's age — and quoted-versus-actual variance is tracked per product, because a landed cost you never true up is a guess wearing a suit. And freight reality is named, not hidden: quotes carry validity windows, and beyond them prices index to a public freight benchmark. Ocean rates have moved fourteen-fold in a single cycle; anyone promising fixed landed cost through that is selling you their own bankruptcy.
Pick the model — all three are published
Same OS, same transparency, three shapes of commitment. The middle one is the thesis expressed as a price.
Classic, kept honest
A unit price with our margin in it — recalculated openly after the pilot under the published terms: undercalculation is ours to absorb, overcalculation is refunded. Some units we win, some we lose; the trend is the OS's problem to fix, and the numbers are yours to see. For one-time builds and first engagements.
Production as a membership
Units at documented cost plus a flat, published handling rate — and a yearly subscription that buys what nobody else sells: reserved capacity with a two-week order-to-ship promise, at any hub your product is qualified for. The margin lives in the membership, not the unit. Wholesale proved this shape retains customers for decades; the chip industry signs it as billion-dollar capacity agreements; cloud made it the default pricing of computing. Ours is the same contract, for physical production.
How the network earns
The long game from the thesis: hubs pay for certification and the factory OS, deployments of a blueprint to a new hub carry a fee, and routed production carries a network transaction fee. This is the revenue that scales without owning every factory — the framework earning as the framework, exactly as the thesis prices it.
The subscription is designed against its own known failure modes, because they are documented. Capacity is never oversold — at most 85% of a hub's nameplate is subscribable. Half the fee is deductible against handling on units you actually ship; unused reserved units roll over one quarter; you can resize ±20% at renewal. The two-week promise penalizes us, not you — misses convert to service credits. And the "any qualified hub" right comes with one second-hub qualification included, because two chained hubs capture nearly all the protection of many. Customers who pay for capacity deserve mechanics that make unused capacity survivable — the alternative is the resentment economics of gym memberships, and we have read that literature too.
Reserved capacity only works if demand is honest in both directions. Every production customer gets a live forecast they can see, comment on, and correct — the model learns from their comments and from actual orders, and the contract binds both sides to it. Lean hubs are not an efficiency slogan; they are what makes the subscription price defensible.
What has to be proven
Model B is conditional, and the conditions are published: it needs at least two certified hubs with chained qualifications before "any hub" is a promise rather than a poster; it needs 12–24 months of utilization data before hard take-or-pay pricing — until then the fee is anchored to the cost of capacity actually held; and it needs a written definition of "cost" (attested BOM plus the published rate card, with audit rights) or the model's best sentence becomes its biggest dispute. The known failure modes have names: the chip industry's capacity fees were renegotiated in the last downturn, and roughly a third of cloud commitments go to waste — the deductibility, rollover, and resize mechanics above exist precisely because we studied how this dies. Logistics ownership carries the one problem software cannot fix: being importer of record is entities, bonds, and real liability, per country — legal infrastructure, built hub by hub. All of it lands in the register as dated claims when the first contract is signed.
Where these numbers came from
Turnkey vs consigned EMS — materials markup practice · Freight forwarder margin structure · Landed-cost practice — the unclosed loop · Freightos — freight-rate volatility · Buyer's consolidation — the savings mechanics · GlobalFoundries–Cirrus capacity reservation agreement (SEC) · AWS Reserved Instances — commitment pricing · Costco — margin in the membership · Flexera — cloud commitment waste · DellaVigna & Malmendier — the gym-membership problem · Cachon — push, pull, and advance-purchase risk sharing · Jordan & Graves — chained flexibility · Amazon FBA — published logistics fee precedent · Optimum Design — the cost-plus quote model · Zetter — what the EMS quote is not telling you · VentureOutsource — should-cost: direct labor under 2% · Miller & Vollmann — The Hidden Factory (HBR, 1985)