← Journal
MOD-01 · Field notes

When the rate becomes the margin

Research

A part priced at USD 1,200 and quoted as EUR 1,200 carries a 13.80% uplift at the ECB reference rate for 29 July 2026 — 19.74% at the euro’s high this year. Conversion is a real cost, and it is nowhere near that: 0.90% at a published bank margin, 1.6% all-in by the FSB’s own measurement. This note separates the two, shows why a costed BOM line looks identical whichever produced it, and proposes nine rules that make the difference checkable from both sides of the table.

2026-07-30 · Field notes · 9 min read · By

This started as a section on how we make money and outgrew it. The short version lives there as one line among four; the arithmetic, the mechanisms, and the convention live here.

The arithmetic

Two numbers that are not the same number

13.80% Uplift from quoting USD 1,200 as EUR 1,200 at the ECB reference rate of 29 July 2026 (1.1380)
19.74% The same treatment at the euro's 2026 high, 28 January (1.1974) — no clever stack required
0.90% A published bank FX margin per conversion — what carrying a currency actually costs
1.6% FSB-measured all-in cost of a cross-border business payment, 2024 — against a G20 target of 1%
Conversion is a real cost. It is not this cost.

Somebody has to carry currency risk, and moving money between currencies is genuinely not free. Nordea publishes an FX margin of 0.90% per conversion for the euro and eight other European currencies. The Financial Stability Board measured the all-in cost of a cross-border business payment at 1.6% globally in 2024. Non-financial companies paid an average effective spread of about 0.06% on EUR/USD forward contracts. Every one of those is legitimate, disclosable, and small. None of them is 13.80%. The problem is not the price of currency — it is that a BOM line arrives carrying a number and no rate, no date, no source, and no separate charge, so nobody downstream can tell which of these produced it.

The mechanisms

Six ways the rate becomes margin

01 · The rounded rate Quoting a dollar figure as the identical number of euros is arithmetically the same act as rounding the rate down to 1.00. Rounding the rate is vastly more powerful than rounding the amount: against the true EUR 1,054.48, a rate of 1.10 yields EUR 1,090.91 (+3.46%), 1.05 yields EUR 1,142.86 (+8.38%), and 1.00 yields EUR 1,200.00 (+13.80%) — while rounding the euro amount itself up to EUR 1,055 moves the line 0.05%. Rounding the rate to a whole number is roughly 280 times as powerful as rounding the result.
02 · The stale rate This one requires no decision at all — only the absence of one. A rate is fixed once in a contract schedule and nobody refreshes it. A rate set at the 2025 low of 1.0198 and still applied on 29 July 2026 embeds 11.59% before anyone adds anything. It is arithmetically symmetric — a rate fixed at 1.1974 and applied at 1.1340 makes the customer underpay by 5.29% — but the information is not symmetric: usually only one party is watching the drift, and only one party controls the refresh. For scale, the median month between 2024 and mid-2026 moved 1.98% high-to-low; April 2025 moved 6.38%.
03 · The stacked spread A part listed in dollars, bought by a sub-supplier booking in a third currency, purchased by an assembler in euro and invoiced in kroner can involve three conversions. Compounded at a published 0.90% margin, three hops cost 2.72% — a documented floor, not a worst case. The catch matters as much as the number: a supplier with multi-currency accounts or a group treasury may convert once, or never, while still pricing as though every leg carried a spread. From a BOM line you cannot tell whether three spreads were incurred or three spreads were merely charged.
04 · The buffer inside the markup base Where an FX allowance sits inside the base a materials markup is applied to, the markup earns on the allowance — the gap is exactly base × buffer × markup. Two structural consequences follow, and both are invisible unless the allowance is its own line: margin gets earned on the act of bearing currency risk rather than on the risk itself, and in a fixed-price line there is no mechanism that returns an unspent allowance when the adverse move never comes. AACE International's recommended practice says to segregate currency exchange from escalation, contingency, and allowance, and to document each definition in the basis of estimate.
05 · "The official rate" is not a specification Two official EU-published euro rates for the same month differ. The Commission's InforEuro accounting rate for July 2026 was 1.1406; the ECB's daily reference rate on 29 July 2026 was 1.1380 — a 0.23% gap, same institution's rate observed on two different dates. EU customs law deliberately uses a stale rate: the ECB rate from the second-last Wednesday of a month, applied for the whole following month, so up to about 43 days old. That is correct for calculating duty and wrong as a commercial conversion rate. Naming "the official rate" without naming the publisher, the series, and the date names nothing.
06 · The order of operations is a red herring — the fixed fee is not Convert-then-mark-up and mark-up-then-convert produce identical prices; percentage operations commute, and anyone selling the sequence as the mechanism is selling the wrong one. What genuinely does not commute is anything fixed. A flat per-line charge inside the markup base yields fee × markup more than the same fee outside it — on a 400-line BOM at an illustrative 15%, a EUR 25 per-line charge moves EUR 1,500. Tiered thresholds behave the same way: which side of a threshold a line falls on changes the basis, not just the amount.
The ambiguity

Why the BOM line cannot tell you which one happened

The same 13.80% has at least four decompositions — and the BOM line looks identical in all of them
How the number could have been builtPart that is sourcedResidual still needed
Rate taken as 1.00 — no conversion at all13.80% in one decision0%
Rate fixed 13 Jan 2025 (1.0198), never refreshed11.59%1.98%
An illustrative 5% FX buffer5.00% (illustrative)8.38%
Three conversion hops at a published 0.90% margin2.72%10.78%

Reference: ECB series EXR.D.USD.EUR.SP00.A, 29 July 2026 — EUR 1 = USD 1.1380, so USD 1,200 = EUR 1,054.48. Each row reconciles to 1.1380 to four decimals. Only the sourced column is evidence; the 5% buffer and the 15% markup used above are illustrative parameters we cannot source, and we say so rather than dressing them up. For calibration: 13.80% on materials is comparable in size to the entire gross margin the listed assemblers earn — Sanmina 8.81%, Plexus 10.08%, Benchmark Electronics 10.16% in FY2025.

The convention

Nine rules, fair from both sides of the table

The convention — nine rules, fair from both sides of the table

A manufacturer who adopts these is not left carrying uncompensated currency risk, and a customer who adopts them can reproduce every number. That is the test.

1 · Name one rate, one series, one date, one endpoint The ECB euro foreign exchange reference rate for a stated date, retrieved from a stated address — series EXR.D.USD.EUR.SP00.A, or the daily eurofxref-daily.xml. Free, four decimals, stably keyed, archived to 1999, and reusable with attribution.
2 · State the anchor's weakness in the same breath you name it The ECB says its reference rates are "published for information purposes only" and that using them "for transaction purposes is strongly discouraged". They are a 14:15 CET mid — the average of observed bid and ask — so they carry no spread, and a real conversion cost sits above them. They are not a regulated benchmark, and they will not equal the customs rate, the accounting rate, or any bank's dealt rate for the same shipment. Use them as the disclosure benchmark. That is exactly the role EU payment law already gives them.
3 · Fix the refresh cadence in the contract, not in someone's discretion Name the observation date and the period it governs. Filed agreements do this at every cadence — one fixes the rate on the second Thursday of the last month of each quarter; another resets monthly; EU customs uses the second-last Wednesday for the following month. Any of those is auditable. "Whenever we update the schedule" is not.
4 · One band, both directions, same magnitude, same trigger The US Department of Defense's own contracting policy states that a price-adjustment clause "will not be one-sided, but will be fair to both parties" and must "allow for both upward and downward revision of the contract price with a specified ceiling and floor of the same magnitude". A filed EMS agreement adjusts monthly where rates move more than ±0.75%. If the price only moves one way, the clause is not a currency clause — it is a ratchet.
5 · Say whether the threshold is a trigger or a deductible These are economically very different and the wording rarely says which. In one published clause, a deviation of 5% or more adjusts the price by half of the whole difference, not half of the excess — so a 6% move produces a 3% change, not 0.5%. Write down which one you mean.
6 · Show the conversion cost as its own line, as a percentage over the named rate This is the exact format EU law already imposes on payment providers for card transactions: express the total currency conversion charge "as a percentage mark-up over the latest available euro foreign exchange reference rates issued by the European Central Bank". That regime does not reach a costed BOM, and we do not pretend it does — it is a ready-made template, not an obligation we are under. Fixed fees stay stated as amounts, and any tiered threshold is declared.
7 · Keep the FX allowance outside the markup base, and true it up State any allowance as a percentage of the named reference rate, apply it after markup rather than inside the base, and say in writing what happens to it if the move never occurs. An allowance that cannot be returned is not an allowance; it is a price.
8 · Match the currency of cost first; then name who carries what is left Buy in the currency you sell in wherever the supply base allows it — roughly 40% of world exports are invoiced in dollars and more than 40% in euro, so this is usually achievable before any instrument is bought. Then say in one sentence who carries the residual. Carrying it is cheap and boring: corporates paid about 0.06% on EUR/USD forwards, and forward points are interest-rate mechanics, not a service anyone can bill as margin. And ask two suppliers to quote the same BOM with the rate and its source stated — in the same research, buyers using one dealer paid roughly fourteen times the spread of buyers who could compare quotes side by side.
9 · Reconcile the quoted rate to the settled rate, and publish the variance A filed EMS agreement already does this: it compares the average rate in the month before invoice against the rate on the payment date and settles the difference by supplementary invoice or credit within five days — flowing both ways rather than being retained. EU customs makes the same demand of a fixed rate, counting it only "provided that the settlement is actually based on that rate". This is the currency version of the landed-cost loop on the economics page: quote it, then prove it.
What makes it auditable, in one line

Rate, source, retrieval date, endpoint, and the conversion charge — printed on the BOM line, with the retrieved file kept. Federal US procurement already requires a buyer to insert the "source of rate" and its effective date when evaluating foreign-currency offers, and EU customs law requires additions to a price to rest "only on the basis of objective and quantifiable data". The test is simply whether two parties, given the same inputs, arrive at the same number. And a note on silence: choosing an Incoterm settles none of this — the ICC's own introduction lists the currency of payment among the matters its rules expressly do not deal with, and warns that leaving them unaddressed "is likely to cause problems later".

The honest limits

What this note does not prove

The currency convention above is published as a proposal, not as a report on something already proved — no full production program has run through all nine rules with a certified hub yet. Three evidence limits are worth stating plainly. First, we found no survey, audit, regulator finding, or litigation record measuring how often 1:1 currency treatment actually occurs in a costed BOM, or quantifying FX uplift in contract-manufacturing quotes at all: the mechanism is fully documented and the arithmetic is checkable to four decimals, but the prevalence is a first-hand observation, and we present it as one. Second, the 15% materials markup and 5% FX buffer used in the worked example are illustrative — the ranges in circulation trace back to providers' own marketing pages, not to a filing or a survey, so what survives is the identity (base × buffer × markup), not the parameters. Third, there is no published bank tariff for what a company our size pays for a forward contract or the collateral behind it, so our working band of mid-market plus roughly 10 to 100 basis points is our position, not a benchmark — the FSB's measured 1.6% average sits above the top of it. Two anchors also have licensing gaps: Norges Bank publishes no reuse licence for its rate data, and the main commercial alternative fix is licence-gated with no published price for referencing it in a contract.

Sources

Where these numbers came from

Field notes are research, not decisions — dated, sourced, and open to correction. Republish this in full or in part under CC BY 4.0, with credit and a link back. · All journal entries →