FX Best Execution for SMEs: Benchmarks, Spread Audits and the Paper Trail

Ask a finance director what they paid for a cross-border payment and you will usually get the fee line. Ask what they paid in the exchange rate and the conversation stops. The fee is visible, itemised and budgeted. The spread is embedded in a single number that appears on a confirmation and is rarely challenged.

That asymmetry is where most SME currency cost lives. A treasury desk that can prove its execution quality has leverage with providers, a defensible audit trail for the board, and, in group structures, the documentation that transfer pricing reviewers ask for. A treasury desk that cannot prove it is negotiating from memory.

This piece sets out how to establish best execution as a repeatable corporate process: which benchmarks are defensible, how to run a spread audit in an afternoon, how to measure timing cost separately, and what to keep on file for each trade.

Best execution is a corporate governance question, not a fund management one

The term comes from securities regulation. Under MiFID II, investment firms owe clients a duty to take all sufficient steps to obtain the best possible result when executing orders, and must publish an execution policy and evidence of how it was applied (Article 27, Directive 2014/65/EU). Corporates are not investment firms, so that duty does not bind them in the same way.

The logic transfers anyway, for three practical reasons:

• •Cost materiality. For an SME with €10m of annual cross-border flow, 20 basis points of avoidable spread is €20,000, roughly the cost of an additional finance headcount, spent invisibly.
• •Board defensibility. "We use our bank because we always have" is not an answer to a margin question. "We execute against a documented benchmark and our average cost is 14 basis points" is.
• •Tax and audit exposure. In group structures, intercompany FX rates are transfer-priced. An arm's length rate needs a benchmark and a record of how it was applied, not an assertion that the rate was fair.

Large corporates solved this decades ago with a dealing desk, an execution policy and transaction cost analysis. The mid-market version is smaller but structurally identical: a benchmark, a ticket record, a monthly variance number, an annual policy review. None of it requires new software.

The three numbers that make up your FX cost

Every currency conversion carries three distinct costs, and conflating them is the reason most SME FX reviews fail to change anything.

1. The spread against mid. At the moment of execution there is a mid-market rate. Your executed rate differs from it. That difference, in basis points, is the dealer's compensation for liquidity, risk and profit. On major pairs in liquid hours, competitive electronic execution sits in single digits to low double digits of basis points. A relationship-priced bank ticket on a modest size, handled by phone or a portal at an unhelpful hour, can be several times that.

2. Explicit fees. Wire fees, payment scheme charges, correspondent bank deductions, monthly account fees. Visible, easy to benchmark, usually the smallest of the three on large tickets, and often the largest on small ones.

3. Timing cost. The rate you executed at versus the rate available at a defensible reference point in your process, such as the decision date, the invoice date or the budget rate. Timing cost is the one nobody measures, and on volatile pairs it frequently exceeds the spread.

A 2026 market with central bank paths diverging and policy headlines arriving at odd hours makes point three more expensive than it was in a low-volatility year. The published WM/Reuters and ECB reference series are the standard public anchors if you need to reconstruct historical moves.

Choosing a benchmark you can defend

A benchmark has to satisfy three tests: it is independent of your provider, it is reproducible, and it is available at the time you need it. Four options are in common use.

For most SMEs the workable combination is the third for routine spread measurement, the second for accounting and revaluation, and the first where an intercompany or contractual rate must be externally verifiable. The 4pm fix is the market convention for scheduled flows precisely because it is a single, public, uncontestable number, which is also its weakness: everyone knows when it prints.

Whichever you choose, write down which benchmark applies to which flow category. A benchmark that changes after the fact is not a benchmark.

Running a spread audit in one afternoon

You do not need a system to start. You need twenty tickets, a mid-rate source and a spreadsheet.

Step 1: Pull the sample. Take the last three months of executed FX trades, in every currency pair that matters, and select the twenty largest by notional. Add five small ones, because small tickets often carry the worst percentage cost and you want that visible.

Step 2: Capture the mid. For each ticket, find the mid-market rate at the execution timestamp from a source independent of the dealer who executed it. Your provider's public mid feed, a second provider's quote, or a licensed reference feed will do, as long as it is consistent across the sample.

Step 3: Compute the cost in basis points.

Spread in bps = (Executed rate − Mid rate) ÷ Mid rate × 10,000

Sign convention matters. If you sold EUR and bought USD, a lower executed rate than mid is a cost. Fix the convention before you start, or the analysis will produce a number that argues with itself.

Step 4: Convert to money. A EUR/USD mid of 1.0850 executed at 1.0838 on a €5m sale is 11 basis points, or about €9,000. That is the sentence that changes behaviour in a management meeting. Basis points are abstract; the currency amount in the P&L line is not.

Step 5: Segment and compare. Cut the sample three ways: by provider, by size band, and by execution channel (portal, API, phone, branch). The pattern is usually stark. Electronic execution on mid-sized tickets is cheap. Manual execution on small tickets at month-end is expensive. Almost every SME that runs this exercise for the first time discovers that 80% of its cost sits in 20% of its tickets, concentrated in a channel or a day of the month.

Step 6: Set a reference level. Record what "good" looks like for your flow profile and your pairs. Do not import a number from a large corporate benchmark; your size and channel determine what is achievable. Then re-run the audit quarterly against that level.

Measuring timing cost separately, or the analysis stays incomplete

Spread audits flatter the treasury function because they stop at the point of execution. The larger question is whether the conversion should have happened then at all.

Three timing metrics are worth tracking monthly:

• •Decision-to-execution lag. Days between the business decision to convert and the trade. Anything beyond a day or two is either a workflow problem or an unmanaged position.
• •Invoice-date drift. For receipts and payments tied to an invoice, the rate at invoice date versus the rate at settlement. This is a working capital and hedging question, and it belongs in the same monthly pack as the spread audit.
• •Budget-rate variance. Realised blended rate versus the annual budget rate. This is the number the board already understands.

The first two are operational and fixable. The third is strategic and is what your hedging ladder exists to control. Run all three and the currency line in the management accounts stops being a surprise.

What to keep on file for each trade

An audit trail is only as good as what it can answer six months later, when the provider relationship is being renegotiated or a tax reviewer asks how an intercompany rate was set. Retain, per trade:

• •Trade date and execution timestamp, to the minute
• •Instrument and notional, both legs, with currency pair and direction
• •Executed rate, plus the benchmark rate and its source and timestamp
• •Counterparty or venue, and the channel used
• •Mid-rate reference used for the spread calculation
• •Dealer or platform reference number
• •Approver identity, where the policy requires dual sign-off
• •Purpose code and the underlying invoice or intercompany reference

Eight fields, captured at the point of trade, answer nearly every later question. Reconstructing them from confirmations and emails three quarters later does not.

The intercompany overlap most SME groups miss

If your group has entities in more than one currency, you are running an internal FX market, whether or not you treat it as one. Intercompany loans, management recharges and goods flows all create cross-currency positions that must be priced on arm's length terms.

Two practices keep this defensible:

• •Document the rate source. State in the intercompany policy which benchmark applies (typically the 4pm fix or the ECB reference rate for the relevant date), which date is used (invoice date, month-end, payment date), and who may deviate and how that is approved.
• •Separate treasury from the trading result. A group treasury centre that quotes subsidiaries at mid plus a documented cost, and books its own result separately, produces clean transfer pricing evidence and an honest view of treasury cost. A group that quotes whatever rate suits the subsidiary's budget is building a file it will have to explain later.

The OECD Transfer Pricing Guidelines require arm's length terms; they do not require sophistication. They do require a record.

A one-page execution policy

The whole process fits on a page, which is the point. If it does not fit on a page, it will not be executed.

1. •Scope. Currency pairs and flow categories covered. Everything else converts at spot, documented as out of scope.
2. •Benchmark. Which reference applies to which category, and the permitted window either side of execution.
3. •Thresholds. Ticket sizes above which competitive quotes or dual sign-off are required.
4. •Prohibitions. No speculative positions, no unhedged exposure above a defined level without written sign-off, no execution outside the approved venue list.
5. •Reporting. Monthly: spread variance by provider and channel. Quarterly: full audit against the reference level. Annually: policy review and provider repricing.
6. •Records. The eight-field trade record, retention period, owner.

The prohibitions are the load-bearing section. Nearly every corporate FX loss in the past three decades began as an execution programme that gradually acquired a view.

Four failures that repeat

Auditing without a benchmark. Comparing your rate to "what the market did" after the fact is not measurement, it is argument. Fix the reference before the trade.

Benchmark from the executing dealer. The mid you are measuring against must not come from the party being measured. This is the single most common flaw in SME FX reviews.

Treating all providers as equivalent. Banks, brokers and payment platforms price differently by pair, size and hour. Segment by pair and size before drawing conclusions, or you will keep routing tickets to the cheapest provider for the wrong flow.

Ignoring the small tickets. A 2% all-in cost on a €3,000 supplier payment is €60, which nobody chases, until the same pattern sits across 400 payments a year. Batch them, net them, or move them to a cheaper rail; the arithmetic is unglamorous and reliable.

Where to start this month

Pull twenty tickets, get an independent mid for each, and compute two numbers: total basis points paid, and total currency amount. That is your baseline. Then run the timing analysis on the ten largest flows and see which of the three costs dominates.

Almost every SME finds the same thing: the biggest single improvement is not a new provider, it is knowing which channel and which day of the month are costing the most. The benchmark turns a conversation about relationships into a conversation about numbers, which is the only kind a provider can respond to.

Execution quality is one of the few treasury capabilities that can be built entirely from discipline. The policy is a page, the audit is an afternoon, and the savings land in the margin line every quarter after that.