Withheld settlements and unreleased reserves announce themselves: a number that should have arrived, and didn't. Deductions are quieter. They arrive inside statements you stopped reading line-by-line years ago, each one small enough not to fight, labelled just plausibly enough to pass. Added up across the life of an account — and especially across its last six months — they are frequently a claim in their own right.
Three patterns account for most of what we find. The examples below are illustrative, not drawn from any specific counterparty or file.
Pattern one: the scheme fine passed through without underlying documentation
Illustrative example. A merchant's statement shows a deduction of €14,000 labelled "Scheme non-compliance assessment — Q2". The merchant asks what it relates to. The answer: a card scheme levied a fine on the institution in connection with the merchant's traffic, and the agreement permits pass-through. End of explanation.
Pass-through clauses are real and common. But a pass-through is a chain, and every link in it is checkable: the scheme's actual notice to the institution, the amount the scheme actually levied, the portion genuinely attributable to your traffic rather than spread across a portfolio, and the contractual basis for passing it on at all — some agreements permit pass-through of fines caused by the merchant's breach, which is narrower than fines connected to the merchant's traffic.
What to ask for, in writing: the scheme notice, the calculation allocating the amount to you, and the clause relied upon. An institution that has the chain produces it. An institution that responds with the label again — "this was a scheme assessment" — has told you something useful.
Pattern two: chargeback allocations that don't reconcile against scheme data
Illustrative example. A merchant's own dispute records show 412 chargebacks in a quarter, of which 161 were successfully represented and reversed. The statements for the same quarter show 412 chargeback debits — and 118 reversal credits. Forty-three reversals exist in the scheme's world and not in the merchant's bank account.
Chargeback accounting has many legitimate moving parts: timing differences, currency conversion, scheme fees netted against reversals. That is exactly why it is a place where errors — and worse than errors — go unnoticed. The merchant sees a stream of debits and credits, has no practical way to pair them transaction-by-transaction, and trusts the net number.
The reconciliation is harder than pattern one but more mechanical: your dispute records (from the portal you exported on day one — see the 30-day checklist) on one side, the statement entries on the other, paired by case reference. The gaps that remain after honest pairing are the claim. This sits next to its sibling: scheme recoveries that never appear at all, which we treat separately because the evidence works differently.
Pattern three: the fees that appear at offboarding
Illustrative example. A terminated merchant's final statement carries entries never seen in four years of the relationship: "account closure administration — €2,500", "extended monitoring fee — €450/month, 6 months", "compliance file retention — €1,200".
Offboarding fees cluster at termination for a structural reason: the institution knows the relationship is over, the merchant is distracted by the termination itself, and the amounts can simply be netted against money the institution already holds. Deducting is easier than invoicing. Each fee has the same two questions hanging over it as pattern one — which clause, and where is the underlying cost? — and at offboarding the answers tend to be thinnest, because these line items are often priced nowhere in the agreement at all.
A variant worth separate attention: fees that were always in the agreement but never previously charged. A dormant clause — say, a monthly "non-compliance surcharge" for missing PCI attestation — that springs to life in the final quarter raises its own question: if the condition existed for four years, why did the charging start when the relationship ended? Selective enforcement timed to termination does not automatically defeat the clause, but it belongs in the ledger, dated, because patterns of timing are part of what a file argues.
How to read a settlement file
You do not need forensic software for the first pass. You need the exports, a spreadsheet, and a method.
1. Separate the streams. A settlement file mixes at least four flows: gross settled volume, processing fees, chargeback activity, and "other" — fines, assessments, adjustments, reserves. Tag every line into one of the four. The claim almost always lives in the last two.
2. Baseline the fees. Take a clean month from the middle of the relationship and compute the effective rate you actually paid. Then compute it for every other month. Steps in that rate have dates, and dates have explanations — a repricing you agreed to, or one you didn't.
3. Isolate every "other" line. Each entry that is neither volume nor standard processing fee goes into the deduction ledger: date, amount, label, the clause that supposedly authorises it, and whether you have ever seen underlying documentation. Sort by size. The top ten lines usually contain the file.
4. Match the chargeback pairs. Debits to reversals, by case reference, as in pattern two. Park the unmatched residue in its own list rather than forcing matches.
4b. Watch the currency column. Where settlement and deduction happen in different currencies, check which rate was applied and whether the agreement specifies one. A spread quietly taken on every converted deduction is itself a deduction — small per line, systematic across thousands.
5. Date the change. Deduction claims are strongest where behaviour visibly shifts — after a termination notice, after a risk-review letter, after you first complained. Plot the "other" stream over time. A flat line that bends at a meaningful date is worth more than any adjective you could attach to it. Keep the working: the spreadsheet, the source exports, the pairing logic. A quantum that can show how it was built survives scrutiny; a bare total invites it.
This is financial-controller work, and it is half of how we run a file — counsel and controller under one operative roof, because a deduction claim is won in the reconciliation before it is won anywhere else. The quantum that emerges from steps 1–5 is also, not incidentally, what determines whether the file is worth pursuing at all. Merchants tend to over-estimate the emotional entries and under-estimate the systematic ones; the spreadsheet corrects in both directions.
What this is worth
Individually, deduction entries are sized not to be fought — that is rather the point of them. In aggregate, across the final year of a terminated relationship, we regularly see deduction ledgers that rival the withheld settlement balance itself. They also strengthen everything else in the file: an institution shown to have deducted unsupported amounts has a harder time defending the reserve it is still holding.
If your statements have entries you cannot match to a clause — or your chargeback maths has never quite added up — submit a fit check. Five minutes, eight fields, no upfront cost, an initial view within five business days.
This article is for information purposes only and does not constitute legal advice.