The verdict in three sentences
An Amsterdam wholesale SME processing 25,000 bank transactions a year and matching them by hand spends about 70 hours a month on low-value work. An automation pipeline (PSD2 bank connection, matching rules, AI suggestions, export to Exact, AFAS or Twinfield) costs EUR 8,000 to 18,000 excluding VAT in 2026. With a target of 85% automatic matching, it frees about 60 hours a month and pays back in under a year.
Why manual matching breaks down in wholesale
Wholesale stacks up difficulties: hundreds of customers paying several invoices in one transfer, truncated bank descriptions, early payment discounts, credit notes and FX fees on foreign suppliers. The accounting lead exports the statement, looks for the matching invoice in the ledger, then matches line by line. A gap of a few cents blocks the entry and pushes the work to the next day.
| Bank transaction type | Share of volume (2026 order of magnitude) | Target automatic match rate |
|---|---|---|
| Customer transfer with invoice reference | 40% | 98% |
| Customer transfer covering several invoices | 20% | 85% (AI suggestion) |
| Supplier direct debits and subscriptions | 15% | 95% (fixed rules) |
| Card payments and terminal fees | 10% | 90% |
| Transfers with a gap (discount, fees) | 10% | 60% (proposal to validate) |
| Other items (VAT, payroll taxes, loans) | 5% | 80% |
| Weighted average | 25,000 transactions a year | about 85% |
The remaining 15% do not vanish: they land in a review queue with a probability-ranked suggestion, which cuts handling time per case by three or four.
What the project includes and what it costs
The PSD2 connection goes through a licensed aggregator (Tink, GoCardless Bank Account Data, Ponto or similar) that pulls statements daily with no rekeying. The matching engine first applies deterministic rules (amount, reference, IBAN), then AI proposes invoice combinations for grouped transfers. Matched entries are exported to Exact Online, AFAS, Twinfield or your accountant's software.
| Component | Scope | Indicative 2026 price (EUR excl. VAT) |
|---|---|---|
| Scoping and bank flow analysis | 3 to 5 accounts, 12 months of history | 1,000 to 2,000 |
| PSD2 bank connection | Licensed aggregator, daily sync | 1,500 to 3,000 |
| Matching rules engine | Amount, reference, IBAN, tolerances | 2,000 to 4,000 |
| AI suggestions for complex cases | Grouped transfers, gaps, learning | 2,000 to 5,000 |
| Export to Exact, AFAS or Twinfield | Journal file or API | 1,000 to 2,500 |
| Review screen and dashboard | Exception queue, match rate | 500 to 1,500 |
| Total project | Wholesale SME, 25,000 transactions | 8,000 to 18,000 |
Recurring costs include the bank aggregator (EUR 30 to 100 a month depending on the number of accounts), hosting (EUR 40 to 100 a month) and annual maintenance of about 12% of the project.
Custom build or off-the-shelf software?
Tools built into Exact Online or standalone matching apps already handle simple reconciliation. In wholesale they often reach 50 to 65% automatic matching, because they struggle with grouped transfers and customer-specific tolerances. A custom build targets the remaining 25 points and fits your ERP without changing accounting software. If your current rate already exceeds 75%, advanced configuration may be enough.
Mini case study
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Sanne, accounting lead at a food wholesale SME in Amsterdam, processes 25,000 bank transactions a year. Her team spends 70 hours a month on it. An accountant costs EUR 45,000 fully loaded for 1,600 hours, about EUR 28 an hour. With 60 hours saved a month, the gain reaches 720 hours a year, about EUR 20,200. The project costs EUR 13,000, plus EUR 2,400 of annual recurring costs. Payback lands around month 9, and the monthly close moves forward by 4 days.
FAQ
Is the PSD2 connection secure?
Yes, it runs through an aggregator licensed by a European regulator (DNB in the Netherlands), read-only. Bank consent renews every 180 days.
Does our accountant need to switch tools?
No, the export follows the format of their software. In 90% of cases a standard journal file is enough.
How long to reach 85% matching?
Rules hit 70% at go-live. The AI suggestions then improve over 2 to 3 months by learning from your validations.
What happens to unmatched transactions?
They land in an exception queue with 1 to 3 ranked suggestions. Average time per case drops from about 4 minutes to 1 minute.
What is the project timeline?
Plan 5 to 8 weeks, including a month running in parallel with your current method.
Let's scope your project. Tell us your banks, accounting software and transaction volume, and we will scope an automation between EUR 8,000 and 18,000 delivered in 5 to 8 weeks. Detailed quote within 48 h. WhatsApp +221 77 596 93 33.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
