Reconciliation is the work of proving that what the business recorded matches what the bank, the mobile wallet or the card gateway actually settled. In a Bangladeshi business it rarely involves one source. A single day may include bank transfers, bKash and Nagad settlements, card gateway batches, cash on delivery and a few intercompany entries. Each source has its own file format, fee deduction and timing.
Automation does not remove the need to reconcile. It changes who does the work and how exceptions are found. This comparison sets out four approaches, the criteria that separate them, and the situations where each one is the sensible choice.
What reconciliation actually requires
- Source capture: statements, settlement files, gateway reports and cash sheets arriving on a predictable schedule.
- Matching: transaction ID, amount, date, reference and fee, with tolerance for timing differences.
- Exception handling: unmatched items, duplicate entries, partial refunds, chargebacks and missing settlements, each with an owner.
- Posting: journals or entries in the accounting system, with the right period and cost centre.
- Evidence: a trail showing who matched what, when, and what was changed manually.
- Controls: separation between the person who reconciles and the person who approves write-offs.
Any approach that cannot show these six things is a data entry tool, not a reconciliation control.
Option 1: spreadsheets with imported statements
The status quo for many finance teams: export a statement, paste it into a workbook, apply filters or macros and tick items against the ledger. It is flexible, cheap to start and understood by the team.
The trade-off is fragility. Formulas break silently, files are emailed around, and only one person knows how the workbook works. Evidence is a saved file, not an audit log. As transaction volume grows, the time saved by automation is consumed by checking the spreadsheet. It remains reasonable for a small number of accounts with low volume, provided the workbook is version-controlled, access is restricted and a second person reviews the output.
Option 2: accounting software bank rules and feeds
Accounting platforms can import bank feeds and apply rules that suggest or post a match. Xero and QuickBooks both support bank reconciliation workflows, and the accounting system already holds the ledger, tax codes and reporting.
The trade-off is coverage. Bank rules work well for recurring, predictable entries. They are weaker when a single gateway settlement covers many orders, when fees are deducted before deposit, or when mobile wallet statements arrive as a PDF rather than a feed. You may still need a separate matching step for those sources. Rule sprawl also becomes a problem: dozens of overlapping rules are hard to audit and easy to break when a payment provider changes its description format.
Fits when most money arrives through bank accounts, volumes are moderate, and the finance team already works inside the accounting platform.
Option 3: settlement middleware
Middleware sits between payment sources and the ledger. It ingests gateway and wallet settlement files, normalises the formats, matches against orders or invoices, and posts a summarised journal or a detailed transaction set. Exceptions land in a queue with a status.
The trade-off is another system to own. You need a mapping for every source, a schedule for file collection, a rule for fees and refunds, and a monitoring routine when a file does not arrive. The advantage is that reconciliation logic lives in one place rather than in a workbook and three provider portals. It also gives operations a view of settlement timing, which is useful when cash flow is tight.
Fits when the business takes money through several gateways or wallets, order volumes are meaningful, and finance needs a daily position rather than a month-end reconstruction.
Option 4: a custom matching engine
A custom engine is built around your exact rules: order-level matching for a marketplace, invoice-level matching for wholesale, fee schedules that differ by provider, and approval thresholds by amount. It can integrate with the ERP, the warehouse and the customer portal, and it can produce reconciliation evidence designed for an auditor.
The trade-off is cost and ownership. A custom engine needs a specification, test data, a rules catalogue, monitoring and a change process. The rules will change when a provider changes its fee or file layout. Without a named owner, the engine becomes the new spreadsheet, only more expensive to fix.
Fits when reconciliation is complex enough to be a competitive constraint, the volume is stable, and the business can support a software product over several years.
Decision guidance
- One bank account, low volume, simple entries: accounting rules are usually enough. Spend the effort on review discipline, not software.
- Several gateways, daily settlement, fee deductions: middleware is the pragmatic middle. It solves the multi-source problem without building a platform.
- Complex contracts, marketplace splits or multi-entity accounting: a custom engine may be justified, but only with a written rules catalogue and an owner.
- Cash-heavy operation: no tool fixes missing cash controls. Fix the receipting and deposit process first.
What to measure after go-live
Track unmatched items as a count and an age, not only as a percentage. A single old exception can matter more than a hundred items cleared today. Report the number of manual journals created to force a match, the time to close each day, and the number of settlements that arrived without an order reference. These measures show whether the chosen approach is reducing work or moving it.
Criteria for the comparison
- Match granularity: transaction, order, invoice or settlement batch.
- Source coverage: bank, wallet, card, cash and intercompany.
- Exception workflow: who sees an unmatched item, how it is assigned and how it is closed.
- Audit trail: immutable logs, approval history and exportable evidence.
- Control design: maker-checker, access separation and write-off limits.
- Integration: accounting, ERP and order systems, including failure handling.
- Total cost: licence, implementation, transaction fees, support and internal time.
Where this fits in our work
We build reconciliation and finance workflow automation under Business Automation. The pricing page explains how a scoped automation project is priced. For the day-to-day operating model, read how daily payment reconciliation changes year by year.
Sources
- Bangladesh Bank, Mobile Financial Services statistics, accessed 17 November 2026.
- Xero Central, Reconcile accounts, accessed 17 November 2026.
- Intuit Developer, QuickBooks Online API, accessed 17 November 2026.
- PCI Security Standards Council, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.