Finance reconciliation automation is usually sold as a matching engine. The matching engine is the cheap part. The cost sits in getting complete, consistent data from each source, handling the exceptions the rules cannot resolve and proving to an auditor that the process is controlled.
This guide shows how to scope the work in a way that survives procurement: what to count, how to build the estimate and how to turn the numbers into a budget worksheet. It avoids fixed BDT figures because source count and exception handling vary too much between businesses.
What the automation replaces
A reconciliation process has six steps: collect statements, normalise the data, match transactions, research exceptions, post the result and retain evidence. Automation can replace part or all of each step. Buying all six when only two are painful is a common way to overspend.
Start with a time log. For two weeks, record where the finance team spends its reconciliation hours: downloading files, cleaning spreadsheets, matching, chasing differences, posting to the ledger or preparing audit evidence. The log is more useful than an opinion about "too much manual work", and it gives the vendor a real scope.
The cost drivers
- Source count and format: each bank, payment gateway, mobile financial service, ERP and spreadsheet is a separate connector and a separate set of quirks.
- Identifier quality: a shared order or transaction reference makes matching cheap. Free-text narrations and truncated references make it expensive.
- Match complexity: one-to-one matching is simpler than one-to-many, split payments, partial refunds, batch settlements and many-to-many fee allocations.
- Exception rate: the percentage of transactions that need human judgement drives the workflow, staffing and support model. Measure it from a sample; do not assume it.
- ERP or accounting write-back: reading data is easier than posting controlled journals into a live ledger with approvals and reversals.
- Audit and retention: evidence trails, immutable logs and retention rules add design and storage work.
- Custom rules: allocations, accruals, intercompany entries and fee schedules that exist only in a spreadsheet.
- Migration and parallel run: historical balances, opening positions and a period where old and new processes run together.
- Support: source formats change, banks update portals and gateways revise APIs. Someone must own those changes.
The cost components
| Component | One-off or recurring | What to count |
|---|---|---|
| Process discovery | One-off | Sources, steps, roles, exception types, control points |
| Connectors and ingestion | One-off plus support | APIs, file formats, portal exports, credentials |
| Data normalisation | One-off | Field mapping, date and currency rules, deduplication |
| Matching rules | One-off | Exact, tolerance, fuzzy and many-to-many logic |
| Exception workflow | One-off | Queue, assignment, notes, approvals, escalation |
| Ledger integration | One-off plus support | Journal creation, reversal, period close, access |
| Audit evidence | One-off plus storage | Immutable logs, reports, retention and retrieval |
| Testing and parallel run | One-off | Sample periods, edge cases, reconciliation of the reconciler |
| Training and documentation | One-off plus refresher | Preparers, reviewers, approvers, month-end cover |
| Platform and support | Recurring | Subscription, maintenance, monitoring, change requests |
A scoping model you can calculate
Use complexity points rather than prices. Score each dimension from one to three, then add the points to compare proposals and to explain why two quotes differ.
One-off index = discovery + source connectors + normalisation + matching rules + exception workflow + ledger integration + audit evidence + testing + training + contingency.
Recurring index = platform tier + support tier + internal reviewer hours + monitoring and change budget.
A standard bank CSV with a clear reference scores one. A portal that only offers a PDF statement scores three. A one-to-one order-to-payment match scores one. A settlement that covers hundreds of orders with fees and refunds scores three. An exception queue with two roles scores one. A queue with approval limits, audit notes and escalation scores three.
Add contingency for unknowns, not as a fixed percentage. If a source cannot provide a test environment or a sample file, mark it as an unknown and size the contingency from that count.
Calculator logic
A useful calculator asks for inputs the finance team already knows and returns an effort band, a licence tier and a pilot length.
- How many sources must be reconciled each day or month?
- How many transactions per month, and how many are refunds, chargebacks or split payments?
- What percentage of transactions need manual review today, measured from a sample?
- How many matching rules are needed, and do any involve tolerances or many-to-many allocation?
- Must the result post to an ERP or accounting system, or is an export enough?
- How many months of history must be available for audit?
- How many user roles need access, and who approves adjustments?
- What support window is required around month-end and year-end?
The output should separate build effort from run effort. A low build quote with a high run cost is not cheaper over three years. Ask for both in the same proposal.
A worked example
Assume four sources, 12,000 transactions a month, three matching rules, an ERP write-back, four user roles and twelve months of history. A sample shows that eight per cent of transactions need manual review. That is 960 exceptions a month. If each takes five minutes, the team spends about 80 hours a month on exceptions alone. Treat those figures as an illustration, not a benchmark; replace them with your own sample.
The model now explains the project. Connectors and normalisation are a fixed build cost. The exception workflow is the variable cost, because 960 items a month need assignment, notes and review. A vendor that automates matching but leaves exceptions in email has not solved the expensive part. A vendor that offers a queue, bulk actions and an audit trail is addressing the real workload.
Budget worksheet
| Line | Basis | Owner | Evidence |
|---|---|---|---|
| Discovery | Sources, steps, exceptions | Finance controller | Process map and time log |
| Connectors | Source count and access method | IT lead | Interface list and sample files |
| Rules | Match types and tolerances | Finance controller | Rule catalogue |
| Exception workflow | Volume, roles, approvals | Finance manager | Queue design |
| ERP integration | Write-back, reversals, period close | ERP owner | Integration specification |
| Audit evidence | Retention and retrieval | Internal audit | Evidence sample |
| Testing | Sample periods and edge cases | Project lead | Test report |
| Support | Hours, response, month-end cover | Finance and IT | Service description |
Questions to ask a vendor
- Which steps of our current process will the system replace, and which stay manual?
- How will you measure the exception rate before quoting the workflow?
- What happens when a bank changes its statement format?
- Can the system show who changed a match, when and why?
- Does the quote include ledger posting, reversals and period-close support?
- What is the exit path if we change accounting systems?
A 90-day plan
- Days 1-15: log the current process and sample the exception rate.
- Days 16-30: define the reconciliation dictionary: identifiers, totals, cut-off and approval rules.
- Days 31-45: build one source-to-ledger path and test it against a closed period.
- Days 46-60: add the remaining sources and the exception queue.
- Days 61-75: run in parallel, compare results and investigate every difference.
- Days 76-90: train preparers and approvers, agree support terms and retire the old spreadsheet.
Reconciliation automation is delivered through Business Automation, alongside finance workflow and reporting. The pricing page explains how discovery, build and support are scoped separately. For the approach comparison behind this budget, read the comparison of finance reconciliation automation approaches.
Sources
- ISO 20022, Universal financial industry message scheme, accessed 17 November 2026.
- Bangladesh Bank, Mobile Financial Services data, accessed 17 November 2026.
- PCI Security Standards Council, accessed 17 November 2026.
- COSO, Internal Control — Integrated Framework, 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 *.