← BACK TO CASE STUDY GENERATOR
CREATIVE TECHNOLOGYFinancial services exclusive

PROPOSED PROJECT

Payment Matching Tool

Match payments across bank statements, payment-provider reports, and accounting records. Suggest which entries belong together, show differences in amounts, and give the finance team a list to check with the original rows attached.

PAYMENT REVIEW / EXAMPLE

Three payments. One bank deposit.

Batch P-804
Suggested match · not posted
01 / ACCOUNTING RECORDS

The customer payments

INV-104₱12,000Batch P-804
INV-105₱8,000Batch P-804
INV-106₱5,000Batch P-804
Gross collected₱25,000
02 / PROVIDER REPORT

Explain the difference.

Gross payments
₱25,000
Provider fee
−₱300
Recorded refund
−₱2,000
Expected deposit₱22,700Batch P-804 / net field
03 / BANK STATEMENT

The amount received

₱22,700
Row B-203Reference P-804Difference: ₱0

Amount and batch reference agree. A reviewer still approves the accounting entry.

Suggested match
Fictional invoices, provider rows, and bank entries. The amounts reconcile in this example; no live account is connected and no entry has been posted.

Match the money.
Keep the explanation.

In a fictional daily close, invoices of ₱12,000, ₱8,000, and ₱5,000 form provider batch P-804. The provider records a ₱300 fee and ₱2,000 refund, so the expected bank deposit is ₱22,700, matching row B-203 and its batch reference. A separate ₱5,000 entry could fit two invoices, and batch P-805 is ₱50 short. Those two groups stay unresolved.

WHAT IS USUALLY DONE

The finance team compares accounting rows, payment-provider reports, and bank statements in separate spreadsheets. Fees and refunds can make several customer payments arrive as one different bank amount. Matching by amount alone can pick the wrong invoice or hide an unexplained shortfall.

WHAT SOVRUN CAN DO FOR YOU

Sovrun would build a payment review desk that reads agreed exports and proposes match groups. It uses payment and batch references, currency, dates, and recorded fees or refunds to connect the source rows. AI helps read inconsistent descriptions and explains the proposed grouping; code checks the arithmetic. Reviewers approve supported matches and keep ambiguous or short-paid groups in a separate queue before any accounting export.

Some rows
should stay unresolved.

An exact amount is not enough when two invoices could fit. An unexplained shortage should not disappear into a guessed fee. These two groups stay with the finance team.

B-204 / REFERENCE MISSING

Two possible invoices.

₱5,000Bank entry

INV-118₱5,000

INV-121₱5,000

Both amounts fit. Ask for a remittance reference before selecting an invoice. Neither candidate is marked paid.

Needs a reference
P-805 / AMOUNT DIFFERENCE

₱50 still unexplained.

Provider gross
₱8,000
Recorded fee
−₱100
Expected deposit
₱7,900
Bank row B-205
₱7,850
Amount to investigate₱50

Request the missing adjustment or corrected report. Do not invent a fee or post an automatic write-off.

Needs investigation

A shorter review queue.
Every decision traceable.

The first build reads one bank export, one provider report, and the accounting extract. This example sorts three payment groups into one proposed match and two unresolved checks.

MATCH GROUP / P-804

₱22,700

Accounting
3 invoice rows
Provider
P-804
Bank statement
B-203
Unexplained difference
₱0
01 / A proposed match sheetConnected bank, provider, and accounting rows with fees and refunds accounted for.
TWO GROUPS TO CHECK

Clear next steps.

B-204Request the remittance reference.

P-805Investigate the ₱50 difference.

02 / An exception queueEach unresolved group shows the missing evidence and the action a reviewer needs to take.
EXPORT / NOT RELEASED

No silent posting.

  • Reviewer and decision time
  • Original source row IDs
  • Duplicate import check

Unresolved rows stay outside the export.

03 / A reviewed exportOnly approved matches enter the agreed accounting export, with original rows and reviewer decisions retained.

How we’d make it happen.

  1. Read and normalize the exports

    Preserve original row IDs and amounts while mapping dates, references, currencies, fees, and refunds into an agreed format. Detect duplicate files and repeated rows.

  2. Build explainable match groups

    Use exact references and the supplied settlement rules first. Suggest other candidates for review, calculate differences, and leave unresolved groups outside the matched set.

  3. Approve before exporting

    The finance team checks the original rows beside each suggestion. Export approved matches only and retain reviewer decisions so a repeated import cannot silently duplicate them.

Where AI helps

Document and language models interpret report layouts, normalize descriptions, and propose candidates when references are inconsistent. Exact identifiers, currency rules, duplicate checks, and calculations run in code. Model suggestions keep the original rows attached and cannot turn an ambiguous amount into a confirmed match or create a balancing charge.

What people approve

The finance reviewer approves each match group, investigates missing references and unexplained differences, and authorizes the final accounting export. The system never silently writes off a difference, marks an uncertain invoice paid, or posts to a live ledger. Access and retention settings follow the client’s agreed accounting process.

What we’d need to start

  • Bank statement exports with row IDs, dates, references, amounts, and currencies
  • Payment-provider reports with batch IDs, payment references, fees, and refunds
  • Accounting exports, matching rules, reviewer permissions, and the agreed import format

How we’d test it

Reconcile each match group back to the original exports. Test batched payments, refunds, fees, duplicate imports, missing references, date shifts, currency differences, and two invoices with the same amount. Confirm that ambiguous rows and unexplained shortages remain unposted. Compare review time and mistaken matches with the team’s spreadsheet process.

Prepared for Financial services. We would agree on the work after reviewing your brief, materials, and requirements.

DISCUSS THIS PROJECT

This is a proposed project, not completed client work. The visuals and numbers show example outputs, not measured results.