Financial reconciliation automation
Context
B2B company with a high volume of daily financial transactions. Receives roughly 10 PDF statements per day, each with thousands of transactions. Traditionally, an operator spent about 2 hours daily manually checking which values in the outgoing column had no counterpart in the incoming column, and vice versa.
What we deliver
A Python desktop tool, distributed as a standalone Windows executable. The usage flow is direct:
- User drags one or more PDFs into the interface.
- Clicks process.
- Receives an Excel spreadsheet with two tabs: outgoing without matching incoming, incoming without matching outgoing.
The tool runs 100% offline. No server dependency, no financial data sent anywhere, no Python install or any external dependency required. The client gets a single .exe file that opens with two clicks.
How the algorithm decides what's reconciled
Three criteria across two passes:
- Same numerical value.
- Same text name, with tolerance for names truncated by the bank's system.
- Time proximity: strict filter first; when no match is found, relaxes to any time on the same day or different days.
Criteria are adjustable via a settings screen: time tolerance, minimum name prefix length, second-pass on/off. This lets the client refine behavior as the operation evolves.
Expected outcome
- ~40 hours per month of operational work recovered.
- Full traceability: each run generates an auditable spreadsheet with a summary of matched and unmatched pairs.
- Zero operational risk: offline tool, no third-party dependency.
Stack
How we execute this project
Paid 1-week discovery, scope, timeline and price locked, 4 weeks of construction, delivery with a 90-day warranty. No retainer, no hourly billing, no budget surprises.