Bank Statement Parsing Benchmark

Parsing Challenges

36 real-world formatting problems seen across 81+ real banks and 15 benchmark statements.

Last updated: March 17, 2026
36 challenges
Layout & StructureBank statements only

Separate Credit & Debit Columns

Statements use separate Credit and Debit columns in some formats, while others use a single Amount column with signs. Treating one layout like the other leads to wrong amounts or flipped signs. We automatically detect which layout is used and normalize the output.

How it looks

date
description
debit
credit
balance
ONLINE PURCHASE - AMAZON.COM Order #112-7834
4,954.01
SALARY DEPOSIT - ACME CORP January payroll
8,454.01
UTILITY PAYMENT - ELECTRIC CO Account #98765
8,326.51
ATM WITHDRAWAL
8,126.51
TRANSFER FROM SAVINGS Monthly transfer
8,626.51
↑ Scroll ↓

In 10 synthetic statements

🇸🇬 bsb-001🇭🇰 bsb-004🇨🇦 bsb-005🇲🇽 bsb-006🇦🇺 bsb-008🇬🇧 bsb-009🇮🇳 bsb-010🇭🇰 bsb-011🇰🇿 bsb-013🇹🇭 bsb-014

Often appears with

Test your parser against these challenges

Download the benchmark dataset and see how your parser handles real-world formatting problems.

See how commercial converters score on our benchmark: Accuracy Benchmarks