Bank Statement CSV Column Auto-Detection

Name: Global Bank CSV AI Map
Price: 65.1 USD

Upload a sample, let the heuristic engine map your bank's CSV columns, save a reusable template, and import to bank statement lines - for any bank, worldwide.

Stop hand-mapping CSV columns for every bank

Every bank exports its statement CSV differently: comma or semicolon delimiters, dot or comma decimals, ISO or day-first or month-first dates, a single signed amount column or separate debit / credit columns, headers in any language. This module reads a short sample of a real export and proposes a complete column mapping automatically using a transparent heuristic engine - then lets you save that mapping as a reusable template so every future file from that bank imports in one click.

Dialect sniffing

Field delimiter detection - comma, semicolon, tab or pipe - by field-count consistency scoring.
Header-row detection: knows when row one is labels versus already data.
Decimal / thousands separator inference: tells 1.234,56 (continental) from 1,234.56 (US/UK).
Quote-character detection for fields with embedded delimiters.

Date-format inference

Recognises ISO, day-first, month-first, dotted, slashed, dashed and named-month layouts.
Disambiguates DD/MM from MM/DD by finding a value whose first part exceeds 12 (which can only be a day).
Picks the single strptime pattern that parses the most sample rows, then re-parses every row deterministically.

Column-role detection

Scores every column for date, amount, debit, credit, balance, description, reference, currency and payee roles.
Blends content evidence (numeric / date / sign / cardinality ratios) with multilingual header keywords - English plus common German, French, Spanish, Italian, Dutch and Portuguese bank-export labels.
Works even on header-less files, using the data shape alone.
Chooses automatically between a single signed-amount layout and a split debit / credit layout.

Reusable templates & import

Save any detected mapping as a named, per-bank template.
Review and override the proposed role of each column before saving - every heuristic ratio is shown for full transparency.
Import a full bank CSV using a template straight into account.bank.statement.line records, ready for reconciliation.
Confidence score per role and overall, so you know when to trust the proposal and when to check.

What gets detected

Aspect	Detected values
Field delimiter	Comma, semicolon, tab, pipe
Header row	Present / absent
Number format	Dot-decimal vs comma-decimal, with grouping
Date format	ISO, DD/MM, MM/DD, dotted, named-month, 2- or 4-digit year
Amount layout	Single signed column or split debit / credit
Column roles	Date, amount, debit, credit, balance, description, reference, currency, payee

How it works

Upload a sample - a handful of rows of your bank's CSV export.
Detect columns - the engine proposes the dialect, formats and a role for every column, with a confidence percentage.
Review - adjust any role inline if you disagree; the numeric, date and distinct ratios behind each guess are shown.
Save as template - give the mapping a name tied to that bank.
Import - feed full statement files through the template and the rows land as bank statement lines on the journal you pick.

The detection engine is pure, deterministic Python - no external service is contacted and no data leaves your database. "AI" here means a transparent heuristic scoring model you can inspect, not a black box.

Technical notes

Builds on the community account module only - no Enterprise dependency.
Imported lines use the standard account.bank.statement.line model, so they appear in the normal bank reconciliation flow.
Accountancy parentheses (123.45) and trailing-minus 123.45- are both read as negative amounts.
Compatible with Odoo 18 and Odoo 19.

Screenshots

Import Csv

Detection Samples

Mapping Templates

Update date: 2026-07-02