Rule-based transaction review
Bank statement categorizer with rules you can review
Categorize transactions from a supported native-text bank statement with categories you define and rules you can inspect. Conflicts stay visible, manual changes stay identifiable, and every exported row keeps its source-page evidence.
Private browser review
Category review
Categorized
3
Conflicts
1
Unclassified
1
ACME PAYROLL
Rule · Source page 1
+2,500.00
Income
UTILITY PAYMENT
Manual · Source page 1
-86.42
Utilities
ONLINE TRANSFER
Review · Source page 2
-500.00
Conflict
CARD PURCHASE
Review · Source page 2
-24.00
Unclassified
Two rows still need review. Exports keep them explicitly flagged.
Illustrative product preview. Your private queue reflects the extracted rows and rules in your own job.
How categorization works
From extracted rows to reviewed categories
Extraction and categorization are separate decisions. First make the statement rows traceable; then apply your taxonomy and inspect what the rules could not resolve.
01
Extract reviewable rows
Start with a supported native-text statement. The parser keeps signed amounts, confidence, balances, and source pages before categorization begins.
02
Create your categories
Use labels that fit your own review or bookkeeping process. The tool does not impose a universal chart of accounts.
03
Add transparent rules
Match description, reference, or both. Add direction and amount limits when a text pattern alone would be too broad.
04
Resolve exceptions
Filter conflicts and unclassified rows, then make a single-row or selected-row manual correction without hiding its source.
05
Export with evidence
Download a versioned CSV or XLSX containing category provenance, matched rules, review flags, and source-page references.
Rules you can read before you trust them
Each rule is a bounded condition, not a hidden prediction. Use direction and amount guards to narrow broad description matches. Rules never run executable expressions or regular expressions.
Manual corrections take precedence. If two equally ranked rules disagree, the row remains a conflict until you decide.
- Search field
- Description, reference, or both
- Operator
- Contains, starts with, or exact match
- Direction
- Any, debit, or credit
- Amount guard
- Optional minimum and maximum absolute amount
- Priority
- Lower numbers win
- Conflict policy
- Equal-priority winners across categories stay unresolved
Exception-first review
See conflicts and unclassified rows before export
Filter the review queue by conflicts, unclassified rows, assigned rows, or free-text search. Correct one row, select several rows for the same category, or undo a change made in the current page session.
| Description | Signed amount | Category | Source | Review | PDF page |
|---|---|---|---|---|---|
| ACME PAYROLL | +2,500.00 | Income | rule | No | p. 1 |
| UTILITY PAYMENT | -86.42 | Utilities | manual override | No | p. 1 |
| ONLINE TRANSFER | -500.00 | — | conflict | Yes | p. 2 |
Synthetic sample rows. Categories illustrate review states, not accounting advice.
A rule match is not an accounting decision
The same transaction can need a different category for management reporting, bookkeeping, tax, lending, or personal budgeting. Your rules organize rows; they do not determine deductibility, compliance, or ledger treatment.
Review the source row and your own policy before relying on a category. If an accounting system has its own required fields, map the export deliberately rather than assuming compatibility.
Read QuickBooks guidance on categorizing and reviewing bank transactionsVersioned CSV and XLSX for review, not silent posting
Both exports use the versioned categorized-output contract. Unresolved rows remain exportable only with an explicit review-required marker, so the file never turns an unknown result into a confirmed category.
Choose the bank statement to CSV converter when you need the uncategorized accounting-feed contract, or review the formatted PDF bank statement to Excel converter before adding your own category rules.
What the categorized export retains
- Transaction and posting dates
- Description and reference
- Debit, credit, and signed amount
- Balance, currency, and extraction confidence
- Source page
- Category ID and label
- Category source and matched rule IDs
- Explicit category-review flag
Formula-like descriptions, references, category labels, and rule text are escaped before they reach the spreadsheet.
What we verify
Tested for predictable rules and safe exceptions
The rule engine is checked against 30 bank-neutral synthetic transactions across eight example categories. Separate cases cover rule priority, debit and credit direction, visible conflict, manual override, invalid profiles, and formula-safe exports.
Separate pipeline checks cover private upload, processing, authenticated review data, signed download, and scheduled deletion. These checks verify the documented workflow; they do not measure real-world categorization accuracy.
30
Synthetic rule cases
Bank-neutral descriptions with declared expected categories.
10
Workflow tasks
Conflict, bulk edit, undo, recovery, export, mobile, and expiry.
3
Private-object types deleted
Source PDF, parser workbook, and normalized review JSON.
Review safeguards
What stays inspectable, private, and time-limited
Deterministic by design
The categorizer uses inspectable rules, not a visual model, LLM, or hidden merchant-enrichment service.
Session-scoped draft
Categories, rules, and overrides stay in job-scoped browser session storage and are not written into the URL.
24-hour file lifecycle
The source PDF, parser export, and private review artifact use the documented 24-hour service lifecycle.
Questions
Bank statement categorizer FAQ
How do I categorize bank statement transactions?
Convert a supported native-text PDF into reviewable rows, create the categories you need, add explicit rules for repeat descriptions or references, and review every conflict or unmatched row before export.
Can I use my own categories?
Yes. Categories are defined by you and stored with the private job draft in the current browser session. The tool does not impose a universal tax or chart-of-accounts taxonomy.
Does this bank statement categorizer use AI?
No. The current workflow uses deterministic rules that you can inspect: searchable field, operator, pattern, direction, amount range, and priority. It does not use an LLM, visual model, hidden merchant database, or cross-customer learning.
What happens when two rules disagree?
If the best-priority rules point to different categories, the transaction is marked as a conflict. No category is guessed. You can review the matched rules and apply a manual category.
Can I change several transactions at once?
Yes. Select the rows, choose a destination category, and apply one manual change to the selection. The workspace shows the selected count and supports in-session undo before export.
Can I export categorized transactions to Excel or CSV?
Yes. The browser creates a categorized XLSX or CSV with category provenance, matched rule IDs, review flags, confidence, and source pages. The export is not presented as a ready-made QuickBooks or Xero import template.
Can it categorize a scanned or encrypted statement?
Not in the current workflow. Scanned, encrypted, malformed, and unsupported PDFs are rejected instead of producing guessed transaction rows. Use an authorized native-text PDF before categorization.
Open the private category-review workflow
Upload only statements you own or are authorized to process. Review every category before using the export for financial work.