Automate Bank Statement Extraction for Accurate Books

Bank statements are essential records, but pulling each line into your accounting system by hand is a time‑sink that invites mistakes. The moment you finish copying the last row, you’re already behind schedule for reconciliation, reporting, and strategic analysis.

You describe it

Bank Statement Transaction Extractor

1. Overview

The Bank Statement Transaction Extractor reads a bank‑statement PDF, pulls out each transaction line, and turns it into a clear, structured list of transactions. Each entry is labelled with the date, description, amount, whether it is a credit or debit, and a category that describes what the transaction is for.

2. Business Value

Save time – eliminates manual copy‑and‑paste from PDFs to accounting systems.
Reduce errors – eliminates transcription mistakes.
Speed up reconciliation – provides a ready‑to‑use list for accountants or bookkeepers.
Standardize data – ensures all transactions follow the same format, making downstream reporting easier.

3. Operational Context

When: Whenever a new bank‑statement PDF is received and needs to be recorded in the bookkeeping system (typically monthly, but also for ad‑hoc statements).
Who: Accountants, bookkeepers, or anyone responsible for entering transaction data.
Frequency: As often as statements are received (usually monthly).

4. Inputs

Name / Label	Type	Details Provided
Bank Statement PDF	PDF Document	A PDF file containing one month’s bank‑statement. The PDF must contain a table of transactions with at least these columns: Date, Description, Amount (optional Balance). The file can be multi‑page but must be text‑searchable (no scanned images).

Notes on the Input PDF

The file must contain clearly separated rows for each transaction.
The date must be in a recognizable format (e.g., 2024‑07‑15 or 15/07/2024).
Amounts can include a dollar sign, commas, and parentheses for negatives (e.g., $1,200.00 or ($45.67)).
The PDF may contain other sections (e.g., headers, footers, or summary tables); the extractor will ignore everything that does not match the transaction row pattern.

5. Outputs

Name / Label	Contents	Formatting Rules
Extracted Transaction List	A list where each entry contains the following fields:
• Date – formatted as `YYYY‑MM‑DD`
• Description – text as it appears on the statement, trimmed of extra whitespace
• Amount – numeric value with two decimal places. Positive numbers represent money coming into the account (credits). Negative numbers represent money going out (debits).
• Transaction Type – `"Deposit"` for positive amounts, `"Withdrawal"` for negative amounts.
• Category – a short label that classifies the transaction (e.g., `Office Supplies`, `Salary`, `Utility`, `Transfer`, `Other`). The category is chosen based on keywords in the description (see Appendix C).
• Notes – optional free‑text field for any extra remarks (e.g., check number, memo).	- Present the list as a plain‑text table with columns: Date, Description, Amount, Transaction Type, Category, Notes.

Do not generate any new identifiers (e.g., transaction IDs).
Keep the order of transactions as they appear in the statement (chronological). | | Validation Report (optional) | A brief list of any rows that could not be processed, with the reason for each (e.g., “Missing amount”, “Unrecognizable date”). | - One line per problem row.
Use plain text; no extra formatting. |

6. Detailed Plan & Execution Steps

Open the PDF – open the supplied Bank Statement PDF.
Locate the transaction table – identify the area where rows contain a date, a description, and a numeric amount.
Read each row – for every row in the identified area: a. Extract the date – copy the text that looks like a date and convert it to YYYY‑MM‑DD format. b. Extract the description – copy the description exactly as shown, removing extra spaces at the beginning or end. c. Extract the amount – remove any dollar sign, commas, and any parentheses. Convert the string to a number with two decimal places. If the original amount was in parentheses, treat the number as negative. d. Determine Transaction Type – if the numeric amount is positive, label it "Deposit"; if negative, label it "Withdrawal". e. Assign Category – compare the description to the keyword‑to‑category table in Appendix C. Use the first matching category. If none match, assign "Other". f. Add notes – if the description contains a known pattern such as “check #” or “memo:”, copy that portion to the Notes field; otherwise leave blank.
Create the entry – put together the six fields (Date, Description, Amount, Transaction Type, Category, Notes) into a single row.
Repeat – process all rows in the same order they appear in the PDF.
Validate the list:
- Ensure every entry has a date, a description, and a numeric amount.
- Verify the date is correctly formatted.
- Confirm that the amount is a number and the sign matches the Transaction Type.
- Check that each Category is from the approved list (Appendix C).
- Flag any row that fails a check; do not include it in the final list.
Produce the output – write the complete list of valid transaction rows under the Extracted Transaction List heading.
Generate the Validation Report – list any flagged rows with a brief reason (e.g., “Missing amount – row 12”).

7. Validation & Quality Checks

Date format: All dates must follow YYYY‑MM‑DD. If a date cannot be parsed, flag the row.
Amount: Must be a numeric value; any non‑numeric entry causes a flag.
Transaction Type must match the sign of the amount.
Category: Must be one of the categories listed in Appendix C. If not, assign "Other" but still note it.
Duplicate detection: If two rows have identical date, description, and amount, treat the second as a duplicate and flag for review.
Balance consistency (optional): If the PDF includes a balance column, confirm that the cumulative sum of amounts matches the change from the first to the last balance. If it does not match, flag all rows.

If any validation step fails for a row, that row is excluded from the final transaction list and added to the Validation Report.

8. Special Rules / Edge Cases

Situation	Action
Parentheses in amount	Treat as a negative value (e.g., `(12.34)` → `-12.34`).
Comma‑separated numbers	Remove commas before converting to a number (e.g., `1,234.56` → `1234.56`).
Date in different format (e.g., `15/07/2024`)	Convert to `YYYY‑MM‑DD` if possible; otherwise flag for manual review.
Multiple keywords match	Use the first matching keyword in the order they appear in the description.
No keyword match	Assign category `"Other"`.
PDF contains scanned images	The process cannot read scanned images. Flag the entire file for manual entry.
Multiple pages	Process all pages in order; continue extracting until the end of the document.
Missing essential data (date, amount, or description missing)	Flag the row and exclude it from the final list.
Unexpected currency symbols (e.g., €, £)	Remove the symbol and treat the number as a standard amount.
Large number of rows (more than 1,000)	Process in batches of 500 rows to keep memory use reasonable; the final output remains a single list.

9. Example

Input – a PDF titled “June‑2024‑Bank‑Statement.pdf” containing the following rows (the PDF also includes headers, but they are ignored).

Date	Description	Amount
2024‑07‑01	Salary – June	$3,200.00
2024‑07‑03	Starbucks Coffee #123	(5.25)
2024‑07‑05	Electricity Bill - ACME Utility	(120.45)
2024‑07‑07	Transfer to Savings	(500.00)
2024‑07‑10	Interest Earned	$5.00

Output – Extracted Transaction List

Date	Description	Amount	Transaction Type	Category
2024‑07‑01	Salary – June	3200.00	Deposit	Salary
2024‑07‑03	Starbucks Coffee #123	-5.25	Withdrawal	Meals & Entertainment
2024‑07‑05	Electricity Bill - ACME Utility	-120.45	Withdrawal	Utilities
2024‑07‑07	Transfer to Savings	-500.00	Withdrawal	Transfer
2024‑07‑10	Interest Earned	5.00	Deposit	Interest

Validation Report (none needed – all rows processed successfully).

Appendix A – FAQ

What if the PDF is scanned and not searchable? The extractor cannot read scanned images. The file should be re‑issued as a searchable PDF, or the data must be entered manually.
How are categories chosen? Categories are selected by matching words in the description to a keyword‑category list (see Appendix C). The first keyword that matches decides the category.
Can the process handle different date formats? Yes. Common formats like DD/MM/YYYY, MM/DD/YYYY, or YYYY‑MM‑DD are recognized. If a date cannot be parsed, the row is flagged for review.
What if the amount column contains text like “$1,234.56” or “(1,234.56)”? The extractor removes the dollar sign, commas, and parentheses, then converts the text to a numeric value. Parentheses indicate a negative amount.
What should I do if a row is flagged in the Validation Report? Review the flagged row, correct the missing or incorrect data in the source PDF, then re‑run the extractor or manually add the row to the transaction list.
Can the extractor handle multiple pages? Yes. The process reads all pages in order. Transactions on later pages are added after earlier pages.
How are duplicate transactions handled? If two rows have identical date, description, and amount, the later one is flagged as a duplicate. Review manually to decide whether it should be kept.
What if the description contains a check number (e.g., “Check #1023”)? The check number is copied into the Notes field, while the Description stays unchanged.
What if the transaction amount is zero? Zero‑valued rows are kept but labelled with "Other" in the Category field.

Appendix B – Glossary

Term	Definition
Deposit	Money coming into the account (positive amount).
Withdrawal	Money going out of the account (negative amount).
Transaction Type	Indicates whether a line is a Deposit or a Withdrawal.
Category	A short label describing the nature of the transaction (e.g., “Office Supplies”).
Notes	Any additional information extracted from the description (e.g., check number).
Validation Report	A list of rows that could not be processed, with reasons.
Keyword‑to‑Category mapping	A list that pairs words or phrases (e.g., “Coffee”) with a Category (e.g., “Meals & Entertainment”).
Text‑searchable PDF	A PDF that contains selectable, searchable text (not a scanned image).

Appendix C – Category Mapping & Reference Materials

C1. Category Keyword Table

Category	Example Keywords (any of these trigger the category)
Salary	“salary”, “payroll”, “wages”, “paycheck”, “income”
Interest	“interest”, “interest earned”, “interest income”
Transfer	“transfer”, “to savings”, “from savings”, “between accounts”
Office Supplies	“office”, “supplies”, “stationery”, “printer”, “paper”
Meals & Entertainment	“coffee”, “restaurant”, “dinner”, “lunch”, “cafe”, “entertainment”
Travel	“airline”, “flight”, “hotel”, “taxi”, “uber”, “lyft”, “travel”
Utilities	“electricity”, “gas”, “water”, “utility”, “bill”, “utility bill”
Rent	“rent”, “lease”, “rent payment”
Insurance	“insurance”, “policy”, “premium”
Tax	“tax”, “taxes”, “tax payment”
Credit Card Payment	“credit card”, “credit card payment”, “card payment”
Loan Payment	“loan”, “loan payment”, “mortgage”
Other	(No match)

How it works – When a transaction description contains any of the keywords listed for a category, that category is assigned. The matching is case‑insensitive and looks for the keyword anywhere in the description.

C2. Formatting Guidelines for Output

Date – always YYYY‑MM‑DD.
Amount – use a decimal point with two digits (e.g., -120.45).
Transaction Type – exactly "Deposit" or "Withdrawal" (capitalized).
Category – one of the categories listed in C1, or "Other" if none apply.
Notes – optional free text; if empty, leave blank (no placeholder).

C3. Common Edge‑Case Table

Situation	How to Handle
Amount shown as `-$50.00` (negative sign and dollar sign)	Remove the `$` and keep the sign: `-50.00`.
Amount shown as `($50.00)` (parentheses)	Convert to `-50.00`.
Description is blank but amount present	Flag for review; do not output.
Date is a text month (e.g., “July 5, 2024”)	Convert to `2024‑07‑05`; if not possible, flag for review.
Multiple dates on same line (e.g., “01/03/2024 – 01/04/2024”)	Use the first date; if ambiguous, flag for review.
Amount includes “USD” or “CAD” symbol	Remove the currency symbol; treat as numeric.
Balance column present	Use it only to double‑check the cumulative sum; do not include in output.
Transaction description has multiple potential categories (e.g., “Taxi & Hotel”)	Choose the first category found in the description order; if none, use `"Other"`.
Very large amount (e.g., `$1,200,000.00`)	Keep the full number; ensure two decimal places.
Amount contains spaces (e.g., `$ 1,000.00`)	Remove spaces and symbols, keep numeric value.
Duplicate transaction (same date, description, amount) appears more than once	Flag the second occurrence as a duplicate; do not output it.

Tip – When in doubt about a category, choose "Other" and add a note explaining the uncertainty.

We build it

Try me

Why manual entry is a hidden cost

Every month, accountants and bookkeepers wrestle with three recurring frustrations:

Tedious copy‑and‑paste that steals productive hours.
Transcription errors that ripple through financial reports.
Inconsistent formatting that forces extra cleaning before the data can be used.

These hidden costs compound, especially when statements arrive from multiple banks or for multiple entities.

The Logic advantage

Logic’s AI‑driven extractor transforms a searchable PDF into a ready‑to‑use transaction list in seconds, letting you focus on insight rather than input.

Consistent date formatting – every date appears as YYYY‑MM‑DD.

Automatic credit/debit labeling – deposits and withdrawals are clearly identified.

Smart categorization – descriptions are matched to a curated keyword set, delivering instant, meaningful tags.

Built‑in validation – rows with missing or ambiguous data are flagged for quick review, safeguarding accuracy.

Key Insight

Uniform, structured transaction data removes the bottleneck that typically stalls month‑end close, freeing you to analyze trends instead of correcting entry errors.

What the extractor delivers

Feature	How it helps you
Date normalization	Guarantees sortable, comparable timelines across all statements
Transaction type labeling	Instantly separates inflows from outflows for clearer cash‑flow views
Category assignment	Provides ready‑made tags for expense reports and budgeting
Validation report	Highlights problematic rows so you can resolve issues before they affect downstream processes
Plain‑text table output	Easy import into any bookkeeping or ERP system without extra conversion steps

Who benefits most

Accountants who need reliable data for audit trails and client deliverables.
Bookkeepers who manage high volumes of statements across multiple accounts.
Finance managers looking for clean data to feed dashboards and forecasting models.

When the extractor handles the heavy lifting, you regain valuable time for analysis, decision‑making, and strategic planning. The result is a smoother workflow, fewer errors, and financial records you can trust.

Ready to Automate?

Get started with this workflow template in minutes. No complex setup required.

View Documentation