How to Convert a Bank Statement to CSV

Choose the source that matches your goal

To convert a bank statement to CSV, first use your bank's own transaction download when it covers the exact account and statement period you need. If the issued PDF is the source of record—or the portal no longer offers that period—convert the PDF into one row per printed transaction, then compare the CSV with the PDF before using it.

A bank download and an issued statement are not always interchangeable. The download can include pending activity, use a custom date range, or omit statement-only boundaries. Confirm the account, first and last dates, and transaction count before deciding that a portal CSV represents the same period as the PDF.

Use one transaction row and three core columns

For a general-purpose transaction export, use one header row and three columns: Date, Description, and Amount. Each genuine transaction gets its own row. Page headers, statement summaries, subtotals, advertising text, and blank separators do not belong inside the transaction ledger.

Keep Amount signed and numeric. In this guide's convention, purchases, withdrawals, and fees are negative; deposits, payments, refunds, and credits are positive. Do not store a dollar sign or thousands separator inside the raw amount value. A destination that requires separate debit and credit columns can map the signed Amount later using its own documented import rules.

Quote commas and quotation marks correctly

A CSV is not made by joining fields with commas and hoping for the best. The common conventions documented in RFC 4180 require records to have the same number of fields. A field containing a comma, line break, or double quotation mark should be enclosed in double quotation marks, and a double quotation mark inside that field is represented by two double quotation marks.

That rule matters for ordinary bank descriptions. A merchant such as Coffee, Market Street contains a comma but is still one Description value. The synthetic examples below show the raw lines a CSV writer should create; the quotation marks are file structure and should not appear as part of the description after import.

Save as UTF-8 and keep dates unambiguous

Save the file as UTF-8 so accented characters, non-English merchant names, and punctuation such as an em dash survive. Microsoft notes that Excel opens a UTF-8 CSV normally when it includes a byte-order mark; otherwise, Data > Get Data > From File > From Text/CSV lets you select the encoding explicitly.

Use a date representation the destination can interpret consistently. ISO values such as 2026-08-03 are unambiguous for data exchange. If the statement prints only month and day, derive the year from the statement period and printed order—not from the upload date. Pay special attention to statements that cross December and January.

• Keep descriptions as text and preserve their complete source wording.
• Use a decimal point and no currency symbol in Amount.
• Leave an unavailable value empty rather than replacing it with zero.
• Preserve the original transaction order until the extraction has been reviewed.

Convert the PDF without flattening its meaning

A text-based PDF may expose selectable characters, while a scanned statement first needs text recognition. Either way, the conversion must reconstruct transaction rows rather than copy every visible line. Wrapped merchant text belongs to the same transaction; repeated page headers and printed totals do not become purchases or deposits.

StatementMint converts entitled pages from an issued PDF into Date, Description, and signed Amount, then offers CSV and formatted Excel for a completed full statement. The CSV is generated with comma and quotation-mark escaping, CRLF record endings, UTF-8 text, and spreadsheet-safe descriptions. Automated extraction can still omit, duplicate, misdate, or mis-sign a transaction, so inspect the result against the PDF.

Open the CSV deliberately in its destination

Test the finished file in the system that will consume it. In Excel, Data > Get Data > From File > From Text/CSV shows the detected delimiter, headers, and types before loading. Microsoft documents that Power Query automatically detects these elements, but automatic inference can still reinterpret an ambiguous date or remove leading zeros from an identifier.

If you are importing into accounting software, follow that product's current field-mapping instructions rather than renaming columns by guesswork. A clean source file is not a promise that every destination accepts the same schema. Keep the untouched CSV so you can remap it without repeatedly converting the PDF.

Review the rows before relying on them

Compare the account and period, count transactions, and spot-check the first and last row on every transaction page. Inspect the largest positive and negative amounts, same-day repeated transactions, descriptions that wrap across lines, and dates near a year boundary. Confirm that filters are cleared before counting.

A syntactically perfect CSV can still be financially incomplete. Even a matching net sum is not proof of completeness because a missing positive and negative amount can offset each other. The issued statement remains the authority. Convert one statement, inspect Date, Description, and Amount against the source, and report a problem from the conversion page if a row is wrong.

Frequently asked questions