The method in plain language
Evidence comes from the statement itself. Analysis is only a working layer and must remain verifiable.
Read the source
Text or page imagery is extracted without replacing the transaction wording supplied by the bank.
Check the arithmetic
Credits and debits are summed separately, after which their net effect is checked against legible balance changes.
Label uncertainty
Printed facts remain distinct from calculated categories, repeated-activity signals, and interpretations of cash flow.
Source hierarchy
Legible content becomes rows linked to its source, with bank descriptions kept intact. Every supported format faces the same verification threshold. Before calculating from photographed or scanned pages, use the scanned-statement reader. Testing and training can instead use a watermarked synthetic bank statement, which has no value as financial evidence. An illegible field becomes null with a warning; neighboring patterns are never used to fill it.
Transaction structure
A row may include posting date, bank wording, possible merchant, debit, credit, balance, category, confidence, lookup support, and reviewer note. Standardized fields sit beside the unchanged description, preserving the evidence underneath.
Merchant lookup and category assignment
Unknown, uncategorized, or uncertain debit wording receives a merchant or category suggestion only when inspectable evidence supports it. Identification is performed with reduced descriptor text; the file, date, value, balance, account identifier, and profile are excluded.
Applying a category requires at least medium confidence plus one valid evidence link. Beside the untouched statement text, results identify the possible merchant, category, confidence, rationale, and source domain. Masked wording, marketplaces, contradictions, and unsupported matches stay unresolved for a person to assess.
Reconciliation checks
Debit and credit totals are calculated apart, with net movement defined as credits less debits. If both endpoint balances can be read, compare the printed change after accounting for interest, fees, reversals, and adjustments.
Browser worksheet calculations
The spending analyzer, monthly expense calculator, and financial ratio check use only the numbers entered in the browser. Each worksheet shows its formula, example, input limits, and interpretation boundaries. They do not retrieve account data, infer missing values, apply universal financial targets, or turn an arithmetic result into advice.
Statement review pack
The private workpaper organizes account identity, covered period, balance reconciliation, debit and credit composition, extraction confidence, unresolved categories, possible duplicates, warnings, and references to source rows. Where a page is identifiable, its number follows the transaction into exception exports.
A duplicate flag indicates only that selected fields match. Likewise, zero variance shows merely that extracted arithmetic closes within tolerance. Neither finding proves authenticity, ownership, business purpose, supporting contracts or receipts, or an audit conclusion.
When bank wording lacks context, users may add one statement note and notes for individual entries. Both screens and exports identify these additions as user-provided. Notes cannot replace source wording or warnings, and they do not establish business purpose or tax treatment.
Signed-in reviewers may retain six evidence confirmations and their notes, none of which are checked automatically. The Excel workpaper carries forward the displayed summary, ledger, reconciliation, automated controls, exceptions, evidence checklist, and methodological boundaries.
Interpretation boundaries
Merchant identities, categories, and recurring candidates are interpretations. Online evidence may describe a business generally, yet cannot establish the item purchased, the authorizing person, fraud, current subscription status, or a transaction's tax treatment.
Tax-review candidates and estimates
The tax area searches debit wording and reviewed categories for entries that might warrant business-expense assessment; it never calls them deductible. Allowable share starts at zero because statements do not prove taxpayer status, purpose, attendees, reimbursement, personal use, supporting evidence, or applicable law for a place and year.
For each chosen entry, supply its jurisdiction, tax year, verified marginal rate, and permitted proportion. The browser first applies that proportion to the value and then applies the rate. This scenario promises no refund. It excludes credits and indirect-tax recovery, as well as capital allowances, thresholds, losses, phaseouts, and local levies.
Built-in notes point to official publications. United States material comes from the United States Internal Revenue Service. For British rules, material comes from United Kingdom HM Revenue and Customs. Further sources are the Canada Revenue Agency, Australian Taxation Office, and Hong Kong Inland Revenue Department. Elsewhere, confirm current rules from the jurisdiction's own authority before calculating.
Quality controls
Every page needs a descriptive H1 matching its title, correct anchors, sourced claims, and copy tailored to page intent. Imported work remains outside search indexes during editorial review. Thin or duplicative pages are rewritten or withheld.
Privacy and retention
Encrypted temporary input and output remain bound to the browser session for 30 minutes, during which a profile can be created from the page. Signed-in users choose whether to retain a statement. Free storage ends after 30 days; Premium has no scheduled deletion. Either may be removed sooner. Complete handling details are in the upload and storage privacy policy.
Apply and verify the method
Run the online bank statement analysis tool first. Next, inspect coverage and uncertainty through the extraction and data-quality learning path. Address differences with the bank statement reconciliation guide, and check images through the OCR quality checklist. Turn those controls into a repeatable release routine with the bank statement review checklist. Longer evidence reviews and editorial changes are available in Bank Statement Research.