Frequently Asked Questions

Our parser stack combines pattern templates (for popular banks) with AI-assisted line detection and reconciliation checks. We validate rows against running balances, enforce date/amount normalisation (locale-aware), and auto-repair common issues (wrapped descriptions, thousand-separator drift). On clean, digital PDFs we expect 95–99% row-level accuracy; on scans we degrade gracefully via OCR with confidence thresholds and surface a "review required" banner when needed.

Both. We first attempt text extraction. If a page is image-only, we fall back to OCR. Scans under ~150 DPI or skewed >10° may need re-scans; we flag low-confidence lines instead of guessing.

We de-wrap using regex + whitespace heuristics, merge continuation lines when the amount/date pattern repeats, and strip OCR artifacts (e.g., `,` vs `.` confusion). We never fabricate rows.

Yes. We compute and validate running balances when present and export opening/closing balances as metadata.

Choose Signed Amount (default) or Debit/Credit columns in export settings. We normalise `−` vs `()` vs `DR/CR` styles.

Select MM/DD/YYYY, DD/MM/YYYY, or YYYY-MM-DD (ISO 8601). We also detect columnar day/month swaps to avoid import errors.

We auto-deskew/rotate; stamps and heavy watermarks are masked before OCR. Edge-cases are flagged for review.

Yes—if you know the password. Enter it at upload and we decrypt in memory for processing.

Default limits: 50MB and ≤200 pages per file (configurable on paid tiers). Batch mode supports multiple files.

Yes. Use Batch Upload in the converter or the API with webhooks for large jobs.

Banking PDFs mix headers, totals, footers, and multi-line rows. Generic exporters flatten everything into columns causing column-shift and mis-typed values. We use bank-aware parsing and reconciliation logic designed for statements, not invoices or forms.

CSV for imports to QBO/Xero and spreadsheets. XLSX when styling/formulas matter post-import. JSON for developers and downstream automation.

Use our QBO/Xero-ready presets or lock columns in the converter. Column shift usually means the source PDF was scanned with noise; try OCR cleanup or the signed amount preset.

Locale-aware parsing converts `1.234,56` ↔ `1,234.56` correctly. Choose your output locale at export.

Yes—use Batch Upload or the API for automation.

Yes. Save a custom preset per client/bank.

Yes, we stabilise by date/time and original sequence; you can reverse order if needed for imports.

Use File → Import → Upload and set Detect locale off; pick your date format. Avoid auto-type conversion that changes account numbers.

Default QBO headers: `Date, Description, Amount` (signed) or `Date, Description, Debit, Credit`. Optional: `Payee, Memo, Reference, Check #, Balance`. All presets validated to pass QBO's importer without edits.

Default Xero headers: `Date, Amount, Description` (signed), optional `Payee, Reference, Cheque Number, Balance`. Presets align with Xero's bank import rules.

Most failures are (a) wrong date format, (b) mixed `.`/`,` decimal symbols, or (c) DR/CR when QBO expects signed amounts. Re-export with the QBO preset and ISO date.

Align to DD/MM/YYYY or ISO for EU/UK orgs; ensure currency matches the bank account in Xero; avoid thousands separators on export.

Signed Amount is simpler and less error-prone. Use DR/CR only if your downstream tooling requires it.

We optionally emit a stable transaction hash (date, amount, last-4 reference, normalized text) to help deduplicate across periods.

Yes—save mappings per bank in your custom preset; applies in batch/API too.

We export memo and reference fields; category/class tagging is usually applied inside QBO/Xero rules after import.

Our infrastructure is hosted in the United States. Files are processed on our servers and encrypted in transit (TLS 1.2+) and at rest (AES-256).

Yes. Default is delete-after-processing. Paid users may opt-in to retain files and outputs for convenience; you can delete anytime.

TLS 1.2+ for transport; AES-256 for storage. Keys are rotated and access is strictly limited and logged.

No, except when you explicitly open a support ticket and grant consent for a least-privilege, time-bound review.

Yes. You can request access, correction, deletion, or export by emailing privacy@bankstatementsavvy.com or via your account. We honour Do Not Sell/Share.

No. We do not use customer files to train foundation models.

Yes—delete from your dashboard or email support. We'll confirm deletion and purge from backups on their next rotation.

For Business/Enterprise, we offer an NDA and Data Processing Addendum.

Customers report saving 1–3 hours per statement—more on messy scans—by avoiding manual cleanup and failed imports.

Yes—use the QBO/Xero presets and locale matching. If an import fails, re-export with the guided preset; our support can review sample rows.

Exports include normalized dates/amounts, optional balance column, and consistent Payee/Description mapping—ideal for income/expense analysis.

Yes—upload folders in batch or use the API with webhooks for async processing at scale.

Business plans support client folders, tags, and team seats.

Yes—use tags and metadata on upload; searchable in your dashboard.

We log upload → processing → export events with timestamps; exports are deterministic given the same preset.

Yes—currency is detected and preserved; you can force output currency codes if needed.

Yes—trial starts on signup; the card is authorised but not charged until the trial ends. Cancel anytime in the dashboard.

See our Pricing page for page credits, batch/API access, and seats per tier.

On annual plans, yes (within reasonable limits); monthly plans reset each cycle.

We occasionally offer one-time credit packs. Lifetime deals are limited promos.

If exports fail due to our fault and we can't fix promptly, we'll credit or refund.

Yes—upgrades prorate instantly; cancellations take effect next cycle.

Yes—contact support@bankstatementsavvy.com.

Priority support for paid tiers; Business/Enterprise can add SLA/DPAs.