How accurate is the PDF invoice data extraction and what should I always verify?

When you upload a utility invoice — whether manually or via an email attachment sent through the system — the platform uses AI-powered OCR to extract data automatically. This extraction is a best-effort process and is not guaranteed to be correct in every case.

What can go wrong during extraction

• Meter number misread: Digits in the meter number may be confused (e.g., 0 vs O, 1 vs l). This can cause the invoice to link to the wrong meter or remain unlinked.
• Consumption value wrong: The OCR may read the wrong line item as consumption, miss a unit conversion, or fail entirely on complex invoice layouts. The result can be zero, NULL, or a wildly incorrect value.
• Wrong billing period: Dates may be misread, especially when invoices use non-standard date formats or multiple date fields appear on the same page.
• VAT extracted incorrectly: VAT rate and amounts are inferred from the document; if the layout is unusual, the system may default to 25% or read the wrong figure.
• Scanned/image PDFs: If the invoice is a scanned image rather than a digitally generated PDF, OCR quality drops significantly. Handwritten or low-resolution invoices have higher error rates.
• Multi-page or multi-meter invoices: Only the primary meter number is captured. If your invoice covers several meters, only one will be auto-linked.

What you must always verify after uploading an invoice

Open Invoices → click the invoice → compare every extracted field against the original document:

1. Meter number — does it exactly match what is printed on the invoice?
2. Consumption value — does the extracted kWh / m³ match the total shown on the invoice for that billing period?
3. Billing period — do the start and end dates match the invoice?
4. Total amount and VAT — do these match the invoice to the cent?
5. Linked meter — is the invoice linked to the correct meter in the correct building?

Do not use extracted data for billing without verifying

Invoice data that flows into cost allocation and tenant invoices is only as accurate as what was extracted. An incorrect consumption figure will silently produce wrong tenant charges. The platform displays extracted data as-is — it does not re-check against your utility provider.

Correcting extraction errors

If you find a discrepancy, open the invoice detail page and use the edit fields to correct the extracted values. Changes take effect immediately and will be reflected in the Energy dashboard and cost allocation calculations.

When invoices arrive as email attachments

If you use the email import feature and invoices are sent as PDF attachments, the same OCR process applies. The automated nature of email-based import means errors may go unnoticed for longer. Establish a routine of reviewing newly imported invoices in the Invoices page after each import cycle to catch and correct any extraction issues before they reach billing.