How to Extract French Expense Report
Data into Excel (2026 Guide)

Why French Expense Reports Need Different Extraction Logic

A French expense report (note de frais) carries data obligations that standard receipt-scanning tools were never designed to handle. Unlike a simple till receipt — where one line means one amount — a single note de frais can bundle a train ticket (TVA non-recoverable), a péage receipt (TVA 100% recoverable), a client restaurant bill (TVA recoverable at 20%), and a mileage claim calculated against the tax authority's barème kilométrique (kilometric scale). Each line needs a different compte comptable (account code) under the Plan Comptable Général (PCG, the French national chart of accounts), and each has its own rule for whether and at what rate the taxe sur la valeur ajoutée (VAT) is recoverable — codified in the Code Général des Impôts (CGI, the French General Tax Code).

Generic OCR tools flatten this distinction. They extract "amount" and "date" — which is fine for a simple reimbursement list — but they don't know that a gasoil receipt entitles your company to 80% TVA recovery while an essence receipt gets 0%, or that the barème kilométrique rate depends on the vehicle's puissance fiscale (fiscal horsepower) and annual distance tier. A tool needs to understand that these aren't just data points — they're transaction-level decisions with tax consequences.

The Fields That Matter for French Accounting Compliance

Before extracting anything, you need to know which fields a French accountant actually needs — and which ones carry tax implications. While there is no legally mandated template for a note de frais, the expense report must provide enough information to justify the expense to both the company and, in the event of a contrôle URSSAF (social security audit), to the authorities.

The comptabilisation (accounting entry) for a typical note de frais with a 100 € HT hotel night and a 40 € HT client restaurant meal looks like this in the general ledger: debit 6256 (Missions) for 100 €, debit 6257 (Réceptions) for 40 €, credit 421 (Rémunérations dues) for 150.60 € TTC. The hotel line has no TVA recovery (excluded under Article 206-IV-2-2° of Annexe II of the CGI); the restaurant line carries 8 € of recoverable TVA at 20%, debited to compte 44566. If your extraction output doesn't separate these correctly, someone downstream has to redo it.

Step-by-Step: Extracting Note de frais Data into Excel

The extraction process itself takes under a minute per report — what matters is setting up your columns so the output arrives structured for accounting, not as a flat data dump that needs an hour of reformatting in Excel.

The approach described here uses Custom Column Extraction: instead of training a template on sample documents or drawing rectangles around fields, you type the column names you want — "Employee Name", "Date", "Amount HT", "TVA Rate", "PCG Code" — and the AI locates each value anywhere on the document by understanding what the field means, not where it sits on the page. This is fundamentally different from template-based OCR, which breaks when a report layout changes or a receipt is photographed at an angle.

For teams that receive expense reports from multiple people — field staff, sales representatives, remote employees — a Collection Link removes the email-back-and-forth entirely. You generate a shareable link (e.g., /c/xxxx), send it to each person who needs to submit expenses, and they upload their scanned note de frais and justificatifs directly into your processing queue — no registration or login required on their end. Files arrive in your account ready for batch extraction. This is covered in detail in our collection workflow guide.

Making TVA and Kilometric Allowances Work in Your Output

The hardest part of processing French expense reports isn't the extraction — it's getting the TVA and indemnité kilométrique calculations right for each expense type. Under Article 271 of the CGI, the general principle is that VAT on goods and services acquired for taxable operations is recoverable — but a cascade of specific exclusions in Annexe II of the CGI creates a patchwork that trips up both employees filling out reports and accountants coding entries.

For kilometric allowances, the calculation depends on the vehicle's puissance fiscale and total annual distance tier. A 4 CV vehicle up to 5,000 km: 0.606 €/km. Between 5,001 and 20,000 km: (d × 0.340) + 1,330 €. Beyond 20,000 km: 0.407 €/km. With a computed column, you define the formula once: extract the distance and CV from the report, and the AI calculates the indemnité directly in the output — no manual reference to the barème table required. For electric vehicles, apply the 20% majoration (increase) automatically by incorporating it into the computation rule. The barème for 2026 remains unchanged from 2025 and 2024; the last revaluation was in 2023 (+5.4%).

Watch the URSSAF ceilings: if your company reimburses above the barème rate, the excess is reintegrated into the social contribution base. For meals, the URSSAF exemption ceiling is approximately 21.10 € per meal (2026). For the forfait mobilités durables (sustainable mobility allowance), the ceiling is 600 € per year per employee, or 900 € when cumulated with public transport subscription reimbursement.

From Extracted Excel to Pennylane, Sage, or Cegid

An Excel file with correctly coded expense lines is halfway into your French accounting software — the remaining step depends on which platform you use, but the column structure you set up during extraction determines how frictionless this handoff is.

For Pennylane, the CSV import accepts expense lines mapped to PCG accounts — your output columns for Amount HT, TVA Amount, and Account Code slot directly into the journal des achats. Sage 100 and Cegid Loop support CSV imports with configurable field mapping, meaning your extraction column headers become the mapping reference. EBP Compta includes a native note de frais module that can import structured data with TVA breakdowns. For firms using Quadratus or Sage FRP 1000, the CSV output follows the same logic — structured rows with the correct account codes.

If your accounting workflow demands a different format, the same extraction can export to JSON for custom integrations — useful for firms with in-house ERP connectors or those using N2F, Cleemy, or Expensya as their expense management layer, where the extraction tool handles the data capture step upstream and feeds structured data into the existing platform.

For companies processing French invoices alongside expense reports, the same column-based extraction approach works on factures (invoices) — covered in our guide to French invoice data extraction, which covers SIREN/SIRET and TVA intracommunautaire fields. If a significant portion of your expense reports involves mileage, our guide on French payslip extraction covers URSSAF contribution handling that overlaps with expense-related social charge rules.

FAQ

Can AI distinguish between a French expense report and an English one?

Yes. The vision model reads the document content regardless of language — a receipt from a Paris restaurant is processed with the same accuracy as one from London. Define your column names in English and the AI extracts the corresponding values. For French-specific fields like TVA or Montant TTC, naming your column "TVA Amount" tells the AI to find the tax line — it doesn't need the document to be in English.

What about handwritten French expense reports?

The vision model handles handwritten text — including French cursive — across expense reports and handwritten justificatifs. Accuracy on handwriting is lower than on printed text (as with any OCR system), so a quick review pass is recommended for fully handwritten documents. For mixed printed-and-handwritten reports — the most common real-world format — accuracy is consistently high on the printed sections.

Does the tool file my TVA return?

No. The tool extracts and categorizes TVA amounts per expense line, producing output where each row carries its TVA treatment. The extracted data feeds into your TVA declaration — whether prepared in Excel, by your accountant, or through your accounting platform — but the actual filing (déclaration CA3 or CA12) remains your responsibility. The value is that TVA amounts are already separated from HT amounts per line, correctly mapped to compte 44566, eliminating the step where someone manually determines which expense lines carry recoverable TVA.

Can the tool apply the barème kilométrique automatically?

Yes, using computed columns. Define a column like Indemnité Kilométrique (if CV=4 and distance ≤5000 then distance×0.606; if CV=4 and distance>5000 and distance≤20000 then distance×0.340+1330; if CV=4 and distance>20000 then distance×0.407). The AI reads the distance and CV from the expense report and outputs the calculated indemnité. You can build separate formulas per CV tier or use a more compact expression. When the barème changes in a future year, you update the column definition — all subsequent extractions use the new rates.

What if an expense report has missing justificatifs?

The AI extracts whatever is present on the uploaded documents. If a justificatif is missing — say, a hotel invoice wasn't attached — the corresponding line will show the employee-entered information from the note de frais but lack the verified Amount HT and TVA breakdown that only a justificatif can provide. You'll see this during the review step. French accounting practice requires an attestation sur l'honneur (sworn statement) for lost receipts under certain conditions, but that workflow is separate from extraction.

Is my expense report data secure?

Files are processed through encrypted connections and are not stored after extraction completes. The output Excel file is downloaded directly to your computer. If you need to collect expense reports from multiple employees without sharing login credentials, the Collection Link feature generates a unique upload URL with verification code access — each uploader sees only their own files, and all submissions land in your processing queue.

What Changes When You Stop Manually Typing French Expense Reports

The gap between a scanned note de frais and a properly coded accounting entry has been a manual bottleneck for French finance teams — not because the data is hard to read, but because the rules differ by expense type and generic tools flatten those distinctions into a single "amount" column. When extraction is set up around French accounting logic — TVA per expense type, PCG codes assigned by nature de dépense, indemnités kilométriques calculated from the barème rather than typed from memory — the output arrives coded for the journal des achats, not requiring someone to re-categorize every line in Excel.

Try it on your own notes de frais. See if the stack that usually fills an afternoon becomes a 5-minute review — where you're checking numbers, not creating them.