E-Invoice Finder
Menu
CountriesCompare softwareDeadlinesGuidesCheckerCheck obligation
ENFRDEPLNL
France · post-launch PA review

How to review a French approved platform after go-live

Assess a French approved platform with live delivery, support, ERP reconciliation and workload evidence before deciding to retain, fix or switch.

Quick verdict:
  • • Coverage floor: capture all anomalies plus a cross-day control cohort spanning each high-impact production pathway.
  • • Proof challenge: independently join one invoice’s origin record to its booked result without trusting screen labels.
  • • Choice gate: keep what is controllable, correct what is repairable, invoke contract remedies for broken commitments, and plan a provider transition only when remaining exposure breaches tolerance.
Get an independent France performance reviewOfficial sources
Last checked: 6 September 2026Based on official sourcesClear summaryBusiness guidance, not legal advice
Official sources prioritized
Last-checked dates visible
Free checker, no signup required

What you need to know

Guide

1. Review fitness after the first live flows—not registration alone

France’s reform is now live, so this review should begin as soon as a representative set of genuine invoices has crossed the production chain. Service-Public reported on 2 September 2026 that the reform took effect on 1 September 2026: affected businesses must be able to receive electronic invoices, while large enterprises and ETIs also begin issuing and e-reporting from that date. SMEs and micro-enterprises begin issuing and e-reporting on 1 September 2027. The same official update advises businesses to map flows, adapt procedures and choose a plateforme agréée (PA) suited to their needs. A review is therefore not an abstract software exercise; it asks whether the live operating arrangement fits the company’s obligations, systems and trading patterns. DGFiP explains that affected businesses use an approved platform to transmit and receive electronic invoices and send the required invoice, transaction and payment data. A PA may also convert a supplier invoice into a format suitable for its client while maintaining integrity, authenticity, readability and completeness. Only a registered PA can perform all regulated transmission, reception and reporting functions; a compatible solution still needs a PA for those functions. Confirm the provider’s current status on DGFiP’s official PA list, but do not confuse that status with proof that the service works well for your particular entities, addresses, formats, integrations, volumes or support model. Registration answers an essential regulatory eligibility question. Buyer fitness is a separate evidence question. It includes correct annuaire routing, reliable Factur-X, UBL or CII handling, usable lifecycle information, PA-to-ERP reconciliation, appropriate e-reporting where applicable, support ownership and sustainable manual effort. Start with a day-7 baseline once seven calendar days of live use are actually available, then use a rolling 30-day view as evidence accumulates. On 6 September 2026, a business that started on 1 September has not yet completed either window; it can prepare the controls and record observations without presenting an incomplete period as a finished result. This is an internal operational review, not a statutory audit or an official platform score.

Guide

2. Define the perimeter and build a defensible sample

Write down the review perimeter before looking at tickets or dashboard colours. List each legal entity by SIREN, the SIRET or establishment dimension used operationally, and every reasoned receiving address in scope. Split the population into inbound invoices, outbound invoices and e-reporting lanes; then identify the PA, compatible solution, ERP or accounting application, AP/AR workflow, connector and internal owner at each boundary. Add expected volume, actual live volume, value, tax or cash criticality, peak periods and any unusual format, currency, business unit or customer requirement. A group-level average can otherwise hide a broken route for a small but critical entity. Use a reproducible data extract rather than whichever cases are most memorable. The minimum useful day-7 sample should include every failed, quarantined, duplicated, missing or manually repaired item plus successful items from each live, business-critical lane. For higher-volume routes, add a documented sample of ordinary successes across different days, counterparties, formats and destinations. Keep the full population where exports make that practical. Record the extraction time, source system, query or filter, period covered and excluded records. A support anecdote, user complaint or screenshot may help identify a case, but it must not silently become the denominator for a performance rate. Create one row per business document and retain stable keys across systems: source document ID, invoice number, legal entities, SIREN/SIRET, reasoned address where used, amount, PA identifier or correlation ID, timestamps, format, lifecycle status, e-reporting reference where applicable, ERP document number and final disposition. These are recommended review fields, not a claim that every product or legal process uses one universal field list. Keep technical events, business outcomes and user observations in separate columns. Mark unknown values as unknown rather than converting them to failure or success. Segment before interpreting. A first-pass rate for outbound UBL invoices from one ERP should not be blended with inbound Factur-X conversion, e-reporting exceptions or a newly opened establishment address. Compare like with like by entity, route, format, connector and criticality. This allows the buyer to distinguish one bad mapping from a platform-wide weakness and ensures that a low-volume clean lane does not dilute a high-volume problem. Preserve both the day-7 snapshot and each rolling 30-day extract so later improvement claims can be reproduced.

Guide

3. Recommended internal 100-point PA scorecard (not official)

The following scorecard is a recommended buyer control, not a DGFiP requirement, legal standard, mandatory score or vendor certification. Score each category from 0 to 100 using documented evidence, multiply it by the proposed weight, and add the contributions. A category with no applicable live flow should be marked not applicable and its weight deliberately reallocated; it should not receive an automatic perfect score. Adapt the weights to your risk appetite, contractual commitments and route criticality before seeing the results. | Evidence category | Suggested weight | What earns confidence | |---|---:|---| | Regulated-route reliability and directory evidence | 22 | Correlated send/receive events, correct SIREN/SIRET and reasoned-address routing, dated annuaire evidence, controlled duplicate handling | | Data and format handling | 13 | Agreed Factur-X, UBL and CII profiles processed as expected; conversions remain traceable and complete; actionable validation evidence | | PA-to-ERP workflow and reconciliation | 17 | Connector acknowledgements, visible quarantine, stable identifiers, controlled replay and source-to-ledger matching | | Lifecycle statuses and e-reporting, where applicable | 12 | Status origin and meaning are clear; applicable reporting events reconcile; errors can be corrected and evidenced | | Incident and support quality | 13 | Prompt ownership, cross-party coordination, useful diagnosis, root cause, communication and contractual service evidence | | Evidence, security and reversibility | 13 | Exportable logs and configurations, role control, auditable changes, usable data extraction and credible exit readiness | | Total operating cost and manual workload | 10 | Transparent charges, measured manual touches, exception cost and predictable scaling | | **Total** | **100** | **Weighted evidence from the defined perimeter and period** | Use anchored scoring so reviewers apply the scale consistently. For example, 90–100 may mean complete evidence and dependable control for the observed sample; 70–89 may mean generally effective operation with bounded gaps; 50–69 may indicate material remediation; and below 50 may indicate an unacceptable control gap for that category. These bands are internal examples, not legal thresholds. A single severe issue—such as unresolved routing for a critical entity—may override the numerical total under the company’s risk policy. Conversely, a short observation period should lower confidence, not automatically lower the provider’s score. Require a citation for every category score: export name, report date, ticket ID, contract clause, reconciliation workbook or control result. Add a confidence rating based on sample breadth and data completeness. The weighted total supports a decision; it does not make one. Management should read the category pattern, critical exceptions, trend and evidence quality alongside the number. A 78 supported by complete route-level records may be more actionable than a 92 assembled from dashboard claims.

Guide

4. Calculate useful indicators without inventing legal targets

Define the denominator and success boundary before publishing any rate. For a regulated-route reliability measure, the denominator might be all unique documents released to the PA during the period, excluding documented test or cancelled records. Success might require correlated acceptance, intended routing and downstream receipt evidence by the agreed measurement cut-off. State whether duplicates are removed by source ID, supplier and invoice number, or another documented key. Never mix attempted API calls, files, business invoices and line items in one denominator. | Indicator | Recommended calculation | Evidence and caveat | |---|---|---| | First-pass acceptance | Unique documents accepted without correction ÷ unique first submissions × 100 | Separate document/data defects from transport and routing failures | | Duplicate incidence | Confirmed duplicate business documents ÷ unique intended documents × 100 | Count retries only when they create an additional actionable document, not merely another technical event | | Missing or late event rate | Documents lacking the defined correlated event by the contractual or internal cut-off ÷ eligible documents × 100 | There is no invented universal legal delivery target; publish the chosen cut-off | | Quarantine rate | Unique documents entering quarantine ÷ documents received at that boundary × 100 | Split by rule, entity, format and whether automatic recovery occurred | | Unmatched PA-to-ERP rate | PA-delivered documents without a matched ERP outcome at cut-off ÷ PA-delivered documents × 100 | Reconcile stable IDs and investigate parked or hidden records before labelling them missing | | Manual touches | Count of human interventions ÷ completed documents | Define a touch consistently; separate normal approval from avoidable repair | | Backlog ageing | Open exceptions grouped by age since the first failed or missing expected event | Use contractual and risk-based bands rather than claiming statutory grace periods | | Source-to-ledger reconciliation | Unique source documents with the expected ledger outcome ÷ eligible unique source documents × 100 | State accepted outcomes, such as posted, parked legitimately or rejected with controlled closure | Measure time-to-diagnosis separately from time-to-resolution. Diagnosis runs from detection to a supported identification of the failed boundary; resolution runs until the agreed business and system outcome is verified. A fast workaround with no root cause can improve apparent resolution while leaving recurrence risk. Report medians and high-percentile or worst critical cases where volume permits, because an average can conceal long-tail failures. For low-volume lanes, publish counts and case detail rather than a misleading percentage. Data quality needs its own exceptions log. Clock differences, time zones, overwritten statuses, one-to-many conversion envelopes, missing cross-system IDs and retrospective manual updates can distort sequence and duration. Note whether timestamps are source-generated or export-generated and whether the event history is immutable. Reconcile totals from source, PA, compatible solution, AP/AR and ERP; explain every difference rather than forcing the figures to match. FNFE-MPE professional guidance highlights reasoned addresses, PA connection, duplicate controls, strict treatment of Refusée, core Factur-X/UBL/CII formats, CDAR status messages, AFNOR XP Z12-012 and directory synchronisation. Use that guidance as a useful control reference, not as legislation or a source of invented pass rates.

Guide

5. Review incidents and support through one traceable case

Select at least one success and each material failure type, then ask support to trace them end to end. The test is whether the PA can use a correlation ID or equivalent stable identifiers to connect intake, validation, directory decision, transmission, receipt, conversion, client delivery, lifecycle messages and any applicable e-reporting event. Ask for exportable events with timestamps and origin—not only a screenshot or a coloured status. Confirm who owns the next action when the path crosses the PA, a compatible solution, the ERP, the buyer’s configuration or another PA. A practical ticket could read: “Outbound invoice FR01-2026-1842, source ID AR-90877, buyer SIREN 123456789, submitted 4 September at 10:14 Europe/Paris. Our PA shows acceptance under correlation C-78421, but the customer reports no receipt. Please provide the timestamped routing destination and address used, outbound event, recipient acknowledgement or error, status definitions, current owner, and whether a controlled replay would preserve identity and duplicate protection. Please cite the applicable contractual severity and response record.” This asks for a bounded diagnosis without presuming blame or revealing more invoice content than support needs. Evaluate the answer against the contract and actual evidence. Did support acknowledge the correct severity? Did it distinguish initial response, diagnosis, workaround and verified resolution? Could it coordinate with the compatible solution, ERP team or other PA rather than repeatedly sending the buyer elsewhere? Did the root-cause report name the failed control, explain why monitoring did not detect it, identify affected scope and assign a preventive action with an owner? Compare timestamps with contractual service commitments, but do not describe a commercial support SLA, availability promise, incident deadline, RTO or RPO as statutory unless a current official source expressly says so. Common mistakes are opening duplicate tickets with inconsistent facts, accepting “delivered” without the boundary definition, resetting a case after every hand-off, using Refusée as a generic technical error, allowing an uncontrolled resend, and closing when the dashboard turns green without reconciling the ERP. Test replay in a controlled case and verify that duplicate controls prevent a second actionable invoice. Record communication quality as evidence: clear impact updates and named ownership matter, but polished messages do not substitute for event exports, root cause or durable correction.

Guide

6. Expose hidden workload and total operating cost

Review cost as an operating system, not a subscription line. Start with implementation and setup charges, recurring platform fees, invoice or transaction volume bands, connectors, legal entities, users, premium support, storage or archive services, format conversion and change requests. Map which charges are committed, variable, bundled, capped or indexed. Then add the internal costs created by the live service: AP/AR review, IT monitoring, master-data maintenance, support coordination, manual downloads, re-keying, duplicate investigations, failed-flow cleanup and reconciliation. Measure avoidable manual workload by exception type. A platform may have a competitive headline price yet require frequent intervention for reasoned-address routing, supplier matching, non-standard profiles, e-reporting errors or PA-to-ERP imports. Estimate cost as event count × median active handling minutes × loaded role cost, then add external accounting, integration and support fees. Keep waiting time separate from active labour, but show its working-capital or close impact where relevant. Document assumptions rather than presenting a false level of precision. Include costs that emerge only when evidence is needed. Can the buyer export event history, lifecycle records, configuration, directory-address mappings, invoices and reconciliation data without paid professional services? What archive or export arrangements apply, and which party can read the output? Do not invent a universal retention period; assess current legal, tax, accounting, security and contractual requirements with qualified advisers. Test reversibility by requesting a sample export and data dictionary. Add the expected effort to preserve continuity, remap integrations, retrain users and validate another provider if change becomes necessary. Buyer decision criteria should combine cost with control. Ask whether fees scale predictably with the expected 2027 population, whether critical support is included, whether manual touches are falling, whether exception ownership is clear, and whether exports reduce dependence on the provider. A higher contractual fee may be economically better if it measurably lowers repair effort and close risk; a low fee may remain attractive where flows are simple and internal controls are strong. Compare total operating cost per successfully reconciled business document, not price per submitted message alone. Use sensitivity ranges for future volume and workload, and do not treat switching cost as either zero or a reason to tolerate an uncorrected structural gap.

Guide

7. Decide: retain, remediate, formally escalate or prepare a switch

Start with defect classification. A configuration issue has a bounded correction within the existing design, such as a wrong reasoned address, role assignment or validation setting. A process defect may involve unclear ownership, uncontrolled retries or poor master-data governance. An integration defect sits at a documented boundary between the PA, compatible solution and ERP. A structural PA capability gap exists when a required function or control—such as usable route evidence, supported live format handling, cross-system correlation, controlled export or appropriate support coordination—cannot be delivered reliably under the intended operating model. **Retain** when critical routes work, evidence is complete, workload and cost are acceptable, and residual issues are minor with credible controls. **Remediate** when the cause is bounded, ownership is agreed, a dated correction and retest are feasible, and the provider can demonstrate prevention. **Formally escalate** when a material contractual commitment may not be met, repeated remediation stalls, evidence remains unavailable or senior cross-party ownership is needed. Preserve facts, contract clauses, impact and requested remedy; avoid public accusations or assumptions about fault. **Prepare a switch** when a structural gap remains material after fair validation, the risk exceeds tolerance, or remediation and escalation cannot produce an acceptable, evidenced operating state. Preparation means building options and continuity evidence, not making an automatic recommendation to move. Decree no. 2026-677 of 27 July 2026 codifies mobility safeguards in articles 242 nonies E bis to E quater of Annex II to the French General Tax Code. Cautiously summarised, the decree requires dated, signed formal consent identifying the old and new PAs, affected electronic addresses and effective date. After receiving it, the new PA has two working days to communicate its number to the old PA; the old PA has five working days to object where the request does not reflect the business’s will. Following express or tacit agreement, the new PA has fifteen working days to register the addressing in the directory. After the switch, the old PA maintains specified processing-status management for one year and, when requested, communicates available continuity information within five working days. Those provisions support controlled mobility; they do not prove that switching is the right remedy or eliminate contract, integration, testing and business-continuity work. Use the separate approved-platform migration guide for procedural depth. Before a decision, confirm the current decree text, DGFiP registration status, contractual exit terms, data rights, security requirements and tax or accounting interpretations with qualified advisers. Apply an executive override for severe risks, but document it. The defensible decision links route-level evidence, remediation feasibility, total cost and transition risk rather than reacting to a single incident or vendor comparison page.

Guide

8. Build the day-7 baseline and rolling day-30 evidence pack

Prepare the day-7 pack now, but close it only when seven complete live calendar days exist for the defined perimeter. For a 1 September go-live, that point is not reached on 6 September. Name one accountable review owner plus route owners from finance, tax, AP/AR, IT/integration, security and procurement or contract management. Include the perimeter, extraction method, sample coverage, scorecard, confidence ratings, issue register, critical case traces, reconciliation summary, manual workload and open decisions. Mark immature indicators clearly rather than annualising six days of behaviour. Turn every material finding into an action record with cause class, affected flows, impact, owner, due date, expected evidence and retest method. Retest the exact failed control: corrected annuaire routing with dated lookup evidence; format handling with agreed profiles; connector recovery with matched PA and ERP IDs; e-reporting correction with a reconciled outcome; or duplicate prevention after controlled replay. Keep the original failed evidence beside the passing retest. Review the contract for definitions, included services, response and resolution measures, availability calculation, exclusions, escalation path, data export, security obligations, change control, termination assistance and charges. Contractual values and company risk appetite—not invented legal targets—set the pass/fail boundaries. As live evidence accumulates, update a rolling 30-day dataset each day or reporting cycle. Preserve the day-7 baseline and version the score so management can see whether defects recur, backlog ages and manual effort falls. Because 30 live days have not elapsed by 6 September 2026, the first complete window for a 1 September start can only be assessed after the period finishes. Until then, label it “partial rolling view” and show counts, coverage and confidence. Do not imply that early stability proves month-end or volume-peak performance. Challenge providers in follow-up demonstrations with your anonymised failure patterns, not generic feature scripts. Ask them to trace a document, explain status provenance, export events, locate quarantine, show directory evidence, reconcile to an ERP outcome, prevent a duplicate replay and produce a reversibility sample. The next independent-review action is to have a reviewer reproduce the score from raw evidence and identify which conclusion would change if missing data became available. Management then chooses retain, remediate, contractual escalation or switch preparation by flow and entity, with a dated review after the first complete rolling window.

Guide

9. Fictional multi-entity example with a transparent score

Consider Groupe Mistral, a fictional group reviewing 1–5 September data on 6 September. It has Mistral Industrie SA (SIREN ending 101), which sends 1,200 outbound invoices from ERP-A; Mistral Services SAS (ending 202), which receives 420 invoices through three reasoned addresses into ERP-B; and Mistral Retail SAS (ending 303), which has 310 transaction records in an applicable e-reporting lane. The team correctly labels this a partial baseline, not a completed day-7 or 30-day review. It includes every exception and all available successes rather than sampling only help-desk cases. Three findings differ materially. First, Industrie has 1,176 first-pass acceptances, 18 corrected data failures and six route exceptions. Four route exceptions reflect an obsolete customer establishment address in ERP master data; two lack an exportable recipient acknowledgement after PA acceptance. The mistake was treating all six as PA delivery failures. The ERP-owned address problem is set for remediation, while the two uncorrelated cases are formally escalated for evidence. Second, Services has 414 matched ERP outcomes, four quarantined supplier records and two invoices delivered to the correct PA but hidden in the wrong AP workbasket. The mistake was measuring “not visible to central AP” as non-delivery. Role and address ownership are remediated; the inbound PA service is retained for this flow subject to retest. Third, Retail has 302 reconciled reporting outcomes and eight exceptions: five corrected mapping defects and three events whose support records do not establish final outcome. Repeated hand-offs between the PA and compatible solution make ownership unclear, so those three are escalated and the group begins contingency analysis for that lane only. Reviewers score the categories as follows: route reliability 78 × 22% = 17.16; data and format handling 86 × 13% = 11.18; PA-to-ERP workflow 74 × 17% = 12.58; lifecycle and applicable e-reporting 68 × 12% = 8.16; incident and support quality 55 × 13% = 7.15; evidence, security and reversibility 62 × 13% = 8.06; total cost and manual workload 71 × 10% = 7.10. The total is **71.39/100**. Confidence is medium because only five days are represented and several external acknowledgements are missing. Under Groupe Mistral’s internal policy, the total suggests bounded remediation, but support below 60 and unresolved critical evidence trigger contractual escalation. Those rules are company controls, not official French thresholds. The outcomes are deliberately flow-specific. Services is retained with permissions remediation because receipt, quarantine and ERP evidence are reproducible. Industrie retains its PA while correcting ERP master data, but escalates the two unsupported hand-offs with ticket and contract evidence. Retail keeps processing under enhanced controls while preparing a switch option: it requests exports, maps dependencies, checks the official PA list, estimates transition cost and uses the migration guide, without yet authorising a change. A switch decision would require failure to produce an acceptable outcome after the agreed escalation and retest, plus a safer validated alternative. The cost view changes the priority. Across the group, 67 avoidable manual touches at an assumed 11 active minutes each equal 12.3 hours before management, accountant and connector support time. Most touches come from Retail’s exception investigation, not Industrie’s higher volume. The team had initially compared subscription price per invoice and missed this concentration. Its next decision criteria are complete correlation for the three Retail events, passing retests for address and role controls, reduced rolling manual touches, exportable continuity data and contract-aligned support performance. At day 7 it will refresh counts; after the first complete 30-day window it will rescore with month-end evidence rather than extrapolating the partial result.

Checklist

Freeze the review perimeter by SIREN, SIRET or reasoned address, live lane, system, volume and criticality.

Confirm the PA’s current registration status on the official DGFiP list.

Export one document-level population with stable identifiers, timestamps, outcomes and source-system provenance.

Include every exception plus representative successes from each critical inbound, outbound and e-reporting lane.

Reconcile source, PA, compatible solution, AP/AR and ERP totals without silently discarding unmatched items.

Calculate defined indicators with published denominators, success boundaries, exclusions and data-quality caveats.

Trace material cases through support using correlation evidence, contractual service terms and named ownership.

Measure avoidable manual touches and add them to fees, integrations, accounting effort and change costs.

Assign each finding to retain, remediate, formal escalation or switch preparation, with an owner and retest.

Preserve the day-7 baseline and update a clearly labelled partial view until a complete rolling 30-day window exists.

FAQ

When should we review our French PA after go-live?

Begin evidence capture with the first production flows. Close a day-7 baseline only after seven complete live calendar days exist, then maintain a rolling 30-day view. A company live from 1 September cannot claim a complete seven-day or 30-day result on 6 September 2026. Review sooner after a critical incident, new entity, connector change or material volume shift, but label short periods and confidence honestly.

What data should we request from the PA and our systems?

Request document-level event exports with stable IDs, timestamps, legal entities, SIREN/SIRET and reasoned address where used, routing decision, validation and conversion results, lifecycle events, applicable e-reporting outcomes, quarantine and replay history, and client-delivery evidence. Match these to source and ERP records, contract reports, tickets and configuration changes. Ask for a data dictionary and extraction criteria so the result can be reproduced.

Does DGFiP registration prove that a PA is high quality for us?

No. Current appearance on DGFiP’s official list is essential evidence of registration status, and only a registered PA can perform all regulated transmission, reception and reporting functions. It does not demonstrate fitness for your particular addresses, formats, integrations, volumes, support expectations or cost model. Assess those with live route evidence, reconciliation, workload, contract performance and controlled retests.

Which PA performance metrics are sensible?

Useful measures include first-pass acceptance, duplicate incidence, missing or late defined events, quarantine, unmatched PA-to-ERP items, manual touches, backlog ageing, time-to-diagnosis, time-to-resolution and source-to-ledger reconciliation. Define the denominator, success boundary, duplicate rule, cut-off and exclusions for each. Segment by entity, lane, format and integration, and use counts rather than unstable percentages for small samples.

Is there a statutory support SLA or delivery-rate target for French PAs?

Do not assume a universal statutory support SLA, delivery percentage, incident deadline, availability promise, RTO or RPO. Apply the values in your signed contract and the company’s risk appetite, while checking current official sources for legal obligations. Keep response, diagnosis, workaround and verified resolution distinct. Legal, tax and contractual interpretations should be confirmed with qualified advisers.

How do we tell whether the fault is in the PA, ERP or configuration?

Follow the same document across each boundary. Require a correlated upstream outbound event and downstream inbound event, then locate the first unsupported hand-off. Correct PA receipt followed by no connector intake points downstream; connector intake followed by no ERP outcome points farther into the buyer stack. Compare transmission-time directory evidence and configuration history. Keep the result unproven if the evidence does not support ownership.

When can poor performance justify preparing to switch PA?

Switch preparation becomes reasonable when a material structural capability gap remains after fair diagnosis, contractual escalation and a defined retest, or when residual risk exceeds the buyer’s tolerance. One bounded configuration incident rarely proves that conclusion. Compare remediation feasibility, route criticality, recurring workload, evidence quality, total cost and transition risk. Preparation is not authorisation to switch; use current legal and migration guidance before acting.

What should we export and preserve before escalation or a possible change?

Preserve invoices and stable identifiers, event and lifecycle histories, routing and reasoned-address configuration, directory evidence, validation and conversion records, e-reporting outcomes where applicable, queue and quarantine data, reconciliation results, tickets, root-cause reports, contract evidence, role and security configuration, change logs and data dictionaries. Test that exports are readable and complete. Apply current legal, tax, accounting, security and contractual retention advice rather than an invented universal period.

Key regulations, formats and terms

FranceFrench tax administrationDGFiPimpots.gouv.frapproved platformplateforme agrééePDPFactur-XUBLCIISIRENVATe-reportingSMEmicro-enterpriseaccounting softwareEuropean CommissioneInvoicingEN 16931Directive 2014/55/EUstructured electronic invoiceVAT automationcross-border tradeFrench approved-platform post-launch performance review

France — Country hub

France

Open country hub →

Use the country hub to see every related deadline, checklist, format guide and software comparison for this market.

Continue reading

Official sources

We prioritize official government and EU sources where available and keep last-checked dates visible for mandate-sensitive pages.