A business operating across five European countries in 2026 can easily find itself contracting with a Peppol Access Point in Belgium, a state-accredited provider in Greece, a registered platform in France, a direct API connection to a clearance system in Poland, and nothing at all in Germany — five relationships, five sets of credentials, five rejection semantics, and five archives, for one invoice process. This is not the result of poor procurement. It is the predictable consequence of a legal design choice made in 2014, and no single vendor can undo it.
Multi-country e-invoicing therefore has to be architected around the assumption that you will run several providers for at least the next decade. The question worth asking is not how to get to one vendor. It is which parts of the stack you keep under your own control so that the number of vendors stops mattering.
Why sprawl is structural
Directive 2014/55/EU harmonised two of the four layers of interoperability: semantics, through the EN 16931 data model, and syntax, through a short list of permitted formats. It did not address transmission or governance. The guidelines CEN published on transmission alongside the standard were explicitly non-binding.
The Commission’s own 2024 evaluation of the directive is direct about the consequence. Transmission, it found, emerged as one of the main challenges during implementation. Small businesses are pushed into the closed networks of their large customers, which do not interoperate with one another. Multinationals reported significantly high set-up costs across Member States precisely because they need different providers to cover country-specific requirements. And the report issues a warning that has since become the defining risk in this area: an increasing number of Member States are planning national certification systems for e-invoicing services and solutions, which will create new barriers to the internal market. It names Greece and France as places where country-specific requirements on providers — VAT registration, stringent certification — already generate high investment costs.
In other words, the fragmentation is being produced by the same governments that are mandating the invoices.
Five gatekeeper models you will have to satisfy
Vendor selection becomes tractable once you stop thinking country by country and start classifying by who the state requires you to go through.
| Model | Who you must contract with | Examples |
|---|---|---|
| Network with certified nodes | A Peppol Access Point, certified under a national Peppol Authority | Belgium (exclusive B2B channel since January 2026), Nordics, Greek B2G |
| Central clearance, direct access | Nobody — you connect to the state API, with intermediaries optional | Poland (KSeF), Italy (SdI), Romania (e-Factura) |
| Central clearance, accredited provider | A provider from the tax authority’s published list | Greece (myDATA), where accreditation is granted per channel |
| Decentralised, registered platforms | A privately operated platform registered by the tax administration | France (Plateformes Agréées, formerly PDPs) |
| No platform | Nobody — a format obligation only, channel is your business | Germany |
Two consequences follow immediately. First, a single country can sit in more than one row. Second, the models impose fundamentally different exit costs, which is the part that rarely gets priced.
Greece: one country, two networks, two accreditations
Greece is the clearest illustration of why country counts understate the problem. A supplier to a Greek public body clears the invoice through myDATA, the tax authority’s platform, and then delivers it over the Peppol network through a certified Access Point to the National Interoperability Centre, which acts as Greece’s Peppol Authority and routes the document to the buying entity. For domestic B2B, the same supplier transmits through myDATA using a provider accredited by AADE, or the state’s own free application.
That is a clearance platform, a network, a national routing hub and two separate accreditation regimes — inside one jurisdiction, for one company. Accreditation is also granted per transaction environment, so a provider approved for B2B is not automatically approved for B2G or B2C. Asking a vendor “are you certified in Greece?” is not a sufficient question.
France: registration is a licence with an expiry date
France restructured its model in October 2024, abandoning the public portal as an exchange platform and retaining it as a directory and concentrator. Exchange now runs entirely through privately operated platforms registered by the tax administration — originally Plateformes de Dématérialisation Partenaires, renamed Plateformes Agréées in 2025. More than 130 are registered and the list continues to grow.
Three features of that regime deserve attention from anyone selecting a provider.
Registration is time-limited. It is granted for three years and is renewable, awarded only after successful interoperability testing with the public portal, with ongoing compliance audit reporting. A registration is a licence that can lapse, not a permanent property of the vendor.
Your software vendor may not be a platform. France draws an explicit line between a registered platform and a merely compatible solution. A compatible solution cannot transmit invoices to your customers’ platforms, receive invoices directly, or report transaction and payment data to the administration. Many well-known invoicing and ERP products sit on the compatible side and connect through a registered partner. If your vendor’s answer to “are you registered?” is a description of a partnership, you have two suppliers, not one.
The directory holds the routing. The public portal’s central directory records every taxable person and the platform designated to receive their invoices. Changing platform means changing a state-held directory entry, with a cutover window during which counterparties may still route to the old one.
Where sprawl actually costs money
The obvious cost is licence fees, and it is the least important. Four others do the real damage.
- Status semantics do not transfer. Every platform defines its own outcomes, and the legal effect of a rejection differs fundamentally between systems. Where a clearance platform rejects an invoice, some jurisdictions treat the document as never having been issued, while others treat the rejection as a failure of the report rather than of the invoice. A status model built for one country will silently mishandle another. This is the failure that produces unbooked revenue rather than an error log.
- Directory entries are held by the provider. In both a Peppol setting and the French model, your reachability is a record pointing at a specific service provider. Migration is a re-registration exercise with a live routing cutover, not a contract renewal.
- Archives fragment along vendor lines. If each provider holds the invoices it processed, your retention obligation is spread across five systems with five export formats and five termination clauses — and the retention period outlives most vendor contracts by several years.
- Format drift. Each provider maps from whatever your ERP emits to whatever the destination requires. Five providers means five mapping layers maintained by five teams, and no single place where you can answer what was actually sent.
What ViDA fixes, and what it doesn’t
The convergence is real but slower and narrower than the marketing suggests. Council Directive (EU) 2025/516 makes EN 16931 the standard for electronic invoices from 1 July 2030 and removes the recipient-acceptance requirement for compliant invoices. Domestic real-time reporting systems must align with the EU model by 1 January 2035.
Three things it does not do.
It does not mandate a single standard domestically: Member States may continue to allow other standards for supplies within their territory outside the cross-border reporting scope. It does not mandate a transmission network — nothing in the directive requires Peppol or anything else. And the common electronic message for transmitting reported data is to be determined through an implementing act under Article 263(4), which has not yet been published, so the final transmission layer cannot be built by anyone today.
Separately, the Commission has signalled its intention to convert Directive 2014/55/EU into a Regulation, which is the instrument most likely to constrain national certification schemes. That is a proposal to follow, not a plan to rely on.
The architecture that survives: one canonical invoice, thin adapters
Since the number of providers is not under your control, the design goal is to make it irrelevant. That means drawing the line in a specific place.
- Own the canonical document. Generate one internal representation of the invoice, containing a superset of the fields any destination requires, and version it. Providers receive the canonical payload and adapt outward. If a provider’s own data model is your source of truth, you have outsourced your VAT position.
- Own the archive. Pull your own copy at the point of clearance or transmission and store it under your control. Retention obligations sit on the taxable person, run for up to a decade, and are not discharged by a vendor’s storage promise.
- Normalise status at your boundary. Map every provider’s outcomes into one internal state model that distinguishes transmitted, accepted by the platform, rejected by the platform, delivered to the counterparty, and rejected by the counterparty — and record the raw provider response alongside it.
- Treat certification as a licence, not a capability. Ask when the registration or accreditation expires, which channels it covers, what audit obligations attach, and what happens to your traffic if it lapses.
- Price the exit before signing. Who holds the directory entry, what the migration window looks like, how the archive is exported and in what format, and how long the provider will serve historical data after termination.
The alternative that periodically tempts large organisations is to build it internally. The historical evidence is unkind. Reviewing this pattern, the e-invoicing analyst Bruno Koch observed that in-house projects attempting to clarify legal requirements across dozens of countries “typically never succeed”, with companies eventually switching to third-party solutions. The parts worth building yourself are the canonical model, the archive and the status layer — not the country connectors.
Frequently asked questions
Can one provider cover the whole EU?
Several claim to, and some genuinely cover most of it, but coverage is not uniform and the reason is economic rather than technical. National certification schemes impose real investment costs per country, which is exactly why providers concentrate where their customers are. Verify coverage per jurisdiction and per channel, and verify it against the national authority’s published list rather than the vendor’s website.
Does Peppol solve vendor sprawl?
Partly. Where a country routes over Peppol, a single Access Point relationship can reach counterparties across all participating countries, and Peppol is in use across most of the EU. But several of the largest markets run central clearance platforms that are not Peppol, and a country can require both — as Greece does for public-sector invoices. Peppol reduces the count; it does not reduce it to one.
Is my ERP vendor a certified platform?
Often not. In France in particular, many familiar invoicing and accounting products are classed as compatible solutions rather than registered platforms, meaning they must route through a registered partner. Ask which registered entity your traffic actually passes through, and confirm it on the tax administration’s list.
What happens if a provider loses its certification?
Your obligation does not move. Registrations are time-limited and conditional — France’s runs three years and requires ongoing compliance audit reporting — so the contract should specify notification obligations, a migration commitment, and continued access to historical documents. This is a question to settle at signature, not at renewal.
Should we wait for ViDA before consolidating?
No, but consolidate the right layer. The canonical document model, the archive and the status handling are worth building now and will not be invalidated by anything in the 2030 or 2035 changes. The country connectors are what ViDA may eventually simplify, and those are the parts worth keeping thin and replaceable.