Every provider in this market sells the same three things: a connection to one or more networks, a validation layer, and somewhere for the documents to live afterwards. The demonstration you get will be of the second, because it is the only one that looks like software. The two that decide what the relationship costs you are the first and the third.
This publication has no commercial relationship with any provider, takes no referral income, and names none here. That is worth stating, because most comparison material about access points is written by access points, and the axis a provider compares on is always an axis it wins.
What follows is the order in which the questions actually bind. Take them out of order and you will spend the shortlist stage negotiating price per document with three suppliers, one of which has nothing live in the country that obliges you first.
Establish your own position before you read a sales page
Almost every poor selection decision here is made by a buyer who does not yet know their own scope, and the provider is then allowed to define it.
- The legal entities you invoice from, and where each one is established
- Which mandates bind each entity now, and which will within three years
- Whether each entity must issue, receive, or both
- Annual document volume per entity and per country, credit notes included
- Any document types you send that are not invoices in the legal sense
- Which system holds the invoice when it becomes final, and what it can emit today without development
- Where your archive sits now, and who owns the retention obligation
- The renewal date of any contract you are already in
None of that requires a vendor. All of it changes what a vendor's answers mean.
Scope has two dates in it, not one
The first filter is geography, and not the geography you have today. A provider that covers the countries obliging you now and has nothing live in the one you enter in two years is a provider you will replace, and replacement is the expensive event.
Supported is doing a great deal of work in most coverage tables. It can mean a live connection carrying documents today, a published roadmap, or a subcontracted partner whose name you are not given. Ask which, country by country, in writing. A provider willing to put a date against a connection that does not yet exist is telling you something useful; one that will not is telling you something too.
Receiving is not the mirror image of issuing
Several mandates bring the obligation to receive in ahead of the obligation to issue, and the two are commercially different. Issuing is a project you control: which customers, in which order, at what pace. Receiving is not. Once your entity is registered in a directory on the Peppol network or on a national platform, anything correctly addressed to you arrives, from suppliers you have never spoken to, in whatever conforming form they chose.
A proposal that prices issuing carefully and defers receiving to a later phase has not priced the harder half.
Whether they run the connection or resell it
Ask who operates the endpoint your documents leave from. Some providers are certified participants in the network themselves; some are software companies sitting on somebody else's access point; some are both, depending on the country.
What your ERP can emit without a project
The largest line in most of these programmes is not the subscription. It is the work of getting a correct and complete document out of the system that holds it, and that work is yours whichever provider you pick — the trade-off between doing it inside the ERP and buying it is set out in the build-or-buy comparison.
Providers divide into two camps, and the division tells you more than any feature list. One camp takes your data as it is and maps it: flat file, database extract, whatever comes out. The other hands you a specification and requires the source data to satisfy it before onboarding.
The first is easier to buy and hides your master data problem inside a mapping layer you do not control. The second is harder to buy and is quoting the real cost. Neither is wrong, but only one has told you the truth about your own data, and the deferred cost does not evaporate. It returns as exceptions, which is where a cost model that includes rework earns its keep.
Custody of the archive
Ask what is stored, in what form, and whose obligation that storage discharges. What you need to produce years later is the structured document that was exchanged, not a rendering of it. A provider that archives a PDF has archived a picture of the evidence.
If your archive lives with the same provider that carries your traffic, you have combined two exit problems. Leaving means migrating live registrations and extracting years of retained documents in the same window, under a deadline set by the contract you are trying to end.
The day you leave
This is the clause almost nobody negotiates and almost everybody eventually needs. Your identifiers are registered on your behalf, in a network directory or on a national platform, and in most arrangements the provider made that registration. Leaving is therefore not a matter of stopping payment. Somebody has to move the registration, and until it moves, correctly addressed documents keep arriving wherever the directory still points.
Ask for the migration steps in writing, the expected window, what happens to documents in flight, and the price of a bulk export of everything held. Then ask whether that export is a contractual right or a service quoted at the moment you need it. The difference between those two answers is worth more than any discount on the table.
Whose data caused the failure
A service level covering platform availability is not the service level you need. Documents rarely fail because a platform was down. They fail because something in the document was wrong, and the argument that follows is about whose wrong it was.
Ask what happens to one that fails validation. There are three outcomes and only one is operable: rejected at the gate with a reason somebody can read and act on; queued in a portal nobody has been told to watch; or accepted, failing later at the buyer, surfacing weeks afterwards as a payment that never came. What happens after a failed hand-off is covered in failed deliveries and the retry chain.
The axes, side by side
| Axis | A good answer | A bad answer |
|---|---|---|
| Coverage | Named live connections, dates against anything not yet live | Broad claims of European coverage, no country detail |
| Receiving | Registration handled and evidenced, a route for suppliers you never onboarded | A later phase, priced separately or not at all |
| Connection | The operator of the endpoint is named, subcontractor included | Nobody will say who runs the endpoint |
| Your data | A written list of the fields required, and what they will not derive | Send whatever you have and it will be handled |
| Failures | Rejected at the gate, with a reason and a named place it appears | Visible only inside a portal, with no notification |
| Archive | The exchanged structured document, bulk export, format in the contract | A rendering of the invoice, exported on request |
| Exit | Migration steps, a stated window, export as a contractual right | Unaddressed, or priced at the moment you ask |
| Price | A structure you can map onto your own volumes, exceptions included | A headline rate with the exception work unpriced |
What actually decides it
Price per document is the variable everyone compares, because it is the only one that arrives as a number. It is also the most easily renegotiated: volumes change, contracts come up, and a provider that wants your traffic will move on price. Nothing else on that table moves.
Coverage decides whether you are choosing once or twice. Exit terms decide what the second choice costs. And the position a provider takes on your data — take it as it is, or fix it first — decides whether the programme you budgeted is the one you will run.
Shortlist on those three. Then negotiate the price.
