You pull up to a charger, tap the card that worked yesterday, and get rejected. The charger is online. The cable fits. Your account has money in it. Nothing is broken. You have simply arrived at a charger belonging to a company your provider has no contract with. Charging interoperability in Europe is not primarily a technical problem. It is a commercial one wearing a technical costume.
Two companies are involved, and only one of them is at the roadside
The single most useful thing a driver can learn about charging is that the pillar in front of you and the account you pay with usually belong to different businesses.
The CPO, or Charge Point Operator, owns and runs the physical hardware: installing it, maintaining it, keeping it powered and online. The eMSP, or eMobility Service Provider, is the company whose app or RFID card you carry; it holds you as a customer, authorises your session and bills you. Sometimes one company plays both roles on its own network, which is exactly why its own chargers feel effortless and everyone else's feel unreliable.
When you tap a card at a charger operated by somebody else, the CPO has to ask a question it cannot answer alone: do I know this customer, and will somebody pay me for this electricity? Answering that question across company boundaries is what the industry calls roaming.
The protocol that carries the question
The language used for that conversation is OCPI, the Open Charge Point Interface. It is a free, vendor-independent protocol maintained by the EVRoaming Foundation, and it exists specifically to connect eMSPs, who have the drivers, to CPOs, who have the chargers.
OCPI carries more than authorisation. It moves tariff data, live status and availability, reservations, remote start and stop, smart charging instructions and the billing records that settle the session afterwards. When an app shows you that a third-party charger is occupied, that information usually arrived over OCPI.
Connections come in two shapes. A peer-to-peer link is a direct agreement between one CPO and one eMSP. A hub (GIREVE and e-clearing.net are the established examples) sits in the middle so that each party integrates once instead of hundreds of times. Hubs are why a mid-size provider can plausibly claim access to a large share of a continent's chargers.
Versions matter more than they should for something invisible to drivers. OCPI 2.3.0 is the current production version and is built to comply with EU regulation; 2.1.1 is no longer supported; 3.0 exists as a draft. A CPO and an eMSP on distant versions can exchange less information than either would like, which is one concrete reason a roaming session sometimes starts but arrives with a price you were never shown.
The honest summary: a standard protocol does not create access. OCPI defines how two companies talk. It does not oblige them to talk. Every network boundary you hit is a missing contract, not a missing cable standard.
Why a session fails even when roaming exists
Assuming the agreement is in place, a handful of failure modes remain, and they look identical from the driver's seat:
The authorisation round trip. Your card ID travels from the charger to the CPO's back office, out to the hub, on to your eMSP and back. Every hop can time out. A charger that seems to "think" and then refuse is often waiting on a system three companies away.
Stale token data. Your card is valid; the CPO's cached copy of the token list is not. This is the classic case of a card that works at one operator and fails at another on the same day.
Version and field mismatches. If tariff data does not survive the exchange, some operators block the session rather than start one they cannot price.
Silent commercial expiry. Roaming agreements lapse and get renegotiated. Coverage claimed in an app is a snapshot of contracts, and contracts change without any announcement to drivers.
None of these produce a useful error message. The screen says the card was rejected, so drivers reasonably conclude the charger is broken, and report it as broken.
What EU law now requires
The regulatory answer to all of this is to stop making membership the only way to buy electricity. The Alternative Fuels Infrastructure Regulation, Regulation (EU) 2023/1804, came into force on 13 April 2024 and attacks the problem from the payment side.
Under Article 5(1), newly built public charging points must accept ad hoc payment: no subscription, no app, no membership. Where the point delivers 50 kW or more, that means a payment card reader or a contactless device that can at least read payment cards; a QR code alone is not sufficient at that power level. Below 50 kW the rules are looser: card readers, contactless, or an internet-connected device enabling a secure payment transaction, which is where QR codes remain acceptable.
Existing infrastructure is not exempt forever. Charging points of 50 kW and above on the TEN-T network and at secure parking areas must be retrofitted by 1 January 2027. Separate obligations cover digital connectivity from 14 October 2024, price transparency under Article 5(4), and data provision under Article 20(2). Operators must publish static data such as location, connector types, opening hours and maximum power output, and dynamic data including operational status, availability, ad hoc price and the origin of the electricity.
Read together, that is a regulator saying two things: you must be able to pay like you would at any shop, and you must be able to find out the price before you plug in.
What a driver should actually do
- Carry a second way to pay. Not a second charging card from a similar provider, a genuinely different route, ideally a contactless bank card. Most roaming failures are account-shaped, and ad hoc payment routes around them entirely.
- Check the operator, not just the plug. Before you rely on a stop, note which company runs it. "CCS available" tells you the cable fits; it tells you nothing about whether your account is welcome.
- Expect newer high-power sites to be the easy ones. The AFIR card-reader obligation bites hardest at 50 kW and above, so recently built fast chargers are the most likely to take a plain bank card.
- Treat advertised roaming coverage as approximate. It reflects contracts, and contracts move. It is a good reason to plan a route and a poor reason to arrive with 4% remaining.
- Report the failure precisely. "Card rejected, session never started" and "charger dead, screen black" are entirely different faults. The first is commercial, the second physical. Conflating them is why so much charger data is untrustworthy.
Why this shapes what we show
Charging data has a credibility problem that fuel data does not, and roaming is a large part of why. A charger can be reported as failed by five drivers in one week and be in perfect working order. Five people simply held the wrong card. Databases that treat every failed attempt as a hardware fault slowly fill with chargers that are marked dead and are not.
That is why the useful unit of information here is a driver saying what happened, and when. A report from ninety minutes ago that a session started and delivered power is worth more than any static database entry, because it captures the one thing no specification can promise: that this charger, this operator, this payment route worked for a real car today.
The regulation is moving in the right direction and will make a great deal of this obsolete by 2027. Until then, the gap between what should work and what does work is filled by drivers telling each other the truth.