How NEVI counts uptime — and why your RMS doesn’t agree

Every CPO with a NEVI award has noticed it at some point. The vendor RMS dashboard says the site ran at 98.7% uptime last month. The compliance analyst opens the data, runs the NEVI methodology against it, and the number comes back at 91.4%. Same site. Same month. Same chargers. Seven points of gap.

The first instinct is that someone is wrong. The honest answer is that both numbers are correct — they’re measuring different things. Understanding which one matters for the federal report is the difference between meeting the 97% per-port uptime floor and writing a corrective action plan.

Most charger vendors define “uptime” in their RMS as one of two things: the percentage of time the unit reported a heartbeat to the CSMS (the heartbeat metric), or the percentage of time the unit was not in a known “Faulted” state per OCPP StatusNotification messages (the fault-free metric). Both are useful operational metrics. Both are also self-reported. The charger has to know it’s broken to report it as broken.

The vendor RMS measures the charger’s opinion of itself. NEVI measures whether a driver could pull up and complete a session. These are not the same number.

The failure modes vendor RMS systematically misses include: stuck cable states that block new sessions but leave the unit reporting “Available”; payment-rail timeouts that prevent authorization despite the unit being technically online; OCPP message-level errors that leave the unit in a “ready but unable” state; network-side connectivity drops that prevent CSMS authorization round-trips; and firmware issues that cause the unit to accept a plug and then refuse to deliver energy.

A driver pulling up to a “100% uptime per RMS” charger and being unable to complete a session does not care which of those failure modes was responsible. NEVI does not care either.

NEVI’s Title 23 final rule defines per-port uptime in operational terms: the percentage of time the port was available to deliver a successful charging session. The methodology explicitly excludes time the port was unable to deliver a session for any reason, including reasons the vendor RMS does not flag. That includes time the cable was stuck, time the payment terminal was unreachable, time the CSMS network connection was unstable, and time the unit was technically “online” but refusing to start sessions.

The gap exists because the two measurements have different information sources. Vendor RMS sees what the charger reports about itself. NEVI requires what the port’s behavior toward a driver actually was. The two only converge in the rare case where the charger correctly self-reports every failure mode the moment it happens — and that case effectively does not exist in production.

CPOs trying to manually reconcile the gap typically pull vendor RMS exports, payment-rail logs from the payment processor, ticket data from the field-service team, and CSMS session ledger exports. They merge these by timestamp and try to identify periods where the charger thought it was available but no session could complete. This is a 2-3 week monthly exercise per site, and it produces approximations — not sealed evidence.

Real NEVI-grade uptime measurement requires the telemetry sources to be correlated continuously, not reconciled monthly. Every OCPP message has to be paired with the payment-rail event it triggered (or didn’t). Every StatusNotification has to be checked against the CSMS session ledger. Every minute the port spent in a non-Available state has to be classified: can a driver successfully complete a session here, right now, or not? The output is a per-port “available for session” line, drawn live, against a sealed event stream.

When NEVI quarterly reporting opens, the packet drafts itself: per-port uptime to methodology, excluded minutes with documented causes, raw event stream attached for any auditor who asks. The compliance analyst reviews instead of assembles. And when a vendor’s RMS shows uptime that does not agree with the NEVI measurement, the CPO has the specific evidence of which minutes diverged — and a vendor-management conversation with leverage.