There are two ways to measure an EV charger. One asks whether the equipment was online and able to dispense electricity. The other asks whether a driver who pulled up actually left with a charge. Those two numbers are not close to each other, and almost every argument about charger reliability is really an argument about which one is being quoted.
Uptime measures the station, not the session
Publicly funded charging carries a numeric reliability obligation, and it is worth reading precisely, because the definition does most of the work.
Read the exclusions again. A utility outage, a vehicle-side fault, planned maintenance, vandalism, and hours the site says it is closed all come out of the denominator. Each exclusion is defensible on its own — an operator cannot be held to account for a substation — and the cumulative effect is a metric that can stay high through a period in which drivers repeatedly could not charge.
There are also failures the definition does not see at all. A dead card reader on a station that is otherwise online. A stall blocked by a parked combustion car. A screen that responds but never authorizes. None of those makes the port "down", and all of them send a driver away.
What the field data says
The gap between the two measurements has been tested directly, by people driving to stations and trying to charge.
That is the finding that reframed the whole conversation: not that operators were lying, but that they were reporting a different quantity than the one drivers experience.
The second half of that is the part a site owner should sit with. Reliability does not hold flat and then fall off a cliff. It degrades on a curve as hardware ages in the weather, which means a site commissioned with excellent numbers is not a site that stays that way without anyone touching it.
Failure one: the parts a driver touches
The most common causes of a dead charging port are physical, and they are physical in an unglamorous way.
Cables and connectors take the worst of it. A DC fast charging cable is heavy, often liquid-cooled, dropped on concrete several times a day, driven over, and sometimes cut for the copper. Latch mechanisms wear. Pins corrode where a connector has been left face-up in the rain. Screens fail from ultraviolet exposure, moisture ingress and touch layers wearing through.
None of that is a surprise to the manufacturers, which is why their own schedules treat these items as consumables rather than as failures.
A site running on the assumption that a charger is fixed infrastructure — that it is installed, commissioned and then only touched when something breaks — will find those components failing on roughly the schedule the manual predicted, but unplanned, on a Saturday, with a driver standing next to them.
Failure two: the connection to the back end
The second family of failures is the network link between the station and the management system that authorizes sessions, meters them and bills them.
Connectivity dominating that list is consistent with where chargers live: underground garages, the back of parking lots, rural interchanges. Cellular signal is marginal in exactly the places where installing a charger is cheap.
The protocol matters here too. Open Charge Point Protocol (OCPP) is the standard messaging layer between a charge point and its central system, and equipment that speaks it can, in principle, be moved between networks. Equipment that speaks a proprietary dialect is tied to one vendor's back end, which is a commercial fact until it becomes a reliability fact.
Failure three: payment
Drivers and telemetry disagree about payment, and both are right.
Telemetry sees payment as a small share of failures because most sessions on a given network are started in that network's app by drivers who already have an account. Drivers report it as a large share because the failure happens to the people least able to work around it — someone at an unfamiliar network, without that app, on a site with poor cell service, in front of a reader that does not respond.
Which means a payment failure is a site-design problem as much as a hardware one. A working card reader, legible instructions and enough signal to complete a transaction are what stop a functioning charger reading as a broken one.
Level 2 and DC fast chargers do not fail the same way
It is worth separating the two, because a site with both will see two different maintenance profiles and be tempted to manage them as one thing.
An AC Level 2 unit is comparatively simple: a contactor, a control board, a cable and an enclosure. Most of its failures are the connection to the network, the cable and connector, and the electrical supply behind it — a tripped breaker, a loose lug, a GFCI that keeps opening. They are cheap to fix and easy to miss, because a Level 2 charger that has quietly stopped reporting looks exactly like one nobody used that week.
A DC fast charger is a small industrial power plant. It has rectifiers, contactors, cooling, filtration, communication modules and a liquid-cooled cable on the high-power units, and each of those is a separate thing that can fail while the rest of the unit reports itself healthy. It also fails more expensively and takes longer to fix, because parts are model-specific and often on a lead time.
The practical consequence is that the two need different service arrangements. Level 2 wants monitoring that notices silence and a technician who can attend a few units in one visit. DC fast wants scheduled preventive work, a spares position for the parts with long lead times, and a diagnosis path that does not start with a truck roll.
Failure four: the back end goes away
The fourth failure mode is the one nobody plans for, and it takes out working hardware.
This is the strongest practical argument for buying equipment that speaks an open protocol and for keeping the network contract separable from the hardware contract. A charger whose management system can be replaced is a charger that survives its vendor. One that cannot is a stranded asset the day the vendor stops answering, and reconnecting it costs a truck roll per unit even when the firmware exists.
Why the same fault gets three different names
A site with mixed hardware and more than one network sees the same underlying problem described differently by each vendor, which makes it very difficult to tell whether a station has one recurring fault or four unrelated ones.
For an operator, adopting that vocabulary costs almost nothing and pays back immediately: it turns a pile of vendor-specific alerts into something countable, and something countable is something you can hold a maintenance contract to.
What this means if you own the site
Three things follow from all of the above.
First, measure what the driver experiences, not only what the station reports. Whether a session succeeded on the first attempt is the number that correlates with the complaints you receive, and it is the metric the national laboratories have been pushing toward for exactly that reason.
Second, treat the consumables as consumables. Filters, cables, coolant, connectors and screens have intervals, and planned replacement during business hours costs a fraction of an emergency call-out at the moment a port fails.
Third, keep the paths open — an open protocol, a network contract you can exit, and a service arrangement that includes parts and labor rather than diagnosis alone. The difference between a two-hour fault and a two-week one is usually whether somebody local is authorized to open the cabinet.
We are an intermediary: we connect you with vetted, licensed local technicians who carry out the diagnostics and the repair, and the technician confirms the scope and price before work starts. What we can tell you before that call is which of the four failure modes the symptoms point at, because a connectivity fault, a worn cable and a card reader are three different visits. Start at ev charger repair for the service itself, or at fleet roadside assistance program if the chargers are one line in a larger service account.