Why EV chargers underperform: A diagnostic guide

Companion piece: 23 ways CPOs fix underperforming sites: the remedy playbook

A charger can report 99% uptime and still turn away one driver in three.

That gap is the most useful thing to understand about charging operations right now. ChargerHelp's 2025 analysis of more than 100,000 sessions across 2,400 chargers found reported uptime running between 98.7% and 99.9%, while only 71% of charging attempts succeeded on the first try. Even more uncomfortable: 35% of those failures occurred on chargers that appeared operational. The data is American, and UK numbers will differ, but the mechanism travels. Your dashboard is green. Your driver is gone.

The same gap exists one level up. A site can look acceptable relative to your network average and still pull in half of what it should for its location.

Both gaps cost the same thing: margin. Profitability sits on every board slide in this industry now, and the cheapest route to it runs through chargers you already paid for. Getting there starts with knowing what is wrong with them, because the wrong diagnosis spends real money and moves nothing.

In our business, we speak with heads of operations and network performance of charge point operators across Europe, and with advisors who have worked on thousands of chargers. This guide sets out what those teams look at when a site slips, the five families of cause, the symptoms that separate them, and what each wrong diagnosis costs. The companion piece covers the 23 remedies that follow.

1. Define underperformance properly

A site underperforms when it pulls less than it should for its location. The average is what hides your problem. A mean built from 400 healthy sites and 40 dying ones reads as acceptable right up until the year-end margin lands. Zapmap and the Green Finance Institute found time- and energy-based utilisation rates moving within a narrow band across every power category through 2025, even as the network grew by 13% and delivered energy grew by an estimated 21%. Aggregate stability tells you nothing about the spread underneath it.

Operators who run this well score each site against an expectation built from its own context: local EV density, traffic, dwell time, competing chargers within the catchment. Then they rank by the gap. Some teams keep a worst-against-achievable list so the operations meeting opens with names rather than a dashboard.

That gap is your starting point. The rest of this guide explains what fills it.

2. The five families of cause

Demand

The hardware works. The bay is clear. There are not enough EVs within reach, or the drivers who are there have never noticed the site.

Symptom signature. Healthy success rate, low session count, flat trend with no step change. New sites show a slow ramp rather than a drop.

How to confirm. EV registration density in the surrounding postcodes against the assumption in your site case. First-time-user share at the site: if it sits near zero while repeat use holds, awareness is your problem rather than driver behaviour.

What operators tell us. Kerbside teams keep returning to housing stock. A street where most residents have no driveway behaves nothing like a street where most do, and the second one will underperform for years whatever you spend on marketing.

Cost of calling it wrong. You cut the tariff at a site with no drivers in range. You lose margin on the few sessions you had and change nothing else.

Physical site & visibility

The most common cause that never appears in a fault log is that nothing has failed.

Symptom signature. Sessions well below comparable sites nearby with similar traffic. Long unexplained availability gaps, which usually mean a blocked bay rather than a fault.

How to confirm. Photographs beat data here. Teams that drive their estates turn up payment screens bleached by sunlight that have been losing customers for months. Nobody reports it because a driver who cannot read a screen does not file a ticket, nor do they return.

Cost of calling it wrong. You run a hyperlocal campaign that successfully drives people to a bay a diesel van has occupied since Tuesday.

Technical / uptime & faults

Symptom signature. Success rate below your network norm. Sessions started well above sessions completed. Faults recurring on the same units.

How to confirm. Cut success rate by charger type, by firmware version and by site age. The comparison usually names the cause on its own. A pattern that follows hardware points at firmware or the vendor. A pattern that follows geography points at grid or comms.

Age matters more than most operators budget for. The ChargerHelp analysis found first-time success falling from 85% at new stations to below 70% by year three. Your 2023 cohort is quietly becoming your problem cohort.

Cost of calling it wrong. This is the expensive one. You market a broken site, and every new driver you attract has a failed first experience. J.D. Power's 2025 US study found 14% of EV owners visited a charger and left without charging. You paid to create those visits.

Price & competitor activity

Symptom signature. A step change, not a curve. Utilisation drops on a date. Sessions shift to a competitor while your success rate stays perfect.

How to confirm. Find the date the drop started, then find what changed in the catchment on that date. Benchmark your own tariff while you are looking: Zapmap put the weighted average pay-as-you-go price in July 2026 at 54p/kWh up to 49kW and 80p/kWh at 50kW and above. If a competitor launched a time-boxed promotion, your problem has an end date and may need no response at all. Several operators describe watching a rival run a time-boxed deal, holding their own tariff, and waiting for volume to return.

Cost of calling it wrong. Two ways to lose. You match a promotion that would have ended anyway and start a race down. Or you sit still through a permanent repricing and lose a year waiting for a recovery that never arrives. The difference between those two mistakes is knowing whether the competitor's move has an end date, so establish that before you touch your own tariff.

Location: wrong from the first day

Symptom signature. No event explains the weakness, because there was never a strong period. The site launched flat and stayed flat.

How to confirm. Compare actual traffic and dwell time against the assumption in the original site case. This is uncomfortable work, since it audits a decision somebody in the business already made and defended.

Cost of calling it wrong. You spend three years cycling remedies through a site that no remedy reaches. Operators who run this well set a limit: after a defined set of interventions, a chronic loser gets switched off, sold, or moved, and the freed capex goes to something that can pay it back. Operators running large mixed estates describe exactly this call. Zapmap has started to see it in its own data, noting that poorly performing chargers are being taken out of public use.

3. The moves your own data can't see

Some causes hide inside your estate. Others sit entirely outside it, and those are the ones your dashboard is built to miss.

Your reporting shows your own sessions, uptime, and tariff. It shows nothing about the catchment around a site. A competitor opening two miles away. A rival cutting its pay-as-you-go rate last Tuesday. Either can pull a healthy site's volume down within days, and neither leaves a mark in the system you use to watch for trouble. The site reads fine on every internal measure. Success rate perfect, hardware green. The drivers just went somewhere cheaper or closer.

This is why a step change matters. When utilisation drops on a given date while your success rate remains steady, the cause is almost never within the unit. Something shifted in the market that day. Zapmap recorded on-street charger numbers growing 18% year on year through H1 2026, the fastest period yet, so the chance that a rival appeared in your catchment keeps climbing. Benchmark your own price while you look. Sit at or below the market rate and still lose volume, and price is not the whole story; a nearby deployment probably is.

You cannot respond to a move you never saw. Most operators learn a rival cut prices weeks later, from a sales rep or a quarterly review, long after the volume left. Watching the catchment, not just the estate, is what lets you make that call while it still matters, instead of writing it up afterwards.

4. The revenue you have already paid for

A charger's cost is fixed. The hardware, the installation and the grid connection behind it all get paid whether anybody charges or not. That single fact decides where the money leaks.

It means an empty socket is never neutral. Every session that walks to a competitor you did not spot, and every quiet hour that passes with the bay idle, is a contribution you have already paid for and did not collect. A busy site can still lose money, and a half-empty one bleeds it slowly enough that nobody flags it until the year-end margin lands.

The first leak is price. When a rival's move is real and lasting rather than a short promotion, holding your tariff out of principle just funds their growth. When it is a time-boxed deal, matching it starts a race down that nobody wins. The difference is whether the move has an end date, which is why you establish that before you touch your own price.

The second leak is time. Most sites do not underperform across the whole day. They fill at peak and sit dead in the trough, and that trough is capacity you are paying for and giving away. Time-of-day pricing pulls price-sensitive drivers into the quiet window. A fleet or depot deal contracts the dead hours to a local operator who values a guaranteed slot more than a low headline rate. Neither touches your peak margin, and both turn idle capacity into sessions. The test is simple: did off-peak volume grow without eroding what you make when the site is busy?

None of this shows up as a fault, because nothing failed. It shows up as a site that quietly earns less than the one down the road, and closing that gap is the cheapest margin you will find.

5. Where the manual version stops working

An hour per site is affordable. A thousand sites a week is not.

You cannot rank a thousand chargers by hand every Monday, hold a likely cause for each weak one in your head, and still recall six weeks later whether the tariff change at site 412 moved the number or whether the season did. So the loop degrades into a rescue: whoever shouts loudest gets attention, the same three sites get fixed twice, and the check-back disappears.

Operators with large estates describe some version of this.

The cost is not abstract. A weak site you never diagnosed keeps paying its grid connection, its maintenance contract and its capital charge while returning a fraction of the sessions it should. Multiply that across the tail of any estate, and you have the gap between the profitability number on the board slide and the one in the accounts. Diagnosis is the cheapest lever in the business, and almost nobody runs it systematically.

What we built

Dodona Network Optimisation ranks every site in your estate by the gap between what it pulls and what it should pull, then lets you quickly analyse the likely symptoms: demand, price, competitive activity, a fault, or placement.

We are opening a beta to a select group of operators in September, and several networks are already in.

Next: 23 ways CPOs fix underperforming sites: the remedy playbook

Sources

• ChargerHelp, first-time charge success rate analysis, 2025 (100,000+ sessions, 2,400 chargers)

• J.D. Power, US Electric Vehicle Experience Public Charging Study, 2025

• Zapmap and Green Finance Institute, UK charge point utilisation report, 2026

• Zapmap Price Index, July 2026

• Zapmap UK charging infrastructure statistics, H1 2026

• The Public Charge Point Regulations 2023, regulation 7