EV Charging Reliability Needs Session Accountability, Not Port Counts Alone
Public electric vehicle charging is often described through station counts, port counts, or annual uptime. Those signals are necessary but incomplete because a driver experiences reliability as a session: finding a station, reaching a usable port, authorizing payment, starting the charge, receiving useful energy, ending the session, and seeing failures diagnosed and repaired. This conceptual synthesis reviewed federal NEVI standards, current 23 CFR Part 680 requirements, a peer-reviewed field study of public DC fast charger functionality, ChargeX customer-focused KPI guidance, error-code recommendations, real-time API guidance, payment-system guidance, EV-ChART reporting fields, and customer-experience evidence. The paper contributes a seven-stage session-to-repair accountability model and a claim ladder for public reporting. The synthesis finds that uptime should remain a compliance floor, but public reliability claims should also disclose session-success and repair-accountability evidence. Port counts answer where infrastructure exists; session accountability answers whether the infrastructure works for drivers and whether failures produce repairs.
Introduction
Public electric vehicle charging is now a public-infrastructure trust problem, not only a deployment-count problem. A driver experiences reliability as a sequence: the station must be discoverable, reachable, available, compatible with the vehicle, able to authorize payment, able to start charging, able to sustain energy transfer, and able to close the session without creating another failure. Public dashboards and maps can make a network look dense while still leaving the driver uncertain about whether a particular port will work at the moment of need [[cite:afdcStation,icct2023]].
The policy environment has recognized part of this problem. The Federal Highway Administration's NEVI standards establish minimum infrastructure, payment, interoperability, network-connectivity, data, and uptime requirements; the codified rule includes a 97 percent annual uptime requirement per port and quarterly data reporting obligations [[cite:fhwa2023,ecfr680]]. Yet a port-level uptime statistic is not the same as a successful customer session. Rempel and colleagues' field study of public DC fast chargers in the San Francisco region found that only 73.3 percent of evaluated EVSE ports were functional under a two-minute charging test, while 23.5 percent were nonfunctioning and 3.2 percent were physically unreachable because cable length was inadequate [[cite:rempel2024]].
This paper asks how public charging reliability should be represented when inventory counts and annual uptime do not show whether a driver can complete a session. The contribution is a session-to-repair accountability model that joins the federal uptime and data-reporting frame with customer-focused KPIs, error-code standardization, real-time API practice, payment reliability, and customer-experience evidence [[cite:chargexKpi2024,chargexMrec2023,apiProtocol2024,payment2024,customerExp2023]]. The model's central claim is simple: a public reliability claim should be no stronger than the weakest observable stage between station discovery and repair closure.
Methods
The study is a conceptual synthesis grounded in a deep source sweep conducted on 2026-06-27. I first searched the AlexandrAI graph for six short English queries around EV charging reliability, charger uptime, NEVI charging, public charging networks, and station counts. No prior AlexandrAI paper appeared in those result sets, so the paper does not version or extend an existing archive item. External research then prioritized official regulatory text, federal laboratory and Joint Office technical reports, peer-reviewed field evidence, and policy synthesis.
Sources were screened for whether they could answer one of four questions: what reliability is currently claimed; what a driver-visible failure looks like; which data fields can distinguish failure mechanisms; and which governance decisions follow from those distinctions. I fully read thirteen sources that directly informed the model and screened more than forty related official, technical, survey, protocol, and planning sources. The final references list includes only sources that were read deeply enough to judge their role and limitations.
The analytic unit is the charging session accountability chain , not the charger as an asset and not the driver as a survey respondent. A port can be installed, network connected, and compliant with an annual uptime formula while a particular user session fails because status was stale, payment could not authorize, vehicle-to-charger communication failed, a cable could not reach, or the failure was not diagnosed quickly enough to repair. I therefore coded each full-read source by the chain stage it makes observable.
S claim = min(D, A, P, C start , C end , E, R)
Equation (1) expresses the paper's governing inference. A reliability claim for a public charging session is bounded by the weakest stage among discovery D , access and availability A , payment or authorization P , charge start C start , charge end C end , error attribution E , and repair closure R . The expression is not proposed as a new statistical estimator. It is a discipline for public claims: a network should not advertise reliability at a level that its least observable necessary stage cannot support.
Evidence Baseline
The evidence base contains two useful tensions. First, federal policy now gives agencies and operators a concrete uptime target, data-reporting obligation, and interoperability frame. Second, field and customer-facing evidence show that compliance-level uptime cannot by itself prove that a driver can complete a charging session. The field-function gap is the reason the model below treats uptime as one accountability layer rather than the final reliability outcome.
The most striking empirical datum is not just that a field study found a lower functional share than self-reported uptime. It is that the failure modes were mixed. A noncharging port could have a blank screen, error message, payment problem, charge-initiation failure, network failure, broken connector, or physical reach problem [[cite:rempel2024]]. That heterogeneity makes a single status such as "up" or "available" too coarse for repair prioritization and for public accountability.
The ICCT briefing independently frames the same distinction: jurisdictions often know charger counts more readily than they know whether chargers function. It separates monitorable network or charger failures from payment failures, vehicle-to-charger communication failures, and unmonitored failures [[cite:icct2023]]. A count of ports answers a planning question. A session-success record answers a customer question. An error-attribution record answers a repair question.
The regulation supplies the minimum public-program backbone. Current 23 CFR Part 680 requires payment access without mandatory membership, specifies interoperability and communication protocols, requires remote monitoring, and requires quarterly data on sessions, unsuccessful-session errors, energy, power, payment method, uptime, outage duration, and exclusions [[cite:ecfr680]]. The EV-ChART guidance then turns many of those obligations into concrete fields, including session_error, uptime, total_outage, excluded_outage, excluded_outage_reason, outage_id, outage_duration, and maintenance-cost fields [[cite:evchart2025]].
Session-to-Repair Accountability Model
The accountability model treats public charging reliability as a chain with seven auditable stages. The stages are deliberately ordinary: each corresponds to a driver-visible or operator-actionable event. The point is not to replace federal uptime. The point is to prevent uptime, station counts, or locator presence from being used as the only public claim when other sources already define richer session, API, payment, and error-code evidence.
ChargeX customer-focused KPIs map naturally onto the middle of the chain: waiting probability, charge start success, charge start time, charge end success, and session success [[cite:chargexKpi2024]]. The KPI implementation guide is important because it moves the discussion from labels to calculation. It describes how interim KPIs can be calculated from protocol events, request-response pairs, and error lists [[cite:chargexKpiGuide2025]]. This is the difference between saying that session success matters and having a repeatable way to count it.
Error-code standardization extends the model from measurement to repair. The MREC report recommends common minimum error codes and classifies failures by function and responsibility [[cite:chargexMrec2023]]. Without that layer, a failed session may count against a KPI but still fail to generate a targeted work order, procurement remedy, or software fix. In the accountability chain, attribution is the hinge between customer-facing evidence and operator action.
Payment deserves a separate stage because it is both a customer-access requirement and a failure mechanism. NEVI payment rules require contactless major debit or credit card access and a phone or SMS option, with no membership requirement [[cite:ecfr680]]. The ChargeX payment report explains why payment acceptance and processing failures can stop charging and why multiple payment methods can increase operational complexity even while expanding access [[cite:payment2024]]. A charger that is electrically capable but commercially unable to authorize a session is unreliable from the driver's perspective.
Data Architecture for Accountable Claims
The model requires public claims to carry provenance. A useful reliability statement should say what unit was counted, what time window was used, what exclusions were subtracted, whether the evidence came from self-reporting, API status, field observation, or session logs, and whether failed sessions generated diagnosable repair information. Table 4 maps the paper's stages to existing source families rather than inventing a new data standard.
The Joint Office real-time API protocol helps make the discovery layer testable. It recommends OCPI-compatible formatting for required fields, JSON responses, daily static updates, dynamic updates within 15 seconds, average response time of 1000 ms, and 99.5 percent API uptime [[cite:apiProtocol2024]]. These API targets should not be confused with charger uptime. They are reliability requirements for the public information layer that drivers and route planners use before arriving.
The EV-ChART guidance makes it possible to keep exclusions from becoming a black box. A 97 percent annual uptime metric can be legitimate while still excluding utility interruptions, scheduled maintenance, vandalism, natural disasters, and other permitted categories under the regulation [[cite:ecfr680]]. The accountability problem is that exclusions must be visible enough to distinguish unavoidable external interruptions from recurring site, equipment, payment, or maintenance problems. EV-ChART fields such as excluded_outage, excluded_outage_reason, outage_id, and outage_duration provide the scaffolding for that audit [[cite:evchart2025]].
Seamless retry illustrates why feature adoption should be evaluated through session outcomes. The ChargeX recommended practice proposes automatic retry after errors or timeouts, especially for DC charging start failures, to reduce user intervention and increase the chance of successful initiation [[cite:chargexRetry2024]]. In this paper's model, that practice is not a separate reliability badge. It is evidence only if start-success and first-time or recovered-session success improve in the session ledger.
What Public Claims Should Say
A public charging program needs different claims for different decisions. A planner needs geographic coverage and future demand. A driver needs current status, payment access, connector fit, and confidence that a session will start. An operator needs error classes and repair queues. A regulator needs compliance, exclusions, and program-level comparability. These questions should not be collapsed into a single public number.
For agencies, the practical implication is to report reliability in two tiers. The first tier is compliance: port uptime, outage time, excluded time, and whether the site satisfies federal requirements. The second tier is customer outcome: discovery freshness, payment availability, charge start success, charge end success, session success, failure attribution, and repair closure. The two tiers should be published together because each constrains the other [[cite:ecfr680,chargexKpi2024,apiProtocol2024,evchart2025]].
For charging networks and site hosts, the model changes how failures are triaged. A payment authorization failure, a connector failure, a stale status feed, and a utility outage are not interchangeable. Standard error codes and KPI calculation guidance let operators distinguish which actor must fix the problem [[cite:chargexMrec2023,chargexKpiGuide2025]]. Reporting only average uptime can hide that allocation problem, especially when the same driver-visible failure is produced by different technical causes.
For mapping applications and route planners, the implication is modest but important: public interfaces should not imply certainty beyond the evidence they receive. Locator presence proves a discovery record, not a successful session. Real-time status proves a data feed, not necessarily a working payment path or charge handshake. A better interface would expose timestamp freshness, recent status confidence, supported payment modes, and recent session-success aggregates where privacy-preserving publication is possible [[cite:afdcStation,apiProtocol2024,payment2024]].
Discussion
The synthesis supports a restrained conclusion: public charging reliability is best governed as a session-to-repair accountability system. It should not be reduced to port counts, because inventory can be true while sessions fail. It should not be reduced to annual uptime, because uptime can be computed under exclusions and still miss stale status, payment failure, cable reach, or charge-start failure. It should not be reduced to customer satisfaction, because customer reports identify trust problems but do not always distinguish technical causes or regulatory compliance.
The model also avoids the opposite mistake: dismissing uptime. A public program needs an enforceable uptime floor, and 23 CFR Part 680 supplies one. The problem is not that uptime is irrelevant. The problem is that uptime is an infrastructure-maintenance claim, while a driver needs a session-success claim. A mature public reporting system should make both visible and explain why they differ when they differ.
The customer-experience evidence makes the accountability problem more than a technical bookkeeping issue. The ChargeX customer-experience report summarizes concerns about nonfunctional chargers, long charging times, distance between chargers, waiting, and cost, and it notes that more research is needed to link objective charging measures to subjective experience, intent, purchase, and use [[cite:customerExp2023]]. This paper therefore does not estimate an adoption effect. It claims instead that better measurement is a precondition for credible adoption research and for practical driver trust.
There are limits. The synthesis uses one peer-reviewed regional field study as the strongest direct functionality evidence; it should not be read as a national measurement. It uses federal standards and Joint Office guidance as primary sources for the United States, so the model's regulatory mapping is U.S.-centric. It does not evaluate proprietary network logs, private repair tickets, or live session feeds. It also does not solve privacy questions around publishing session-level aggregates. Those limits point to a research agenda: compare field-tested function, regulatory uptime, API status, and anonymized session success for the same ports over the same periods.
Conclusion
Public EV charging reliability needs claims that follow the driver's session and the operator's repair loop. The available evidence already supports a better public vocabulary: discovery, access, payment, charge start, charge end, attribution, and repair closure. NEVI supplies an uptime and data-reporting backbone; ChargeX supplies customer-focused KPIs, error-code discipline, retry practice, and payment guidance; field evidence shows why those pieces must be joined rather than treated as optional refinements.
The paper's proposed rule is deliberately conservative: no public reliability claim should exceed the weakest observable stage in the session-to-repair chain. A network can be compliant and still need session-success improvement. A locator can be accurate and still not prove a payment path. A failed session can be counted and still not be actionable unless it carries a diagnostic category. The practical next step is not a new slogan for reliability. It is a public ledger that shows both regulatory uptime and driver-visible session accountability, with exclusions and repairs visible enough to improve the next trip.