How the Symbioen Index works

Four principles

Independence. The index cannot be bought or influenced - and that is precisely where its value comes from. The data behind every score comes exclusively from AFIR-mandated National Access Points: government feeds that operators are legally required to publish. The independence is in where the value comes from.

Driver experience and market first. The index is not designed to make operators look good. It is designed to measure what drivers actually experience. Every metric - from the availability penalty to how Performance is defined - is chosen because it catches the gap between what an operator reports and what a driver encounters at the charger. A healthy market for EV infrastructure requires honest measurement.

Integrity. Every number is traceable to a raw event. The same methodology applies to every operator, at every station, at every level of aggregation - country, region, operator, station, EVSE. One methodology, applied uniformly, verified algorithmically.

Regulatory alignment. Definitions of Performance, availability, and reliability are still settling across the industry. When consensus forms or new regulatory standards take hold, Symbioen updates the methodology accordingly. The methodology is continuously refined to better reflect real-world conditions - and stakeholders are kept informed of any changes as they happen.

The problem with Uptime

The standard metric in EV charging is Uptime - the percentage of time a charger reports itself as available. Operators publish it. Tenders require it. Investors use it in due diligence. The problem is that it is self-reported, and the number almost always looks good. Networks consistently report 95-99% Uptime while independent field studies find that drivers successfully complete a charge on the first attempt only 71-73% of the time - a figure documented across multiple independent studies, including UC Berkeley's 2023 field assessment, the 2024 ChargeX report, and the 2026 Stanford/NREL analysis.

The gap is structural. A charger can report Available at the moment a driver arrives, fail to start a session, and that failure never touches the Uptime figure. The charger was, by the operator's definition, up. From the driver's perspective it was broken.

Where the data comes from

Symbioen data comes from National Access Points - government-mandated feeds that operators are legally required to publish under the EU's AFIR regulation.

What the index measures

The composite score combines metrics about each EVSE, plus a penalty for chargers that report Available but cannot charge.

Device Reliability

This measures the technical stability of the EVSE. A reliable EVSE stays in a consistent state. Erratic state transitions indicate a failing connection or a malfunctioning hardware or software or connectivity issues. The score penalises these erratic transitions. Hardware that is noisy in the data feed is noisy for drivers in real life.

Anomaly Avoidance

This measures the integrity of the data stream itself. Some chargers produce duplicate timestamps. Some cycle through impossible status sequences - physically unreachable transitions that indicate a software or connectivity problem rather than a real state change. A charger that floods the feed with updates every few seconds is throttled by the pipeline, and each such throttling event is recorded as a Broken Episode and penalises the score.

Availability

A charger that is technically stable in the logs but routinely offline or unreachable is not serving drivers. Availability measures the percentage of time a unit is functional and reachable - and directly penalises ghost stations that appear clean in the data while being unavailable in practice.

Utilization Rate

If drivers are successfully charging at a station, the station is working. Utilization Rate tracks the proportion of time a charger is actively delivering energy. It acts as real-world validation: a high use share means real people are arriving, connecting, and charging. Low use share on a theoretically available charger is a warning sign - it suggests a station that is logged as present but is not attracting or completing sessions. Drivers learn quickly which stations to avoid. The data eventually reflects what word of mouth already knows.

Performance

Symbioen computes a dedicated Performance score that builds on the methodology proposed by ChargeUp Europe. AFIR mentions Uptime only in a non-binding recital - there is no hard EU regulatory definition yet. ChargeUp Europe proposed a formula without exemptions: all time a connector spends in a faulty OCPP status counts as downtime, regardless of cause - vandalism, grid outage, or maintenance. Their model is straightforward: when the charger reports an error, failure time accumulates; when it reports Available, failure time stops.

Symbioen goes further. We treat a return to Available status as insufficient proof of recovery - the charger must demonstrate it actually works by completing a real charging session of at least two minutes. We call this approach “Until Next Success.” It catches a specific failure mode that pure status-based measurement misses: chargers that briefly flip to Available between errors without ever delivering energy. Neither a brief flip back to Available nor a session under two minutes resets the clock. Only a confirmed, successful delivery of energy proves the charger has recovered.

ChargeUp Europe acknowledged explicitly that their formula, if applied honestly, would mathematically lead to lower Uptime percentages - below the 97-99% figures operators routinely publish. Symbioen’s stricter approach lowers the number further still - and that is by design. The 2-minute threshold and the persistent penalty rule are Symbioen’s engineering choices, informed by related industry benchmarks (California’s 5-minute SCAR threshold, P3’s 4-minute short-session definition) but not derived from any published standard.

The analysis window is a rolling seven days. The figure is computed per EVSE and aggregated per station.

How the index is composed

The Symbioen Index brings together all metrics into a single score, covering the full range of failure modes a driver encounters: hardware instability, software and data issues, and outright unavailability.

The index is calculated at the level of the individual EVSE. The result is a single score per EVSE, ranging from 0 to 100.

A station score is the EVSE-weighted average of all EVSEs at that location. A 20-port station contributes proportionally more to its operator's average than a 2-port station. Simple arithmetic averages would let a small high-performing site mask a large underperforming one.

Operator, municipality, county, and country scores follow the same logic - each level aggregates the level below it, weighted by EVSE count. The result is a consistent set of numbers across every granularity: you can compare a single charger against its station, its station against its operator, and its operator against the national average, all on the same scale and calculated by the same method.

Cleaning the data before scoring

Raw status feeds from National Access Points are not clean. Chargers generate duplicate events, rapid oscillations between states, and brief apparent recoveries that are immediately followed by another failure. Scoring these events naively would reward chargers for a five-second Available signal between two hour-long outages.

Before Device Reliability is calculated, the timeline is passed through a process called Truth Smoothing. Once a charger enters an error state, it must produce a stable working signal for at least two minutes before the system accepts that it has recovered. If it returns to error within five minutes of an apparent recovery, the recovery is invalidated - the charger is considered to have been broken throughout. Brief positive signals between failures are treated as noise, not evidence.

The same logic applies to update storms. When a charger generates status changes at a rate that indicates a software or hardware problem rather than real state changes, the pipeline throttles the feed. Each throttling event is logged as a Broken Episode and applied as a penalty to the Anomaly Avoidance score. The silencing itself becomes a reliability signal.

Fair operator rankings

An operator running a single charging point that had a good week should not sit at the top of a national leaderboard above operators running hundreds of stations. Small samples produce unreliable averages - a one-station operator is a coin flip, not a data point.

Operator leaderboards use Bayesian averaging to correct for this. Each operator's score is pulled toward the global mean in proportion to how small their sample is. An operator with one station gets pulled significantly; an operator with two hundred stations is barely affected.

All group-level aggregations - across stations, municipalities, or operators - are also weighted by the number of EVSEs in each group rather than counted equally. A station with twenty chargers contributes proportionally more to an operator's average than a station with two.

When a score is not published

Not every EVSE receives a score. If the data for a given EVSE is too sparse - fewer than five significant status changes in the analysis window, or coverage of less than thirty percent of the window - no score is emitted. A charger that the feed barely notices cannot be reliably rated. Publishing a number in that case would be misleading, so Symbioen returns an ineligible result with the reason rather than an uncertain estimate.