ChargeX customer KPIs from OCPP logs
The four interim key performance indicators of the ChargeX Consortium (Session Success, Charge Start Success, Charge End Success and Charge Start Time): what each asks, which OCPP 2.0.1 messages the open calculator reads for it, and the assumptions it makes.
Where the KPIs come from
The National Charging Experience Consortium (ChargeX), run by Idaho, Argonne and the National Renewable Energy Laboratory with industry members, looked at public charging from the driver's side. Its June 2024 report (Quinn and others) splits the experience into six parts: finding a charger, accessing it, starting a charge, completing it, getting help, and feeling safe and comfortable. It defines two sets of indicators: ideal ones, which need data the industry does not collect yet, and interim ones, which can be computed today from the OCPP messages chargers already send.
The Implementation Guide (Savargaonkar and others, December 2024) explains how the interim KPIs are computed from OCPP. ChargeX then published a calculator for OCPP 2.0.1 logs under the MIT licence (version 1.0, May 2025).
The four interim KPIs
| KPI | The question, in words | Unit | How the calculator counts it |
|---|---|---|---|
| Charge Start Success | Can the driver pay, plug in and get energy flowing into the battery? | Percent | Attempts that reach power delivery, divided by attempts to start |
| Charge End Success | Once energy flows, does the session end because the battery is full or someone ended it on purpose, rather than prematurely? | Percent | Sessions with power delivery that end for a normal reason, divided by sessions with power delivery |
| Session Success | Both together: from the attempt to a normal end | Percent | Attempts that end for a normal reason, divided by attempts to start |
| Charge Start Time | How long from starting a session until charging begins? | Seconds | Time from the authorization answer (or from plug in) to the first charging state; reported as the 10th, 25th, 50th and 75th percentile |
What the calculator reads from OCPP 2.0.1
| Message and field | Read as |
|---|---|
| AuthorizeResponse, idTokenInfo.status | An authorization attempt before the transaction (plug in after authorization) |
| RequestStartTransactionResponse, status and transactionId | A remote start attempt, for example from an app |
| TransactionEventRequest, eventType Started with triggerReason CablePluggedIn | An attempt that began with plugging in |
| TransactionEventRequest, eventType Started with triggerReason Authorized, and its response idTokenInfo.status | An attempt authorized in the transaction event itself, counted as accepted or rejected from the answer |
| TransactionEventRequest, eventType Updated with chargingState Charging and triggerReason ChargingStateChanged | Power delivery started; its time ends the Charge Start Time |
| TransactionEventRequest, eventType Ended with transactionInfo.stoppedReason | A normal end when the reason is EnergyLimitReached, SOCLimitReached, Local, Remote, StoppedByEV, LocalOutOfCredit, TimeLimitReached or EVDisconnected |
| StatusNotificationRequest, connectorStatus | Parsed into the table, not used by the KPI equations in version 1.0 |
Each attempt is sorted into one of four start paths, and each path has its own equation in the Implementation Guide: plug in first (equations 1 and 12), authorization before plug in (3 and 14), remote start (4 and 15), and authorization in the transaction event (5 and 16). The first number of each pair feeds Charge Start Success, the second Session Success. Charge End Success is equation 10 and Charge Start Time equation 9. The result is a workbook with one sheet per KPI, each path shown separately and, for the two success rates built from several paths, a weighted average.
In version 1.0 of the calculator the list of normal end reasons spells one value LocalOutofCredit, while the OCPP 2.0.1 schema spells it LocalOutOfCredit. A session that ended for that reason may therefore not count as a normal end. This was observed in the code and has not been confirmed with the authors.
Assumptions
Real logs leave room for interpretation. The calculator README lists the choices it makes; in short:
- Responses are searched only after their request, and the first response with the same message ID is taken.
- An Authorize is bound to a transaction that starts with the same token within 300 seconds after it; failing that, within 300 seconds before it. An Authorize that binds to nothing, or a remote start without a transaction, is an orphan and still counts as an attempt.
- An attempt needs a valid start before it can count a power delivery, and a power delivery before its end can count for Charge End Success. This keeps rates from exceeding 100 percent or dividing by zero.
- Charge Start Time is recorded only when both the start and the power delivery time are known.
- The log reader knows two raw log formats. Any other log has to be turned into a table of device ID, OCPP message and timestamp first, with a full date in each timestamp.
- The days at the start and end of the chosen date range are cut off so that transactions are not counted half; duplicate rows are dropped.
Reading the numbers
- The KPIs count what the charger reports. A start that fails before the charger opens a transaction (for example, no plug in event because the configured TxStartPoint was not reached) does not appear at all.
- TxStartPoint and TxStopPoint decide when a transaction starts and ends, so two networks with different settings produce different numbers from the same driver experience. The example in the calculator README begins by reading both settings with GetVariables.
- A low Charge Start Success says that starts fail, not why. The why is on the cable: the capture of a failed start shows the stage where it stopped (Troubleshooting by symptom, Reliability of public charging: what field data says).
In this tool
The analyzer computes the four KPIs from OCPP 2.0.1 and 2.1 logs with a port of the INL calculator (MIT), checked against the calculator on the same log. The OCPP tab shows them for the session in front of you, with the transactions they were counted from; batch mode adds up every OCPP log of a run, with or without a capture, and reads the device_ID,message,timestamp table of the calculator with one charger per device.
- Times keep their milliseconds. The calculator cuts them to whole seconds, and on a time without a trailing Z it drops the last digit of the seconds as well.
- An Authorize and a Started event in the same second are taken in the order of the log; the calculator leaves such an Authorize unbound, which counts it as a failed attempt.
- An Ended event without stoppedReason counts as a Local stop, which is what OCPP 2.0.1 means by leaving it out; both spellings LocalOutOfCredit and LocalOutofCredit count as a normal end.
- The whole log counts; the calculator can leave out the first and last day of a data set.
Sources
- Quinn, C. W., et al.: Customer-Focused Key Performance Indicators for Electric Vehicle Charging, Idaho National Laboratory, June 2024
- Savargaonkar, M., et al.: Implementation Guide of Customer-Focused Key Performance Indicators for Electric Vehicle Charging, Idaho National Laboratory and Argonne National Laboratory, December 2024
- ChargeX Consortium: OCPP 2.0.1 Interim KPI Calculator, version 1.0, MIT, Battelle Energy Alliance (Idaho National Laboratory); README assumptions and the code of calculator.py and transaction_parser.py
- Idaho National Laboratory: ChargeX Consortium guidelines help industry quickly measure, improve EV charging for drivers, news release on the Implementation Guide: the four interim KPIs in words
- Idaho National Laboratory: ChargeX Consortium helps industry measure, understand EV drivers' charging experience, news release on the June 2024 report: six components, interim and ideal KPIs
- Open Charge Alliance: OCPP 2.0.1 JSON schemas, MeterValuesRequest and TransactionEventRequest, SignedMeterValueType; as shipped in the Mobility House ocpp library