VW EU Data Act Dashboard

Turn the raw data export Volkswagen gives you under the EU Data Act into a readable, interactive, fully offline dashboard — driving, charging, battery health, thermal system, backend activity, and an audit of how complete the package actually is.

One Python script, no server, no accounts, no telemetry. The output is a single dashboard.html you open in any browser — including a plain-language battery assessment when the export contains battery-current and cell-voltage evidence, which makes this tool especially useful when buying a used VW EV (see below).

The portal serves the participating Volkswagen Group brands — Volkswagen Passenger Cars, Volkswagen Commercial Vehicles, Audi, Škoda, SEAT, Cupra and Bentley — and this tool reads their common export format. Best supported is the MEB EV family: VW ID.3/ID.4/ID.5/ID.7/ID. Buzz, and their platform siblings Cupra Born/Tavascan, Škoda Enyaq/Elroq, Audi Q4 e-tron (same platform, same 96-series battery packs, so the battery analysis applies). Model and pack sizes are detected from the VIN. Built and reverse-engineered against a VW ID.7 export; the high-frequency diagnostic channels were validated on that car, so treat other models' inferred channels with a little extra skepticism — and please open an issue with your results on other Group cars. Non-MEB vehicles (e.g. Audi's PPE models, combustion cars) will still get the documented fields and the package audit, but the battery inferences won't apply.

What you get

• Overview — current vehicle snapshot (SoC, range, closure and lock states, window opening, lights, charging policy, climate, connectivity, service due) and headline figures: distance, consumption, energy charged, idle drain, top speed
• Driving & charging — distance per day (reconciled to the odometer), driving heatmap by weekday/hour, speed distribution, energy consumption per day with ambient temperature context, a charging ledger (AC / DC fast / scheduled — classified from the power-curve shape, with estimated kWh and power), a trip ledger with an independent ∫I·V consumption cross-check and regen/traction split, and parked-drain events annotated with the thermal-mode mix (conditioning vs quiet parks)
• Battery — a plain-language battery health verdict (healthy / normal wear / worth checking) when battery-side evidence supports one — including a session-by-session table of how the usable capacity was measured (SoC window, current coverage, AC/DC) — state of charge, highest/lowest cell voltage, cell imbalance over time and by SoC band, HV current
• Thermal — ambient temperature, operating modes, seven thermal sensors as small multiples, coolant flow and valve-actuation timelines
• Backend & config — remote actions, vehicle reports, backend errors, and a decoded configuration snapshot
• Package audit — what VW's own Data Dictionary says exists vs. what your export actually contains (useful if you want to complain — see below)

Every value is labelled observed (directly in the export), derived (calculated from observed samples) or inferred (undocumented channel whose meaning was reverse-engineered). Solid line segments are measured; dashed segments bridge periods where the car wasn't reporting.

Getting your data from the Volkswagen Group portal

1. Go to the Volkswagen Group's EU Data Act portal: https://eu-data-act.drivesomethinggreater.com/de/en
2. Select your vehicle's brand (Volkswagen, Audi, Škoda, SEAT, Cupra, Bentley, VW Commercial Vehicles) and sign in with that brand's ID account. The vehicle must be linked to your account as owner/primary user.
3. Request the historical data package for your vehicle.
4. Wait for the notification that the package is ready (this can take a while), then download it. The download link expires after ~7 days, so don't sit on it.
5. ⚠️ Check the file size — first attempts often come back near-empty. The tool no longer refuses such packages: it builds a snapshot-only dashboard from whatever arrived, flags the delivery as incomplete (it is VW's export service failing, not your car), and tells you to re-request and complain. In practice the portal frequently delivers an incomplete package on the first request (a JSON of a few KB with only a handful of snapshot fields). A real package with diagnostic history is tens of MB with hundreds of thousands of records. If yours is tiny, simply request the export again — it can take two or more attempts until a complete file arrives. The script prints the record count when it runs, so you'll see immediately whether you got a real one.
6. Unzip the package. Inside you should find one or more .json files named after your VIN (e.g. WVWZZZ...123_20260720080125.json). That JSON is the only file this tool needs — everything else (including the field descriptions from VW's Data Dictionary) is built into the script.

Usage

Requires Python 3.9+ and nothing else — no packages to install, and the only input file you need is the export JSON (field descriptions from VW's Data Dictionary are bundled into the script). The bundled V4.0 index contains 4,429 field descriptions, including every numeric configuration field and its -0/-10 variants.

Put your export JSON next to the script, then:

python3 build_dashboard.py

That's it — open the generated dashboard.html in a browser. No configuration is needed: the script detects your model from the VIN, your timezone from the vehicle's own clock, and measures usable capacity when battery-current and cell-voltage coverage permits it. Structured charging-only exports receive a labelled energy/SoC proxy instead of a fabricated SoH verdict. Everything it decided is printed as it runs. To be explicit instead:

python3 build_dashboard.py export.json -o dashboard.html

Reports default to English. Generate the complete report in German, Dutch or Lithuanian with --language de, --language nl or --language lt. Official Data Dictionary descriptions, raw field names and raw payload values remain verbatim as source evidence.

Options

Flag	What it does
exports...	one or more export JSONs — multiple files are merged and deduplicated
-o / --out	output HTML path (default dashboard.html next to the export)
--language {en,de,nl,lt}	report language (default en)
--csv	also write cleaned per-series CSV files (SoC, odometer, charges, reporting activity, time-/SoC-based power curves, capacity evidence, monthly energy, trips…)
--price-kwh 0.21	electricity price — adds cost estimates to the charging ledger
--currency €	currency symbol for --price-kwh
--utc-offset 2	override the auto-detected display timezone (hours from UTC)
--pack-kwh 86	override the measured/assumed usable battery capacity
--vehicle-title "VW ID.4"	override the VIN-detected vehicle name
--include-identifiers	keep full VIN and backend/user IDs in the HTML (redacted by default)

Build an archive over time

VW retains only about two months of the high-frequency diagnostic history. The script therefore accepts any number of exports and merges them:

python3 build_dashboard.py export_may.json export_july.json export_september.json

Request a fresh export every ~6 weeks, keep the JSONs, and your dashboard grows a history deeper than any single package VW will ever give you — seasonal consumption, battery-health trends, the lot.

Buying a used MEB EV? (ID.3/4/5/7, Buzz, Born, Enyaq, Q4 e-tron…)

This might be the tool's best use. The battery is the single most expensive component of a used EV, and dealers rarely show you a real state-of-health figure. But the seller can get the data for free: under the EU Data Act, any owner can request their vehicle's data export from the Group's portal (see "Getting your data" above). So:

1. Ask the seller to request the historical data package for the car and send you the JSON — it costs them nothing but a few clicks and a few days' wait. Ask early: the export takes days to arrive and covers roughly the last two months of driving.
2. Run this script on the file. Diagnostic-channel exports can provide:
   • Measured usable battery capacity and state of health — measured from actual charging sessions, matched against the pack sizes that model shipped with (a pack measuring 80 kWh can only be a degraded 86 kWh pack, never a healthy 77 — the tool reasons accordingly)
   • Cell imbalance — an early-warning indicator a range test won't show
   • Odometer cross-check — the diagnostic odometer channel vs. the reported mileage field, a basic tampering sanity check
   • Usage patterns — DC fast-charge share, charges past 80%, deep discharges, idle drain: how the battery was actually treated
3. A seller who refuses to share a free, privacy-redacted export about the car they're selling is also telling you something.

The export contains no location data (verified on the export this tool was built against), and the dashboard redacts the VIN and account identifiers by default. It still contains detailed vehicle history, so it should be shared deliberately.

Structured charging-only exports still provide the observed charging ledger and curves, but correctly withhold an SoH verdict.

Battery health & capacity evidence

The dashboard gives a direct verdict — healthy / normal wear / worth checking — built from two measurable indicators:

• Measured usable capacity. For every charging session that gained ≥30% SoC while the car was reporting, the script integrates battery current × pack voltage (mean cell voltage × 96 series cells, the MEB layout) and divides by the SoC gained. The median across sessions is your pack's measured usable capacity. The nominal pack for the state-of-health percentage is chosen from the sizes that model actually shipped with (from the VIN): ID.3 → 45/52/58/77/79 kWh, ID.4/ID.5 → 52/77/79, ID.7 and ID. Buzz → 77/86 — picking the smallest option the measurement can fit, since packs only lose capacity. All energy figures in the dashboard use the measured value automatically.
• Cell imbalance. The spread between the highest and lowest cell voltage. A healthy pack stays in the single-digit millivolts; a consistently large or growing spread is an early warning sign worth a service check.

When battery-current and cell-voltage history is present, merging several exports over time can turn repeated battery-side capacity measurements into a degradation trend. Charging-energy/SoC proxies are not suitable for this.

The ∫I·V integration happens at the battery terminals, so charging overhead and any climate or conditioning load while plugged in never enter the figure. Structured-format exports carry no battery-current or cell-voltage history. For those files the dashboard shows reported session energy divided by SoC gained as a descriptive consistency proxy only. The JSON does not document where that energy is metered, and AC/DC losses, auxiliary loads and 1% SoC rounding materially affect the ratio. It is not used as usable capacity or state of health. Ratios beyond the largest pack option are flagged as evidence of that uncertainty. Battery-terminal estimates remain diagnostic rather than official warranty certificates — treat "worth checking" as a prompt to investigate, not a diagnosis.

Two export formats

The portal delivers (at least) two different package formats, and the tool handles both:

• Diagnostic-channel format (seen on a VW ID.7): hundreds of thousands of records in undocumented numeric channels — odometer, SoC, cell voltages, speed, current — from which trips, battery health and everything else are reconstructed. Roughly two months of history.
• Structured format (seen on a Škoda Enyaq): no numeric channels at all; instead documented chargingSession.[n] records (start/end, SoC window, energy, average/peak power, AC/DC), time- and SoC-based powerCurve samples for recent sessions, daily/monthly charged-energy aggregates — and high-volume event records (speed, ignition) delivered without values, timestamps only. Around eight months of history. Here the charging ledger and power curves are the vehicle's own figures (labelled observed). The energy/SoC ratio is shown as a proxy but does not produce an SoH verdict. A reporting-activity heatmap is built from the value-less events; odometer/trip history simply isn't in the package and the dashboard says so.

Data quality notes

• The export contains sensor error values the script filters out: SoC readings of 0, odometer readings of 1048574, temperatures of 0 K.
• Over 90% of the records sit in numeric channels that VW's own Data Dictionary does not document. Their meanings here (odometer, SoC, cell voltages, speed, …) were inferred from units, value ranges and cross-checks against documented snapshot fields, and are labelled inferred in the UI.
• Energy figures (kWh, kWh/100km, charge power) are estimated from SoC deltas and the assumed pack capacity — good for trends, not billing.
• No location or route data: a systematic search of every channel in the reference export — including the opaque numeric channels and long encoded values — found no coordinates, addresses, route objects or serialized journey payloads. The only movement evidence is the odometer, speed and battery-current telemetry the dashboard already uses; kilometres inside odometer sampling gaps are reported with whatever sparse timing evidence exists, never reconstructed into trips.
• Timestamps in the export are UTC; display uses --utc-offset.

Is the package complete? (spoiler: probably not)

The Package audit tab compares your export against VW's own Data Dictionary. In the export this tool was built on, VW delivered 233 of 5,140 documented data point keys, and categories the dictionary explicitly defines — warning lights, diagnostic trouble codes — were missing entirely, while the same warnings were visible in the VW app. No location data was present.

Under the EU Data Act (Regulation 2023/2854, Arts. 4–5) and GDPR (Arts. 15/20) you are entitled to the data VW holds, in a usable format. The audit tab gives you the concrete field names and numbers to cite if you decide to push back through the portal's contact form.

Updating the bundled Data Dictionary

The dashboard never reads the PDF at runtime. If Volkswagen publishes a new dictionary, maintainers can regenerate the embedded Python and browser maps with the development-only importer (requires pypdf):

python3 -m pip install pypdf
python3 tools/import_data_dictionary.py dictionary.pdf \
  --web-dictionary ../ev-data-dashboard-web/dictionary.js

The importer separates the PDF's Description column from its measurement unit and technical datatype columns, repairs wrapped Data Point Names, validates the expected row count, and handles the small number of malformed table rows via a text-stream fallback.

Privacy

• Everything runs locally; the dashboard is a static file and makes no network requests.
• The VIN and backend/user identifiers are redacted by default in the HTML (--include-identifiers to keep them).
• dashboard.html still contains your driving history — share it deliberately.
• The included .gitignore keeps exports and generated dashboards out of version control. Never commit your export.

License

MIT — see LICENSE.