Methodology & validation
Trust in this product depends entirely on whether you can defend the numbers in writing. This page lays out exactly what RangeCheck models, what we cite, and — critically — what we have not yet independently validated. If you want to dig deeper, every formula in the source model is open for inspection on request.
1. What we do (and don't) measure
RangeCheck predicts battery state-of-health (SoH). We do not plug into your car. We do not pull data from manufacturer telematics. There is no app. The output is a model estimate of the most likely current usable kWh for a given vehicle, age and odometer.
For physical pack measurement, we are building a partner network of UK OBD test providers. When that ships, you will be able to upgrade any £9 Predicted Report to a £49 OBD-Verified Report.
2. The model
SoH% = 100 − calendar_loss(age, chemistry)
− cyclic_loss(odometer, chemistry, fast_charge_share)- Calendar ageing: ~2.3% per year for the first 24 months, then ~1.4% per year. LFP packs scaled ×0.7. NCA packs scaled ×1.1.
- Cyclic ageing: ~0.012% per equivalent full cycle, with a fast-charge stress multiplier of
1 + fast_share × 0.4. LFP cycle loss scaled ×0.55. - Equivalent cycles are derived from odometer assuming a fleet-typical 3.4 mi/kWh consumption rate.
- Fast-charge share defaults to 40% for retail vehicles and 70% for fleets.
- Chemistry is inferred from make and model. Where ambiguous (e.g. trim-dependent), we default to NMC.
- Total loss is capped at 35% — beyond that the model declines to predict and recommends physical inspection.
3. Datasets we calibrate against
We do not own these datasets. We cite them as the published evidence base our coefficients are tuned to:
- Geotab — Fleet EV Battery Degradation Tool (2023): 6,300 vehicles, multi-year telemetry. Public methodology and aggregate curves.
- Recurrent Auto — State of Used EV Batteries (2024): 15,000 vehicles, US-skewed but model-mix overlaps strongly with UK.
- Smith et al. — NMC and LFP cycle-aging models (peer-reviewed journal literature).
- UK Department for Transport — National Travel Survey 2024: 8,500 mi/yr UK average, used as the default annual-mileage assumption in the public Index.
4. Coverage
145 UK-market models are scored — every electric car, van and truck currently on UK sale plus the most common used-market models from the last seven years. New models are added as official UK kWh and chemistry data becomes available.
5. Accuracy — what we claim and what we don't
Target accuracy: within ~2–3 percentage points of measured SoH for cars in the 1–7 year, sub-100,000 mile range. This target is based on the published accuracy of the Geotab and Recurrent models we calibrate to (their reported MAE is in the 2–4pp range).
Independent validation status: in progress. We are sourcing a panel of measured-SoH OBD readings to publish a RangeCheck-specific MAE figure here. Until that figure is published on this page, please treat the headline accuracy as a target, not a measured result.
Refund policy: any £9 Verified Report is fully refunded if our predicted SoH is more than 5 percentage points away from a manufacturer or OBD-tool readout taken within 30 days of the report date.
6. Known limitations
- Vehicles with battery replacements are not detected — the model assumes the original pack.
- Heavy-duty and commercial use cases (taxi, last-mile delivery) age faster than the retail default; use the fleet preset.
- Cars stored long-term at 100% SoC or in extreme heat will degrade faster than the model predicts.
- First-generation Nissan Leaf (24 kWh, air-cooled) degrades non-linearly and is flagged separately.
- Pre-2018 Tesla Model S and X with NCA chemistry have higher variance than NMC peers.
7. What's modelled vs measured vs aspirational
| Claim | Status |
|---|---|
| State-of-health % output | Modelled — peer-reviewed coefficients |
| Geotab / Recurrent dataset sizes (6,300 / 15,000) | Cited — public, third-party figures |
| ±2.1pp accuracy | Target, derived from cited datasets — RangeCheck-specific validation pending |
| Vehicle catalogue (kWh, range, chemistry) | Sourced from manufacturer UK specs and EV Database |
| Listings sell ~22% faster with a report | Aspirational — to be measured once partner marketplaces are live |
| Buyers save ~£1,400 in negotiation | Aspirational — based on average UK used-EV price gap, not measured outcomes |
| OBD-Verified Reports (£49) | Coming soon — partner network in development |
8. Reproducibility
Researchers, journalists and procurement teams can request the full coefficient table, the vehicle catalogue and a CSV of every Index entry by emailing methodology@rangecore.app. We respond within two working days.