How we verify EV data

EVz turns manufacturer brochures into structured, comparable data. This page explains exactly where that data comes from, how we treat real-world range, and what each verification label on a listing means. Our goal is that every figure on the site is traceable to a source and dated.

Where our data comes from

Specifications and prices come from official brand sources - brochures, spec sheets and official listings - cross-checked against reputable outlets for tested figures. Each model records its source and the date it was captured, so any figure can be traced back to where it came from and when.

Real-world range vs ARAI-claimed range

We record the manufacturer's ARAI-claimed range and, separately, a real-world range figure based on field conditions. We never present the claimed figure as real-world. Where both are known, our listings show the gap between them, because that gap is what actually affects a buyer.

What our verification labels mean

• Extracted - AI-parsed and not yet checked by a person. Extracted data is not published or indexed; it waits in a review queue.
• Sourced - gathered from an official brand listing or a reputable outlet, with the source link recorded, but not yet individually re-checked against the document by a person. Sourced models are published and clearly badged "Sourced", so their provenance and its level are always visible.
• Human-verified - an EVz reviewer has checked the data against the source document.
• Brand-confirmed - the manufacturer has confirmed the figures directly.

The EVz Score

The EVz Score is a 0-10 rating computed entirely from catalogue data - there is no opinion component, and this page is the published formula. It is a weighted blend of five components: real range (25%), value as real kilometres per lakh of ex-showroom price (25%), data confidence (20%), efficiency in km/kWh (15%) and charging time (15%). Each component is scaled against the market: the 10th-90th percentile of the models EVz publishes in that segment, recomputed on every data update. Value is scaled against price-band peers where the band has at least five scored models (smaller bands fall back to segment-wide anchors), so a premium car competes with other premium cars on value, not with a budget hatchback. Data confidence has four fixed tiers: 10 for an independently road-tested model, 8 for human-verified or brand-confirmed data, 6 for complete brand-published specifications, 4 for partial ones. If a component is missing, the remaining weights are rescaled; a model missing its price or any range figure is shown as "not yet scored" rather than given a guess.

Every range-based component uses a real range basis, never the raw claim - because claims in India run consistently optimistic (see the real-world range vs ARAI page). The basis ladder: the model's independently tested figure where one exists; else the claim adjusted by the model family's own measured tested-vs-claim ratio; else by the brand's measured record within its vehicle class; else by the segment's median ratio (used only when the segment has at least three tested models); else by the fleet-wide median across everything EVz has tested. A tested figure above the claim never inflates other models (the ratio is capped at 1). This also means the score rewards verification: a brand whose claims measure honest lifts its whole lineup, and adding a road test raises data confidence to the maximum. Charging-time comparisons stay within a segment (two-wheeler figures are AC home-charging times, car figures are usually DC fast-charge times). E-cycles are not scored: their range claims have no certification basis to verify against, so a score would be false precision.

Data licence & citation

EVz's published datasets - starting with the real-world range vs ARAI gap data - are free to reuse under Creative Commons BY 4.0: use the numbers in your article, chart or research, with attribution to EVz and a link to the source page. The licence covers the data itself; EVz's editorial text and page design remain ours. Each licensed dataset offers a CSV download and a suggested citation on its page. Where we cannot yet point at the independent source behind a figure, the dataset's source column is left blank rather than filled with a substitute - absence over misattribution, always.

Freshness

Every model shows the date its data was sourced. When a brand updates a brochure or price, we re-check and re-date the model.

Confidence on fast-moving policy and third-party data

Some data is not a fixed vehicle spec but a moving, often opaque, third-party fact: state EV subsidies that lapse when funds run out, or battery-swapping networks whose station counts and tariffs are rarely published. For these we show a confidence label instead of a verification tier:

• High - corroborated by an official source or two independent reputable sources, current.
• Medium - from a reputable source but with a caveat (a figure not officially published, or a policy near its expiry). Verify before you rely on it.
• Low - the data is dated or single-sourced; we flag it and, where it would otherwise read as thin, keep the page out of search until we can refresh it.

Every such figure also carries an "as of" date and a source link. Because subsidies and swap tariffs change frequently, we say plainly when to confirm a number with the state transport department or the operator before acting on it, rather than presenting a stale figure as current.

What we publish

Only models that clear our data-quality bar are published: a sourced, human-verified or brand-confirmed specification (sourced models carry their source link), a price, and at least one differentiated data point such as a real-world range figure or a citable subsidy fact. Anything that does not clear the bar stays unpublished rather than shipping thin or unverified.

Found something wrong? See our corrections log.