How we calculate the ChinaCars Score and rankings

Methodology v1.0 · effective from [publication date] · data last verified: [DD.MM.YYYY]

The ChinaCars Score is a 0–100 rating that summarises, at a glance, how a Chinese car sold in Romania stands on price, range, performance, safety and warranty, practicality and charging. This page explains exactly how it is calculated, where the data comes from, what we do not measure, and how the platform makes money — so you can check every number yourself.

1. Principles

1. The score is absolute, not relative. Every value (e.g. 456 km of range) is scored against a fixed, published scale. Adding a new model to the catalogue does not change anyone else's score. A model's score changes only when its data or the methodology version changes.
2. We only compare what is comparable. We publish three separate rankings — battery-electric (BEV), plug-in hybrid (PHEV) and combustion (petrol and conventional hybrids). A petrol SUV is not "worse" for lacking fast charging.
3. Missing data is neither rewarded nor punished. A model enters a ranking only if it has real data on every axis that applies to it. Otherwise it appears in the catalogue as "Unranked — insufficient data", with no score.
4. Every value has a source and a date. On each model page you can see, for each specification, where it comes from and when it was verified.
5. Weights are public and adjustable. The default weights are our editorial choice, explained below. Sliders on the ranking page let you recompute the ranking with your own priorities.
6. Nobody can buy a position. See section 9.

2. Which version of a model we score

We score the entry version — the one whose price appears as "from" on the model page. Every value in the score (power, range, battery, acceleration, charging) comes from that same version. We never combine the entry price with the top version's power.

Where a higher version changes the character of the car (e.g. MG4 XPOWER), we treat it as a separate model or mention it explicitly in the text, but we do not blend it into the score.

3. The axes and what feeds them

Each axis is the arithmetic mean of its available component scores (0–100).

What is not in the score, and why: real-world measured consumption (ADAC, Green NCAP) — shown under "Facts", but not available for enough models; software quality and updates; resale value; long-term reliability; service network (dealer counts are shown, not scored). These are real limits of the score, not accidental omissions.

4. Scoring scales (v1.0)

Each value is converted linearly to 0–100 between two anchors. Below the "0" anchor the score is 0; above the "100" anchor it is 100. Anchors were fixed in September 2026 from the actual distribution of Chinese models sold in Romania (roughly the 5th and 95th percentiles) and stay unchanged until the next methodology version.

Discrete scales:

• Seats: 7+ → 100 · 5–6 → 70 · 4 → 50 · 2–3 → 30
• Euro NCAP: 5★ → 100 · 4★ → 75 · 3★ → 50 · 2★ → 25 · 0–1★ → 0. A model not tested by Euro NCAP gets no score on this component (the axis is computed from warranty alone) and is visibly flagged "Not tested by Euro NCAP". We assume neither that it is safe nor that it is not.
• "Unlimited km" warranty is shown as such; kilometres do not enter the score.
• Conditional warranty (e.g. "5 + 5 years" tied to in-network servicing): we score the unconditional part and display the full formula.

5. Weights and overall score

When an axis does not apply (charging for PHEV; range and charging for combustion), the remaining weights are scaled proportionally to sum to 100 %. Overall score = weighted sum of the axes, computed to one decimal and displayed rounded to an integer.

Why these weights. Value carries the most weight because price is every buyer's first filter. Range follows because it is the main anxiety with an electrified car. Performance, safety and practicality are equal — we do not think we can decide for you whether 0–100 matters more than the boot. Charging has the smallest weight because it depends heavily on the infrastructure you have access to, not only on the car.

How much do the weights matter? We tested the BEV ranking with five weight sets (default, equal, "family", "city/budget", "motorway"). The top three positions are identical in four of five; the largest move (5 places) appears when value gets 40 %. The ranking is robust at the podium and sensitive in the middle — exactly where your priorities should decide. The sliders on the ranking page let you do that.

6. When a model enters a ranking

A model is ranked if it meets all of the following:

1. it has a "from" price published by the Romanian importer;
2. it has at least one real value on every axis that applies to it;
3. at least 70 % of its applicable fields are filled.

Otherwise it is unranked. It still appears in the catalogue, in comparisons and in search, but without a score or position, flagged "Insufficient data" with the list of missing fields. Every model page shows data completeness (e.g. "11/12 fields").

Models with no public price (importer offers "request a quote" only) cannot be ranked — value is a quarter of the score.

Commercial vehicles (vans, pick-ups) are shown in the catalogue but do not enter the passenger-car rankings; their prices are usually quoted without VAT and Euro NCAP tests them under a different protocol. The same goes for quadricycles (L7e): they are type-approved under different rules from cars and do not have to pass the same crash tests, so they are not comparable.

7. Where the data comes from

Accepted sources, strongest first:

1. Price list and technical sheet of the official Romanian importer — for price, warranty, versions.
2. Manufacturer's European technical sheet — for technical specifications (WLTP, power, dimensions, charging).
3. Euro NCAP (euroncap.com) — exclusively for safety stars. We do not use C-NCAP, ANCAP or other programmes.
4. Manufacturer's warranty manual — when the importer does not publish terms.
5. Reputable motoring press — only to confirm a value from 1–4, or when an official value is not published online; flagged as such.

Rules:

• Price is the list price including VAT, excluding time-limited promotions, subsidies (Rabla) and stock discounts. If the importer publishes only a promotion, we flag it "promotion" and do not score it as list price.
• Range is combined WLTP from type approval. We do not use NEDC, CLTC or unsupported "up to" figures.
• Every field shows, on the model page: value, source (type and link), verification date.
• Prices are re-verified monthly; specifications at every line-up change and at least quarterly.
• When two official sources disagree, we show both and score the more conservative value.

8. Versions, history and corrections

• The methodology carries a version number (this is v1.0). Any change to a formula, scale or weight becomes a new version, with a public changelog and a before/after table for all models.
• On the first day of every month we save a dated snapshot of the rankings. "#3 in October 2026" stays verifiable after the data changes.
• Data corrections are listed in the public log, with date and reason.
• Found an error? Every model page has a "Report an error" button. We check and reply; the fix appears in the log.

9. Independence and conflicts of interest

• The platform is built by Websem, a Romanian digital-marketing agency. Websem has no contract with any car manufacturer or importer and has received no payments, test cars or other benefits from them. [To be confirmed/adjusted by Dan before publishing.]
• The platform may earn from quote and test-drive requests that visitors submit through the form and that we forward to authorised dealers in the chosen county, and from services for dealers described on the "For dealers" page.
• No dealer or importer can buy a ranking position, score, label or placement. There is no sponsored content in rankings or on model pages. If we ever accept advertising, it will be explicitly labelled and kept apart from rankings.
• Quote requests do not influence a model's score, order or visibility.
• Editorial labels such as "Recommended" have been removed. Themed collections ("Cheapest EVs", "Long range") state their exact filter and sort criteria in the header.
• The scoring code and dataset are public [GitHub / CSV link], so anyone can reproduce the ranking.

10. Limits you should know about

• The score measures the spec sheet, not the driving experience. We have not driven every model.
• WLTP range is the type-approval figure; real consumption, especially in winter, can be 30–45 % higher. Where independent tests exist (ADAC, Green NCAP), we show them under "Facts".
• The "from" price is the entry version's; standard equipment varies widely between brands and is not scored.
• A model "not tested by Euro NCAP" is not necessarily unsafe — but we have no way of knowing.
• The scales in section 4 are an editorial choice. We derived them from the market distribution, but someone else could choose differently; that is why we publish and version them.

11. The comparator

The comparator (the /compare and /compare/A-vs-B pages) adds no opinions: it applies arithmetic to the same data and the same axes as the score, with published thresholds. The thresholds and weights below are part of methodology v1.0; any change follows the rules in section 8.

What we compare. The entry version of each model (section 2), up to 4 models at once. The position shown on a card is the one inside the model's propulsion group, computed over all published models, not only those in the comparison.

Who leads on what. On each score axis a model "leads" if it is at least 5 points ahead of the next one; otherwise we write "tight". A model without a value on an axis does not take part in it: missing data is neither an advantage nor a disadvantage.

The duel in points. The scoreboard adds one point for every criterion of the specifications table that a single model wins clearly: it has the best value and every other model with data is more than 5 % behind it (that is, not "≈ equal"). Ties, identical rows, rows with data for only one model and informational rows give no point, and missing data is never counted as a loss. Points are not weighted and do not replace the ChinaCars Score: they only show how many concrete criteria each model wins.

Labels in the specifications table. Every numeric value is compared with the best in its row (highest for range, power, boot space and so on; lowest for price, 0–100 acceleration and charging time):

A missing value shows as "n/a" and is never counted as "behind". Informational rows (length, width, height, drive) are not judged. The "EU homologated" row appears only when homologation is recorded; its absence means "we do not know", not "no".

Ratios. We show three simple ratios as the second line of a cell: €/km ("from" price ÷ WLTP range), €/hp ("from" price ÷ power) and km/kWh (WLTP range ÷ battery capacity — a catalogue efficiency, not real consumption). They appear only when both figures exist for all compared models; we estimate nothing.

Strengths and trade-offs. For each model we list up to 3 strengths and 3 trade-offs. A strength (trade-off) is a score axis in which the model is at least 5 points above (below) the best rival, together with the figure that produced the gap, for example "range +40 km (+9 %)". The figure shown is always the component of the axis in which the model really leads (or really trails). Price is added to the trade-offs when it is more than 20 % higher than the cheapest model in the comparison. We use no adjectives.

"What matters to you?" profiles. Three presets, with fixed and published weights:

The rest of the weight (100 % minus the fixed sum) is spread over the other applicable axes, in proportion to the default weights in section 5. Axes that do not apply to a propulsion type (charging for combustion cars, for example) are dropped and the weights renormalised. The score for your profile is the weighted sum of the axes that have data, renormalised over the available weight; when part of the weight has no data we show how much of it was covered. It is computed in the comparator only, in your browser. ChinaCars Score v1.0 does not change, and the rankings do not depend on the profile.

Different categories. When the selection mixes propulsion types we show each model's position inside its own group, say explicitly that overall scores are not directly comparable, and limit the axis verdict to the axes common to all models. A mix of segments (for example a hatchback and a large SUV) is flagged, not blocked.

Incomplete data. Below the 70 % completeness threshold (section 6) the comparison shows a warning listing the missing fields: the model's strengths may be understated.

What we do not do. We do not generate text with language models, we do not write "we recommend", and we do not order models by score unless you pick a profile; the default order is the one in the page address.

12. Car finder (the fit with your needs)

The car finder (/alege) orders the models by the criteria you give: free text ("family, 7 seats, under €45,000, no charger at home") or the choices in the panel. The text is read by fixed rules (words and figures: budget, powertrain, body, seats, home charging, long trips, commuting, county, priorities), without language models; everything we understood is shown as labels you can remove or change. The same rules turn a description typed into the catalogue search ("electric SUV under €35,000") into filters, and the page says what it understood.

What we leave out. Models not sold new in Romania (withdrawn or never sold officially), commercial vehicles (vans, pick-ups) and quadricycles unless you pick them, models over budget, models without a list price when you gave a budget, and models that miss a requirement you wrote (body, powertrain, brand, minimum seats, an all-wheel-drive version, minimum range, power, boot, warranty or charging). A missing value does not meet a requirement: we assume nothing.

The score on your priorities. We start from the ChinaCars Score axes (section 3). Each priority you pick gets a fixed weight: 35 % for one, 25 % each for two, 20 % each for three. The rest is spread over the other axes in proportion to the weights in section 5; axes that do not apply to a powertrain are dropped and the weights renormalised. With no priorities, the weights are those of section 5. Only axes with data count. From the text, "family" adds practicality and safety, and "long trips" adds range and charging (only when electric or plug-in cars are in play), after the priorities you wrote explicitly; three at most.

Adjustments for your situation (points added to or taken from the score on your priorities):

The fit is the score on your priorities plus the adjustments, kept between 0 and 100 and rounded. Ties go to the higher score on your priorities, then to the lower price. On every model we show the calculation ("How we got this") and at most three reasons and three trade-offs, all figures from the catalogue: price, WLTP range, DC charging, Euro NCAP, warranty, boot, seats, power, acceleration, dealers in your county. A priority's figure is listed as a reason when its axis scores at least 65 out of 100, and as a weakness below 40.

A little over budget. Models with a list price at most 10 % over your budget are not in the list; we show up to two of them apart, only if their fit would have put them in the top three.

What it does not mean. The fit is not a recommendation and no substitute for a test drive. It is computed in your browser on the published data; ChinaCars Score v1.0 and the rankings do not change.

13. Alternatives

Every model page lists up to 4 alternatives: the closest models by a distance with published rules.

• Price. The relative difference in the "from" price, divided by 25 % (a model 25 % dearer or cheaper adds 1 point). Beyond 40 % it is not an alternative. A model with a price gets no alternatives without one.
• Powertrain. Same kind: 0. Otherwise the distance in the table, multiplied by 1.2:

• Other. A different seat class (5 versus 6 or more): +0.5. A different body: +0.8. An unranked model: +0.2. Commercial vehicles are compared only with commercial vehicles, quadricycles only with quadricycles, and models not sold new in Romania are never shown as alternatives.

We pick the closest, at most one from the same brand and one from each other brand. Under each alternative we show up to 3 differences, in figures and without adjectives: price (from €300), powertrain, WLTP range (from 20 km, same powertrain), seats, Euro NCAP, power (from 15 hp), boot (from 40 l), DC charging (from 20 kW), warranty.

Questions about the methodology: info@chinacars.global