How Our Expert Ratings Are Calculated

Last updated July 2026.

At CarExplore we publish a single score out of 10 for every new car we review, built from five sub-scores. This page sets out exactly how each one is calculated, so you can check our working.

We think that matters. Plenty of sites publish a number without saying where it came from. Every figure on a CarExplore scorecard traces back to a source you can open yourself, and if we cannot derive a score from a stated input, we delete that row rather than guessing at it.

What the CarExplore Score is

An aggregation, not a road test.

We do not drive the cars (unless stated in the review). CarExplore does not conduct first-hand vehicle testing. What we do is read every Australian review of a car, combine their verdicts into one score, whether an outlet published a number, a written grade or a recommendation in prose, and set that alongside government safety, emissions and recall data. Every subjective statement in a CarExplore review is attributed to the outlet that made it or to the manufacturer.

That is the whole point of the format. One page tells you what four or five professional testers concluded, where they agreed, where they split, and what the official Australian data says, without you opening seven tabs.

The five sub-scores

1. Media Expert Rating (40% of the total)

This is the combined verdict of every major Australian outlet that has reviewed the car, expressed as one score out of 10.

It used to be a straight average of the scores outlets published as numbers. That no longer works, because outlets have moved away from numbers. CarExpert, one of the largest, dropped its scores in 2025 after readers pointed out that almost every car it reviewed landed between 7.8 and 8.4, which told buyers very little. It now rates cars in words, from Excellent to Below Average. If we only counted the sites that still print a number, this rating would speak for three or four outlets and call it the industry.

So we count every credible Australian review, whatever form its verdict takes, and we do it with a documented rule rather than a guess. This is the same approach as our Reliability and Eco scores: a real published input, run through a stated formula, to produce a number you can reproduce.

One scale for everyone. We place every outlet on a single 0 to 10 scale. On that scale a genuinely average car sits around 7.5, which is how car scores have always behaved in practice, rather than down at 5. A 9 is exceptional. A 9.5 is the car its rivals are measured against. Below 6 means real flaws.

We weight by how solid the evidence is. A published number is stronger evidence than a grade we converted, which is stronger than a written verdict we read. So the rating is a weighted average, not a flat one.

Here is how that applies to the outlets we read:

How we convert CarExpert's words.

This is our reading of CarExpert's published grade, not a number CarExpert assigned.

Why Carsales is not simply halved. Carsales defines 10 out of 20 as class average and 16 as best in class. Dividing by two would turn a perfectly average car into 5.0 out of 10, when an average car on the 10-point scale sits closer to 7.5. So we map it on the line those two anchor points set: 10 out of 20 becomes 7.5, 13 becomes 8.5, 16 becomes 9.5. Converting it linearly would drag every aggregate down and make it wrong, so we do not do it.

When an outlet only writes a verdict, with no number and no grade, we place it on the scale from the reviewer's own recommendation, and we hold a quote from the review that justifies where we put it. We label it on the page as our read of their words, and it carries the least weight of the three.

The formula. We multiply each outlet's score by its weight, add those up, and divide by the total weight. That gives the Media Expert Rating, rounded to one decimal.

When we hold it back. We publish this rating only when at least three outlets have reviewed the car and there is enough solid evidence behind the number to stand on. On the rare car where all we have is a couple of written impressions, we will not manufacture a score. We tell you what each reviewer said and leave it there.

We show our working. In every review you can see how each outlet was counted. A site that scored the car shows its score. A site that graded it shows the grade and the number we mapped it to. A site we read from its written verdict says so. We also note which variant each outlet drove, because a score awarded to a flagship is not a score for the base model.

2. Safety Rating (15%)

Taken from the ANCAP result and nothing else. We verify it at ancap.com.au.

Score = the mean of the four published stage percentages, divided by 10.

ANCAP changed its protocols on 1 January 2026, so there are now two eras of rating on the road:

• Ratings datestamped 2025 and earlier report Adult Occupant Protection, Child Occupant Protection, Vulnerable Road User Protection and Safety Assist.
• Ratings datestamped 2026 and later report the four Stages of Safety: Safe Driving, Crash Avoidance, Crash Protection and Post Crash.

Both are scored out of 100 and published as percentages, so the arithmetic is the same. But the two are not comparable, and we never average or compare sub-scores across them. A five-star result from 2021 and a five-star result from 2026 are different achievements, so we always print the protocol year beside the rating.

ANCAP ratings expire six years after their datestamp. We state the expiry.

If a car is unrated, or the rating excludes the variant we are reviewing, we delete the Safety row, say so plainly on the page, and note in the scorecard footnote that the overall score was calculated without it.

3. Eco Rating (15%)

Taken from the Australian Government's Vehicle Emissions Star Rating at vesr.gov.au, which rates every light vehicle sold here since 2004 in half-star increments up to six stars, based on tailpipe CO2 in grams per kilometre drawn from the Green Vehicle Guide.

Score = (VESR stars ÷ 6) × 10, rounded to one decimal.

Only zero-tailpipe vehicles reach six stars, so a petrol or diesel car cannot score 10 here. That is correct and intended.

If a car is too new to appear on VESR, we say so and leave the row out rather than substituting an estimate.

4. Reliability Rating (15%)

We only score this where there are hard inputs. Brand reputation is not an input.

Starting from a baseline of 6.5, which reflects the five-year, unlimited-kilometre warranty that is now the mainstream Australian standard:

• Add 1.0 where the warranty extends conditionally on dealer servicing, or where battery cover runs materially longer
• Add 1.0 where the standard warranty is seven years or more with no conditions
• Subtract 1.0 for each open recall against this generation of the car, checked at vehiclerecalls.gov.au

We filter recalls by build date. A recall covering the previous generation is not a recall against the car being reviewed, and treating it as one would be misleading.

Where a model is too new to have an ownership record, we write the Reliability section without a number rather than inventing one.

5. Cost of Ownership (15%)

Five-year total = capped-price servicing total + (annual fuel or energy cost × 5).

Annual fuel cost is the ADR combined consumption figure at 15,000km a year, priced at the fuel grade the car actually requires. If a car has moved to premium unleaded, we price it as premium, because that is what the owner pays. We use the NRMA weekly fuel report and state which average and which week.

The five-year total converts to a score on a fixed scale:

• $12,000 over five years scores 5.0
• $6,000 scores 10.0
• Linear in between, using 10 − ((total − 6000) ÷ 1200), clamped to the 0 to 10 range

We show the working in one line in every review.

We do not include insurance. Premiums vary so much by age, postcode, history and insurer that a single quoted figure would mislead more readers than it helped.

How the five combine

The CarExplore Score is a weighted blend, calculated from the rounded sub-scores shown on the page so that any reader can reproduce it with a calculator.

The Media Expert Rating carries the largest single weight because it represents the combined judgement of several professional testers who have actually driven the car. The four data-derived scores together carry 60%, so no single measure dominates.

Worked example, all five rows present

CarExplore Review

7.8 /10

Expert Scorecard

Media Expert Rating (blend of 4 expert verdicts)

7.6/10

Safety (ANCAP 2026 protocol stage scores)

8.4/10

Eco (Vehicle Emissions Star Rating)

7.5/10

Reliability (warranty and open recalls)

7.5/10

Cost of Ownership (5yr servicing and fuel)

8.3/10

The maths:

(7.6 × 0.40) + (8.4 × 0.15) + (7.5 × 0.15) + (7.5 × 0.15) + (8.3 × 0.15) = 3.04 + 1.26 + 1.125 + 1.125 + 1.245 = 7.795, which displays as 7.8

Worked example, one row deleted

When a row cannot be derived, we delete it and redistribute its weight proportionally across the rows that remain. We always name the missing row on the page.

The 2026 Toyota RAV4 is a live example. ANCAP has not rated the sixth-generation car, so there is no Safety row. The remaining weights total 85, and each is divided by 85 to rescale:

(7.9 × 0.4706) + (8.2 × 0.1765) + (7.5 × 0.1765) + (8.3 × 0.1765) = 7.95, which displays as 8.0

The overall score will never come out identical to the Media Expert Rating. If the arithmetic ever produces a tie, we show an extra decimal place so the two are clearly different numbers doing different jobs.

What we do not do

• We do not drive the cars (unless stated). No CarExplore review contains a first-hand driving impression. Anything about ride, steering, noise or seat comfort is attributed to the outlet that reported it.
• We do not include user reviews. Owner ratings are valuable but self-selecting, and mixing them with professional scores would make the average mean less, not more.
• We do not weight newer reviews more heavily, or rank one masthead above another. Recency and brand do not change how much a review counts. What does change it is the form of the verdict: a published number counts for more than a grade we converted or a written verdict we read, as set out in the Media Expert Rating section. We show each review's date and how we counted it so you can judge for yourself.
• We do not use overseas ratings as a substitute for Australian ones. Euro NCAP, NHTSA and IIHS results apply to differently specified cars. Where an Australian rating does not exist, we say the row is missing rather than filling it with a foreign one.
• We do not estimate insurance costs.

Where the data comes from

• Pricing, specifications and servicing: the manufacturer's Australian website
• Safety: ANCAP, at ancap.com.au
• Emissions: the Vehicle Emissions Star Rating at vesr.gov.au, and the Green Vehicle Guide
• Recalls: the federal register at vehiclerecalls.gov.au
• Fuel prices: the NRMA weekly fuel report
• Expert scores: CarExpert, Drive, Carsales, CarsGuide, Chasing Cars, CarSauce and WhichCar/Wheels

Where two sources disagree, the manufacturer's own Australian page wins for specifications and pricing, and the government register wins for safety and emissions. We note the disagreement rather than quietly picking one.

How often ratings change

Every scorecard carries a "Scores updated" line with the month and year.

Reviews sit on permanent URLs and we update them in place rather than publishing a new page each model year. A score is recalculated whenever pricing changes, an ANCAP result is published, a recall is issued against the current generation, or a new expert review appears.

FAQs