How CarsbySpec turns specifications into useful research
Accuracy begins with being explicit about what the database contains, what a number means and what we do not know. This page explains how profiles, reviews and comparisons are built.
One stable page for each model year
A CarsbySpec car URL identifies a brand, model and year. Several trims can sit behind that one address, because splitting every engine, battery and equipment combination into near-duplicate pages would make the catalogue harder to use and easier to misread. The trim table preserves those differences while the canonical URL remains stable.
Existing addresses are treated as permanent catalogue references. New editorial text, corrected figures and additional trims are added to those pages instead of moving them to a new URL.
Specification pages and reviews have different jobs
A car profile is the permanent model-year reference: dimensions, performance, range, capacity, prices and version differences. Its text explains the figures without assigning a buying score. The separate review URL turns those figures into category ratings and compares the car with the same body style and powertrain from nearby model years.
A CarsbySpec review does not claim that we have driven the vehicle unless a named road test and first-hand notes are explicitly published. We do not infer steering feel, ride quality, seat comfort, braking endurance or build quality from a specification table. Those questions belong in a physical inspection and road test.
How review scores and safety are handled
Performance, range or economy, practicality and value are scored only against a relevant peer group. Each category uses the available comparable fields, and a missing category is excluded rather than replaced by an average. Value combines price with usable specification outcomes, but local taxes, incentives, equipment and warranties still require a market-specific check.
The peer group contains one version per model with the same body style and powertrain within two model years of the subject. The nearest year is selected, then the lowest recorded price. This groups catalogue entries; it does not establish that they have identical equipment or suit the same buyer.
For each eligible metric, the score is 3.8 + 5.8 × the share of comparable entries with an equal or worse figure, capped at 9.8. At least two usable entries are required. Category scores average the available metric scores. Overall weights are performance 30%, energy 25%, practicality 20% and value 25%; weights are redistributed when a category is unavailable. A score can change when the peer data changes.
Performance uses power, torque, acceleration and top speed. Energy uses eligible range and consumption figures, charging power for electrified cars, and recorded emissions and tank capacity for combustion cars. Practicality uses cargo capacity, towing, seats and turning circle. Value uses price and specification-to-price ratios. These are specification scores, not measured comfort, reliability, running costs or resale-value assessments.
Range and consumption comparisons require matching verified test cycles and powertrains. Price comparisons require a matching recorded market. A field with an unknown basis is displayed for reference but excluded from those comparisons. An energy category may therefore reflect charging power alone; read the individual figures and source coverage alongside the score.
Every car receives a safety-data status. A sourced crash-test result is shown on the original authority, market and protocol scale; it is not converted into a CarsbySpec score or compared as if Euro NCAP, NHTSA, ANCAP and IIHS used the same test. A result may cover later model years only when the authority explicitly extends it to an unchanged vehicle.
Older cars are labelled by the test era they belong to. NHTSA consumer crash testing began in 1978 and its five-star system in 1993, IIHS crash ratings began in 1995, and Euro NCAP published its first results in 1997. A car that predates those programmes is marked as having no contemporary NCAP rating; that absence is never presented as a zero safety score. Vehicle weight, drivetrain and driver-assistance marketing are not substitutes for crash-test evidence.
Comparisons need context, not just winners
The comparison tool calculates exact differences for shared fields and keeps measurement direction clear: a shorter acceleration time or lower price is treated differently from greater power or cargo capacity. The prose also explains where a larger number is not automatically better, such as exterior length or vehicle weight.
Figures are most reliable when both cars use the same regional test method. WLTP, EPA, WLTC and CLTC range values are not interchangeable, and market prices can differ through taxes, standard equipment and model positioning.
Range leaderboards retain historical catalogue entries and sort their reported numbers. They explicitly include mixed and unverified cycles, so their order is not a controlled range test. The two-car tool withholds range, consumption and price winners when the relevant basis is missing or different.
Source coverage, research dates and corrections
Legacy records do not all have a linked primary source. The specification and review pages show that gap explicitly. A source citation identifies its scope: a handbook can support operating advice without verifying a price, and a launch release for one market cannot verify another market’s trim.
Model-specific buying research is published only for its stated model years and links to the source used. Its checked date refers to that research section, not every number on the page. Automatically calculated reviews do not receive an invented review date. General inspection checklists are labelled as such and are not presented as known faults.
Corrections are made on the existing model-year address so bookmarks and search links remain useful. A missing fault entry or crash-test result is not proof that a vehicle is fault-free or safe. Confirm current recalls against the exact VIN with the manufacturer or relevant authority.
Expanding the historical archive
The long-term catalogue is designed for cars from 1970 onward. Historical pages are published only when the model identity, year, variant and core specifications can be checked and paired with an appropriate image. Bulk-generating empty pages would increase the URL count but reduce the value of the site, so the public year range expands as complete records pass validation.
Every historical import must include source records and pass checks for impossible values, duplicate model-year URLs, duplicate trims and missing image assets before it can be applied.