Data methodology

Vehicle data comes from manufacturer specification sheets, manufacturer websites and configurators for the South African market. Each recorded figure keeps a reference to the source it came from and the date it was confirmed.

Prices are stored as dated observations. When a price changes, the new figure is added and the old one is marked superseded — it is never overwritten, so the history remains available.

Equipment is recorded as one of six states, and two distinctions matter:

• Standard — included in the price.
• Optional — can be added on its own.
• Package only — available only inside a bundle, which is a different purchase.
• Conditional — depends on another choice, such as the gearbox.
• Not available — a fact about the car.
• Not confirmed — a gap in Camberly's data, not a limitation of the car.

The last two are never merged. Reporting an unconfirmed feature as unavailable would be publishing a false claim about a vehicle.

Two credible sources can quote different figures for the same car — most often because they measure differently, such as a fuel consumption figure on the newer WLTP cycle against the older NEDC one. Neither source is wrong.

Camberly never picks one quietly. Both figures and both sources appear on the vehicle page, the fact is marked as disputed, and Camberly lowers its own confidence in it — a figure two credible sources disagree about is not settled, however authoritative either one is.

Until an editor rules on it, the figure shown is chosen in this order, and each step must be decisive before the next is used:

1. Authority over this kind of fact. A manufacturer is the authority on its own trim equipment. A safety organisation, not the manufacturer, is the authority on a crash test result.
2. Confidence recorded against the observation.
3. Recency — the more recently confirmed figure.
4. A fixed tie-break if nothing above separates them. It is arbitrary, and deliberately stable: an arbitrary rule that always gives the same answer is honest, whereas a figure that changes between visits is a fault.

A single completeness score hides where the holes are. Camberly records completeness separately for identity, pricing, powertrain, safety, media and ownership, and a vehicle page states its own gaps rather than leaving blank cells.

A vehicle is measured against relevant competitors, never the whole catalogue, and every benchmark names the group it was measured against. “Above average” with no stated comparison set is meaningless.

• A benchmark needs at least three comparable vehicles. Below that, none is published.
• A difference smaller than 2% of the group median is reported as “in line with”, not “better than” — topping a flat range is not a strength.
• Bakkies are compared within their own body configuration. A single cab is measured against other single cabs, not against double cabs.

Every score carries its methodology, that methodology's version, the raw inputs it used and how much of its input was available. Value deliberately does not treat the cheapest car as the best value: it weighs equipment, warranty and running costs against price.

A score built on less than half of its inputs is withheld rather than published, because a confident number resting on a minority of its inputs invites more trust than it has earned.

When a methodology changes, the version changes with it, so an older assessment stays explicable rather than becoming mysterious.

A change to a price or specification is recorded as an event with its old value, new value, source, reason and author. The system then marks every derived value that depended on it — scores, ladders, assessments — as stale, and recomputes them.

Anything written by a language model is marked stale and held for human review rather than silently regenerated. Replacing one unreviewed claim with another is not a correction.

Product usage is measured first-party, in aggregate, to answer whether the product helps: search-to-vehicle rate, guided-discovery completion, which tools get finished. §148 of Camberly's specification lists the metrics; none of them require knowing who anyone is, and the implementation records nothing that could identify a person.

Commercial relationships do not and will not alter factual data, scores, rankings, comparisons or suitability assessments. Camberly currently has no commercial relationships of any kind.