Methodology
The K15 Market Benchmark™, explained.
One number per model, powered by proprietary market intelligence and built from verified worldwide auction sales. No opinions, no unverified estimates — just the market record, measured.
Definition
The K15 Market Benchmark™ is Karbon Fifteen's proprietary measure of a model's current market value — powered by proprietary market intelligence and built from verified auction sales worldwide, in every major currency, normalized to USD including buyer's premium.
"Verified" means every input sale was confirmed against the primary auction-house record. Sales in GBP, EUR, CHF, JPY, and other currencies are converted at the exchange rate at the time of sale, so every model is comparable on one scale. The computation itself — how sales are screened, weighted, and combined — is proprietary to Karbon Fifteen. When a model has too few verified sales to support a reliable figure, no benchmark is shown.
From hammer to benchmark.
Collect
We continuously ingest public auction results from the world's leading auction houses and sale platforms, across every major collector-car market.
Verify
Each sale is checked against the primary auction-house record — price, car identity, sale date, and seller location. Unverifiable results never enter the dataset.
Screen
Cars that don't represent the production model are excluded and documented: prototypes and factory mules, salvage or rebuilt titles, engineless project cars, replicas, and parts lots.
Compute
Our proprietary K15 algorithm weighs each verified sale — screening outliers, adjusting for recency and sale context — to produce one benchmark per model.
Track
Trends and forecasts are computed only from that verified base, and an automated integrity sweep re-examines every published figure nightly.
Exclusion policy
A real sale isn't always a representative one.
Some genuine, verified sales still say nothing about what a standard production car is worth. A pre-production factory prototype sold on a bill of sale, a car rebuilt after being declared a total loss, or an engineless project pulled from decades of storage will all trade far from the real market — and would quietly distort any average that included them.
When our audits identify such a car, the sale is excluded from every benchmark, trend, and forecast, and the exclusion is documented with the reason and the primary source we verified it against. The sale price itself is never altered — we exclude, we don't edit.
Excluded categories include: prototypes and development mules, salvage/rebuilt and branded titles, non-running project cars, replicas and tributes, race cars sold as parts, and memorabilia lots.
Regional intelligence
Global, USA, and Rest of World — three honest views.
The same model often trades differently in America than it does in Europe, the Middle East, or Asia. Every verified sale is assigned a sale region from its primary record — the physical auction venue, or the seller's documented location for online sales — so each model carries three independent reads: Global, USA, and Rest of World.
Regional pools are smaller by nature, so regional figures are computed more conservatively and always display their sample size — you can judge the depth of the data behind every number.
Proprietary computation
Open about the inputs. Protective of the recipe.
We publish exactly what goes into the K15 Market Benchmark™: verified public auction sales, worldwide and in every major currency, normalized to USD including buyer's premium. Every input is traceable to a primary auction-house record you can check yourself.
How those sales become one number is Karbon Fifteen's intellectual property. The K15 algorithm screens statistical outliers, weights results by recency, and adjusts for sale context — restorations, competition variants, provenance-driven premiums — before producing the benchmark. The specific weightings and screening thresholds are proprietary and continuously refined against realized market results.
What we will always disclose: the benchmark uses only verified sales, it is powered by proprietary market intelligence, and it is withheld entirely when the data is too thin to support it.
The rules we never break.
Verified sales only
Every price behind the benchmark traces to a published, primary auction-house result — hammer plus premium in the sale's original currency, normalized to USD.
Outlier-resistant by design
A single headline record can distort a market read. The K15 computation is built to reflect what the middle of the market actually pays.
Recency-weighted
The computation favors the latest verified results, so the benchmark moves with the market rather than anchoring to history.
No number without data
If a model lacks sufficient verified sales, we show 'Insufficient history' — never a filler figure or an unverified estimate.
A benchmark never exceeds the record
Hard rails bound every computation: a benchmark can never sit above the model's verified record sale, and markets with no recent sales publish no figure at all.
Composition-aware trends
When a swing in the mix of cars sold — rather than prices themselves — would distort a trend, the computation detects it and switches to a mix-insensitive method.
The numbers around the benchmark.
12-Month Trend
Annualized price movement derived from recent verified sales, with guards against thin data and composition swings. Requires enough sales history to be reliable; otherwise we show 'Insufficient history'.
Forecast
A 12-month projection from proprietary trend models trained on decades of transactions. Always labeled as a projection — never a promise.
Record
The highest verified public sale for the model, with the auction house and date it was set.
Sample size
Every regional figure shows the number of verified sales behind it, so a deep market and a thin one are never confused.
Private / off-market estimate
Shown only where verified data supports it; quiet sales historically clear below auction headlines.
Continuous auditing
Every published number is re-checked nightly.
An automated integrity sweep re-derives every benchmark, trend, and forecast from the raw verified sales each night and tests them against a battery of invariants: no benchmark above its record, no figure published from stale or insufficient data, regional pools that reconcile exactly with the global pool, and trends that stay within plausible bounds.
Anything the sweep flags is investigated against primary sources before it can reach the site. Extreme results — unusually cheap or expensive lots — are individually verified on the original auction-house page and either confirmed as genuine market data or excluded with a documented reason.
Frequently asked questions
Common questions, straight answers.
See the benchmark in action.
Every covered model — Powered by Proprietary Market Intelligence.