Coverage & Methodology

How to trust an ALT/FNDATA answer

What the connector's data covers, how current it is, how prices become comparable in USD, and the handful of rules that keep an answer honest. Read this alongside the limits and caveats in the technical docs.

What the data is

ALT/FNDATA is a transaction database for luxury and collectible assets. It records what things actually sold for, drawn from two kinds of transactions:

Public auction results from the major houses and hundreds of regional houses worldwide.
Private-sale transactions where the sale is disclosed, which most competitors do not capture.

It is a secondary-market record. It is not retail list pricing, not dealer asking prices, and not a listings feed. When a figure appears, it reflects a transaction, or in the case of an estimate, a house's pre-sale forecast clearly labeled as such.

Sources & coverage

Sources include Christie's, Sotheby's, Phillips, and Bonhams, alongside hundreds of regional and specialist houses. Coverage spans categories that share one wide schema, so the same query works across them: watches, jewelry and gems, handbags, fine art and works of art, wine and whisky, classic cars, motorcycles, aircraft, and more. Each category maps to its own table; see the dataset reference.

Check the source before you generalize Coverage is deep but uneven across houses and categories. When a result matters, group by vendor to see which houses the lots came from before drawing a market-wide conclusion.

Freshness & the recency caveat

New sales flow in continuously through the ingestion pipeline, but they do not all arrive at once. Results are captured, classified, priced, and reconciled before they land in the production tables, so the most recent one to two quarters are systematically undercounted. This is the single most important thing to hold when reading a trend.

The recency rule A drop in volume or lot counts in the latest quarters is almost always ingestion lag, not a real market decline. When categories fall in lockstep at the newest edge, that is the signature of under-ingestion. Treat the trailing periods as provisional and lead with the stable history.

Practically: ask Claude to flag the most recent quarters, compare against the same period a year earlier rather than the immediately prior quarter, and avoid turning a provisional edge into a headline.

USD normalization

Lots sell in many currencies. To make prices comparable, usd_price_decimal is normalized to USD using the nearest-date exchange rate to each sale, not a single fixed rate and not necessarily the exact same-day rate. Legacy currencies that predate the euro are routed through EUR. A converted figure can therefore differ slightly from a same-day conversion you might compute by hand; the trade-off buys consistency across decades and dozens of currencies.

Sold, unsold & estimates

Three fields decide whether a number means what you think it means:

status is sold or unsold. Filter to sold before reading any price.
usd_price_decimal is the realized price, populated on sold lots. On unsold lots there is no realized price to read.
sale_estimates_high_usd_price is the house's pre-sale high estimate. It is a forecast, useful for measuring pricing power as a ratio, but it is not an amount anyone paid.

Keeping unsold lots in view matters too: sell-through rate, the share of offered lots that sell, is a demand signal in its own right, and it needs the unsold lots in the denominator.

What we do not claim

It is not real-time. Expect the recency lag described above.
It is not retail or asking-price data. Marketplace list prices and dealer asks are out of scope.
It is not investment advice. The data is provided for research and reference; decisions and diligence remain yours.
Coverage is not uniform across every house and period. Verify the source mix before generalizing.

Questions on methodology Reach us at info@altfndata.com.