Index Methodology
Understand how Artmetria structures market and object-level signals through a careful analytical framework. Scores are signals, not verdicts, predictions or recommendations.
Methodology · Hub
How Artmetria Works
A clear introduction to Artmetria's method: how the platform reads signals, uses AI, explains coverage and makes its limits visible.
Artmetria is built as a method before it is presented as a product. The platform connects looking, learning, market context and structured analysis without claiming certainty where the art market remains partial, fragmented or difficult to compare.
This page brings together the main methodological notes behind the beta.
A signal, not a verdict.
Artmetria is educational and analytical. It does not authenticate, appraise, predict future value, or give financial advice. Every output is a reading meant to support judgement — read it through its limits.
- A signal is not a verdict.
- A recommendation is an invitation to explore, not an instruction.
- A limitation is part of responsible interpretation, not a hidden weakness.
From a raw record to an explainable reading
The same six steps run behind every value. You can follow the process without seeing the proprietary weights — the point is that it is inspectable, not that it is secret.
- 01Input
Auction results and a museum-scale reference corpus arrive as raw records — prices, lots, artists, images, provenance lines.
- 02Normalisation
Records are cleaned and mapped to a common scale: duplicates removed, currencies and units aligned, artists and mediums reconciled.
- 03Qualitative layer
Each object is read against explicit criteria — artist, period, execution, condition, provenance, critical resonance — not price alone. How those criteria are weighted stays proprietary.
Go deeper → - 04Aggregation
The per-object readings are composed into a synthetic value and a market reading, so one number reflects many signals rather than a single data point.
Go deeper → - 05Coverage & confidence
Every output carries how much data backs it — a coverage state (canonical → insufficient) and a 95% confidence interval — so a thin dataset can't hide behind a confident-looking number.
Go deeper → - 06Output
An explainable reading — a value, its band and its coverage — meant to inform judgement. Never a verdict, a valuation, or financial advice.
Go deeper →
Coverage & Explainability
See how Artmetria distinguishes active coverage, monitored sources, partial data and confidence context. Coverage is part of the method, not a hidden technical detail.
AI Methodology
Learn how Artmetria uses AI as an interpretive assistant for aesthetic profiling, artwork reading, discovery and context - not as an authority, certificate or financial advisor.
Scope & Limits
Read how Artmetria handles beta status, partial coverage, data limitations, AI uncertainty and responsible use without overclaiming.
Artmetria Engine
Discover the structured market memory behind Artmetria: the layer that helps organise market context, coverage and uncertainty.