Methodology

How we collect data, compute statistics, and construct our proprietary indexes.

Principles

  • Statistical slices, not prices. We publish medians, percentiles, standard deviation, and aggregated metrics — never individual listings or sales.
  • Cross-platform aggregation. Sales data is normalized across multiple secondary-market platforms and auction houses, then quality-checked for outliers.
  • Transparent methodology. All construction rules — universe, weighting, rebalancing, chaining, buffers — are documented and versioned.
  • Public data only. Brand websites (via Wayback Machine), public auction results, free APIs (eBay Browse, Google Trends, Reddit). Photos sourced from press kits or licensed.

Data sources

Brand websites + Wayback Machine

Current and historical retail prices via direct scraping and Internet Archive snapshots back to 2015.

Auction houses

Sotheby's, Christie's, Bonhams, Heritage Auctions, Phillips — publicly published auction results, our most reliable data source.

Secondary market platforms

Fashionphile, Rebag, Vestiaire Collective, The RealReal, Yoogi's Closet — sold listings aggregated into normalized statistics. Individual listings are never republished.

eBay Browse API

Active listings as a proxy for asking-price distributions, accessed via the free official API.

Google Trends + Reddit

Hype signals — search interest and community mentions — combined into our Hype Index.

European Central Bank

Daily FX reference rates for cross-currency price normalization.

Statistical aggregation

For each variant and time window we compute medians, means, percentiles (10/25/50/75/90), standard deviation, and sample size. Outliers are removed using a 3 × IQR filter before statistics are computed. Aggregations require a minimum sample size of 10; confidence is marked high at 100+, medium at 30–99, and low below 30.

Where sample is insufficient at the variant level, we roll up to the model level (across all materials and colors) and document the aggregation in the source metadata.

Index construction

Our flagship Bag Market Index and sub-indexes follow a transparent construction methodology:

  • Universe: 100 most liquid components by trailing 12-month sales volume.
  • Weighting: Square-root volume scaling, capped at 10% per component, floor 0.3%.
  • Rebalancing: Weights every quarter (Jan/Apr/Jul/Oct), composition review semi-annually (Jan/Jul).
  • Buffer rules: ±20% threshold to prevent flip-flopping at the boundary.
  • Chaining: Divisor adjusted on each rebalance to ensure value continuity.
  • Survivorship bias: Historical composition preserved for accurate backward extension.

Versioning and audit

Methodology versions are tracked in our index methodology log. Changes are announced publicly 30 days before they take effect. Each rebalancing event is recorded in our audit log with before/after composition and weights.