Methodology

1. Scope

What Pursely covers, explicitly, and what it does not.

• Covered: Hermès handbags listed on major US-based resale platforms.
• Not covered: Other luxury brands, primary retail, international resale platforms outside the US, accessories beyond handbags, and peer-to-peer marketplaces.

Why this scope: Depth over breadth. Hermès handbags represent the highest-value, most-data-rich segment of the secondary luxury market, and concentration enables more reliable cohort analysis than broader coverage would.

2. Sources

• Currently tracked: Fashionphile, The RealReal, Madison Avenue Couture, Rebag.
• Expanding coverage: 1stDibs, Sotheby's.

All data is drawn from public listings. Pursely does not access private inventory, private sale records, or non-public platform data.

Independence: Pursely maintains no commercial relationships, affiliate agreements, paid placements, or referral arrangements with any of the platforms it covers. Pursely does not receive payment from, and does not pay, any covered platform.

3. Collection cadence and lifecycle tracking

Listings are monitored continuously throughout their lifecycle: from initial appearance, through any price changes, to eventual sale or removal.

Historical observations are preserved indefinitely. A listing that was live in January and sold in March remains in the dataset, with all intermediate price and status events recorded.

Continuous data collection has been in place since December 2025.

Lifecycle states tracked

Each listing transitions through a defined set of observable states:

• Available: actively listed and offered for sale.
• On hold: temporarily reserved, pending payment, or otherwise held by the platform.
• Sold: final transaction recorded by the platform.
• Relisted: previously sold listings that return to available status.

The relisted state is tracked deliberately. A listing that briefly disappears as sold and reappears as available, due to cancelled sales, returns, or routine platform processing, is recorded distinctly from a true sale. Treating relistings uniformly as sales would systematically overstate market velocity and distort sell-through analysis. This distinction is one of several reasons Pursely's lifecycle data differs from snapshot-based collection.

4. Data points captured

What Pursely records for each listing.

5. Normalization

How cross-platform data is made comparable.

• Color taxonomy: All listing colors are mapped to Pursely's color family taxonomy. Multi-color listings, including bicolor or contrast-handle bags, are flagged separately and excluded from cohort-level market analysis.
• Condition tiers: Each platform's condition vocabulary is mapped to Pursely's internal five-tier scale to enable cross-platform comparison.
• Sizes and models: Hermès model names and sizes are normalized to canonical forms, such as Birkin 25 and Kelly Sellier 28.
• Currency: All prices are reported in USD. Non-USD listings are excluded from aggregate market-value calculations.

Taxonomy maintenance

Hermès releases new colors and leathers periodically through seasonal collections and limited releases. Pursely's taxonomies are reviewed regularly and expanded to incorporate new materials as they enter the resale market. New values are recorded in the underlying data from first appearance; their incorporation into the published taxonomy and aggregate analyses follows after sufficient observations are available to support reliable categorization.

Pursely deliberately does not disclose its specific normalization implementation, vendor stack, or extraction process. The taxonomies, tier definitions, and exclusion rules are disclosed; the technical pipeline is proprietary.

6. Data quality and validation

Pursely uses large language models for extraction and classification of listing data. The methodology below establishes how that automated work is verified before data enters the published dataset.

• Multi-stage validation. Every listing extracted into the Pursely dataset passes through multiple automated validation stages before becoming part of published market analyses. Stages include schema conformance checks, distribution-based anomaly detection against historical cohorts, and cross-model agreement review.
• Cross-model verification. Critical extraction and classification tasks are performed independently by more than one large language model, drawn from different vendors to reduce correlated failure modes. Disagreement between independent passes is treated as a high-signal flag and routed for review rather than silently averaged or resolved by majority rule.
• Drift monitoring. Extracted values are continuously compared against historical distributions for the same cohort. Statistical drift in extracted attributes, such as a sudden change in the proportion of listings categorized as a specific leather, triggers review of the affected pipeline stage.
• Held-out evaluation. Pursely maintains a held-out reference set of manually verified listings used for ongoing evaluation of extraction and normalization accuracy. Material changes to the pipeline are evaluated against this set before being put into production.
• Human review of edge cases. Listings flagged by any of the above mechanisms are reviewed manually before inclusion in published analyses.

Pursely does not disclose specific accuracy metrics, model identities, or vendor relationships in this public methodology. Quantitative quality reports are available to institutional clients under data-license agreements.

7. Market analysis

How aggregate market figures, such as price ranges and market values, are produced.

• Cohort definition: Listings are grouped into cohorts by model, size, leather family, color family, and condition tier.
• Minimum sample: Cohorts with fewer than five qualifying listings are not used for published market ranges.
• Pool composition: Cohort analysis uses a combined pool of currently live, transitioning, and recently sold listings. This combined pool is more stable than any single status alone.

Exclusions from cohort analysis

• Multi-color listings.
• Listings missing essential attributes, including size, leather, or condition.
• Listings without a disclosed price.

Outlier handling: Statistical outliers are flagged for review but are not silently excluded from published ranges.

8. Pricing definitions

Plain-language definitions of every pricing concept Pursely publishes.

• Listed price: The current asking price as displayed on the source platform.
• Sold price: The final disclosed price at which a listing was sold, where the platform discloses it.
• Market range: The aggregate cohort-level price range, calculated as described in Section 7.
• Price history: The complete sequence of price changes for an individual listing, with timestamps.
• Tracked market value: The cumulative dollar value of all listings Pursely has indexed to date, including listings that have since been sold.

9. What Pursely does not do

• Pursely does not operate as a marketplace and does not facilitate transactions.
• Pursely does not earn commissions, affiliate revenue, or referral fees from any covered platform.
• Pursely does not guarantee the accuracy of any individual listing.
• Pursely does not provide buy, sell, hold, or trade recommendations.
• Pursely does not provide investment, financial, or sourcing advice.
• Pursely does not aggregate user opinions, ratings, or sentiment.

10. Limitations and disclosures

• Coverage is US-focused. International resale markets, including European auctions and Asian platforms, are not currently represented.
• Pursely relies on covered platforms accurately representing their inventory. Listing errors at the source propagate to Pursely's data until corrected.
• Sold-price data is only as complete as platforms choose to disclose. Some platforms publish final sale prices; others do not.
• Past prices are not indicative of future prices. Market data describes the historical and current market and does not forecast it.
• Information published by Pursely is research and reference material. It is not investment, sourcing, authentication, or legal advice.

11. Versioning and updates

Methodology is versioned. The current version is v1.0.

Material changes, including changes to scope, sources, normalization rules, or analytical methodology, increment the version number and are logged with effective dates.

The current version is always available at /methodology. Prior versions are archived at versioned URLs.

Non-material changes, including typos and clarifications, are made silently with the page's last-updated date refreshed.

12. Contact

Questions about this methodology may be directed to contact@pursely.org.