Igor Grachev
Sections

Marketplace Metrics Dictionary

Listing quality

A construct rather than a single measure: the degree to which a supply unit gives a demand participant sufficient and accurate information to decide. It is measured through a declared model of observable completeness, accuracy, and freshness indicators, published separately.

hypothesisVersion 0.1Published: 25 July 2026Updated: 25 July 2026

Names and aliases

supply unit information quality, listing completeness and accuracy

Formula

no single formula exists; the published output is a vector of indicators — required-attribute completeness, share of discrepancies with the actual item, price and availability freshness, and moderation-edit share

Metric card

Unitvector of indicators; no composite score
Data sourceThe required and recommended attribute schema, listing snapshots, moderation events, disputes about description mismatch, and price and availability updates.
FrequencyBy eligible-supply snapshot and by listing-creation cohort.
Segmentsmicro-market, category, creation method, seller type, cohort

Interpretation errors

  • Equating completeness with quality: filled fields carrying inaccurate content damage the decision more than empty ones.
  • Collapsing the indicators into one score and using it in ranking: the platform will get a rising score without rising informativeness.
  • Treating quality as a property of the listing rather than of the listing-request pair: sufficient information depends on what demand is actually looking for.

Example

Illustrative vector for one cohort: 82% required-attribute completeness, 2.4% discrepancies with the actual item, 6% of listings carrying a stale price. The numbers describe no real platform.

Use

Define the required-attribute schema separately for each micro-market and publish the indicators separately: completeness, accuracy, and freshness are governed by different operational processes and must not hide inside one number. Test the model against outcomes: if listings with high completeness do not earn a higher qualified response rate, the schema needs revision rather than the listings.

If a score is nonetheless used in ranking, publish that as an editorial and product decision together with observation of what happens to the indicators after it ships. A metric that becomes a target stops describing reality; that effect should be measured rather than assumed away.

Anti-example

Requiring at least five photographs raises completeness to 95%. Sellers add repeated and irrelevant images, the share of discrepancies with the actual item does not move, and the qualified response rate stays flat. The completeness indicator rose; the informativeness of the listing did not.

Limitations

  • Accuracy is observable only indirectly — through disputes, refunds, and moderation — so the indicator is biased towards cases somebody contested.
  • Required attributes are defined by the platform schema, so values are not comparable across platforms or across schema versions.
  • The moment a quality score influences ranking it stops being a measurement and becomes a target, with every consequence Goodhart's law implies.
  • The construct has not been validated against data; until independent review it remains a working measurement model.

Related metrics

qualified-response-rate, match-quality, zero-exposure-share, refund-rate

Context sources

These works provide two-sided-market context; the formula above remains an author-defined working definition.

  1. Hanna Halaburda, Mikolaj Jan Piskorski, Pinar Yildirim. Competing by Restricting Choice: The Case of Matching Platforms (2017).
  2. Kong-Pin Chen, Yen-Chi Huang. A Search-Matching Model of the Buyer-Seller Platforms (2012).
  3. OECD, European Union, European Commission Joint Research Centre. Handbook on Constructing Composite Indicators: Methodology and User Guide (2008).