What is measured
Five categories of input feed every score on the platform. Each has its own units, coverage profile and characteristic way of being wrong.
A score is only as good as the fields underneath it. This page documents those fields — what each one means, what unit it is in, and the specific way each category degrades. The sources page covers where the data comes from and how often it refreshes; the pipeline page covers what happens to it in between.
Market data
The densest and most reliable category. Prices and volumes are reported continuously by many venues, and the aggregation problem is well understood.
| Field | Unit | Definition |
|---|---|---|
| price | USD | Last observed price |
| marketCap | USD | Circulating supply times price |
| volume24h | USD | Reported traded value, trailing 24 hours |
| volatility | Fraction | Realised volatility. 0.04 means 4 % |
| marketDominance | Percent | Asset market cap as a share of total market cap |
Market capitalisation depends on a circulating supply estimate, which is a provider judgement about which tokens to count — a revision changes every ratio built on it with no change in the market. Reported volume is unverified and inflated by wash trading where it occurs. Both effects hit small assets hardest. See liquidity research.
Social data
The category with the most processing between the raw observation and the stored field, and the one that drives the most metrics.
| Field | Unit | Definition |
|---|---|---|
| socialVolume24h | Count | Posts mentioning the asset in 24 hours |
| interactions24h | Count | Views, likes, comments, shares and reposts on those posts |
| socialDominance | Percent | Share of all measured crypto social volume |
| sentiment | 0–100 | Aggregated classification of tone across collected posts |
Coverage is limited to collected platforms, so a community elsewhere reads as zero. Ticker collisions add unrelated conversation for assets with generic symbols. Sentiment is a model output whose reliability drops on ironic and crisis-era text. No authenticity filter is applied, so coordinated activity counts like organic activity. See social intelligence and AI agents.
Community data
Off-chain measures of following and participation — subscriber and follower counts, and engagement relative to community size. Slower-moving than social data and answering a different question: not what is being said today, but how many people are attached to the project at all.
Follower counts are cumulative. They rise when someone joins and rarely fall when that person loses interest, so a dormant community retains its headline numbers indefinitely. They are also purchasable. Engagement relative to community size is the more informative measure and the harder one to manufacture. See ecosystem research.
Developer activity
Public repository signals: commit activity, contributor counts, repository counts, and how those series behave over 30 and 90-day windows.
Commit counts are fully controlled by the party being measured and are trivially inflated. Private development is invisible, so a well-funded team working toward a release can appear inactive. Forks and archived repositories inflate counts. Persistence across quarters is more informative than any level reading.
News
Headlines and article metadata from crypto media, deduplicated across outlets, with recency and the assets each item touches. This category supports the platform's news and trending features rather than the numeric scores.
The same story is republished across many outlets, so raw counts measure syndication rather than significance. Deduplication reduces this and cannot eliminate it, because rewritten coverage is not textually identical. Sponsored content is not always distinguishable from editorial coverage.
What is deliberately not collected
Naming the gaps matters as much as listing the fields, because several metric names imply data the platform does not have:
- No on-chain data. No wallet balances, transaction flows, holder distributions or exchange flows. This is why WSD is a turnover comparison rather than a whale-tracking metric, despite its name.
- No order book data. Depth, spreads and book imbalance are not collected, so no metric measures depth — only turnover.
- No positioning or derivatives data. Open interest, funding rates and actual leverage are not collected. Liquidation visualisations are model outputs under assumed leverage, not measurements.
- No personal data from social sources. Metrics are aggregate counts and classifications, not profiles of individuals.
Quality summary
| Category | Coverage | Latency | Main risk |
|---|---|---|---|
| Market | High across tracked assets | Minutes | Supply estimates, unverified volume |
| Social | Uneven — platform dependent | Minutes to hours | Missing communities, synthetic activity |
| Community | Good for major projects | Daily | Cumulative counts, purchasable |
| Developer | Public repositories only | Daily | Private work invisible, counts gameable |
| News | Broad across crypto media | Minutes | Syndication, sponsored content |
Related
- Data sourcesWhere the data comes from and how often it updates
- Data pipelineWhat happens between collection and score
- MethodologyHow these fields become metrics
- Research areasWhere each category is examined in depth