Data quality at DeFi Intel — every fact, sourced. Every entity, linked.
DeFi Intel publishes a typed entity graph of 11,787 entities, 26,104 typed relations and 3,354 dated events covering DeFi, stablecoins, RWA and on-chain infrastructure. Every entity carries a source-confidence score on a 1-5 scale, a last-reviewed timestamp, and at least one corroborating trace from one of our 184 independent research agents. This page is the public dashboard for everything that backs that claim — the methodology, the freshness SLAs, the corrections process, and the gaps we know we still have. If something on this site is wrong, email corrections and we'll fix it on the public record.
Why this page exists. Most DeFi data products are black boxes. We're the opposite: every entity is sourced, every relation is typed, every cross-reference resolves, and the methodology is public. If you want to audit how a number got into our graph, that audit trail starts here.
Live stats
11,787
entities
26,104
typed relations
3,354
dated events
1,800
news items
237
long-form entries
722k
corpus words
184
research agents
100%
xref resolution
Last build: 2026-05-03 · 0 broken cross-references · 12 research scopes · quarterly entity review pass active.
Coverage by entity type
Top entity types by population, with attribute fill-rate, source URL coverage, alias coverage and average source-confidence (range 0.0–1.0, where 1.0 = score-5 official filing).
| Type | Count | Attrs | Source URL | Aliases | Avg conf |
|---|---|---|---|---|---|
| person | 2,359 | 98.4% | 95.5% | 5.6% | 0.92 |
| company | 1,920 | 94.4% | 85.8% | 40.8% | 0.83 |
| paper | 1,268 | 98.3% | 92.5% | 62.4% | 0.92 |
| protocol | 1,247 | 95.7% | 76.0% | 46.8% | 0.83 |
| event | 913 | 96.9% | 87.5% | 13.6% | 0.87 |
| product | 884 | 94.5% | 75.6% | 42.9% | 0.79 |
| token | 698 | 94.0% | 78.1% | 53.0% | 0.83 |
| chain | 352 | 69.9% | 89.5% | 20.7% | 0.90 |
| license | 327 | 100.0% | 99.4% | 14.4% | 0.91 |
| fund | 295 | 98.3% | 90.5% | 77.3% | 0.85 |
| regulator | 174 | 79.3% | 89.1% | 77.0% | 0.92 |
| dapp | 166 | 100.0% | 99.4% | 74.1% | 0.85 |
| exchange | 115 | 87.0% | 88.7% | 36.5% | 0.85 |
| regulation | 90 | 100.0% | 73.3% | — | 0.89 |
| dao | 87 | — | — | — | 0.85 |
Long-tail (78 additional types: bridges, regulations, jurisdictions, lawsuits, audits, etc.) follows the same scoring. Full coverage table available via /api/ or as a CSV in the licensing tier.
Source-confidence scoring (1–5 scale)
Every entity-attribute pair carries a source-confidence score. We aggregate the per-attribute scores into an entity-level avgConf. The scale is deliberately strict: a brand-homepage claim is never enough to clear score 3.
- 5 Official filings, on-chain commits, peer-reviewed papers, audited contracts. Examples: SEC 10-K, FINMA registry, GitHub commit hash, deployed verified contract on Etherscan, working-paper DOI from NBER/SSRN.
- 4 Primary docs from operators. Governance forum proposals signed by core-contributor wallets, official whitepapers, signed press releases on company letterhead, formal tweets from verified founder accounts referencing on-chain events.
- 3 Brand homepage, official blog, verified social. The homepage of a protocol claiming a TVL number; an official Mirror post; a Discord announcement from a verified moderator role.
- 2 Established secondary reporting. CoinDesk, The Block, Decrypt, DL News, Bloomberg, Reuters. Quoted but not the primary source.
- 1 Anonymous Twitter, unverified leaks, single-source rumour. Captured for completeness; never a sole basis for a published claim.
Entities with avg-conf < 0.7 are quarantined out of headline aggregations and footnoted as "preliminary". Entities with avg-conf ≥ 0.85 are eligible for the high-confidence flag.
Last-reviewed dates
Every long-form research entry (237 entries, 722k words across 12 scopes) carries an explicit lastReviewed timestamp in the front-matter. After 90 days, the entry is surfaced for re-review. After 180 days without a re-review and no contradicting events, it is auto-flagged "stale" in the UI. The 12 research scopes — DeFi protocols, stablecoins, RWA, infrastructure, regulators, exchanges, MEV, ZK, restaking, gaming, DePIN, and policy — each have a named scope-owner reviewer.
Corrections process
If you find a fact that's wrong, here's the path:
- Email [email protected] with the entity URL and the disputed fact (a sentence is fine; a screenshot is better).
- We acknowledge within 24 hours and triage within 72 hours.
- We either correct the fact, annotate the entry with the dispute, or respond with the supporting source — whichever the evidence dictates.
- Every correction is logged on the public changelog with date, disputed fact, resolution and source change.
We do not silently rewrite history. Past versions of entries are retained and accessible via the API.
Independent research-agent corroboration
The 184 independent research-agent traces are the structural backbone of DeFi Intel. Each agent runs on a different seed, prompt template and source-mix policy — so a single bad prompt or a single bad source cannot propagate through the corpus undetected.
- Per-entity rule: every entity must have ≥1 corroborating agent trace before it ships to production.
- High-confidence flag: entities supported by 3+ independent agents are flagged in the UI and in the API response (confidence.tier = "high").
- Audit trail: agent traces are versioned and retained. Pro / Enterprise customers can request the trace bundle for any entity (typically returned within 4 hours).
- Methodology document: the full agent-trace methodology is publicly documented at /methodology.
Freshness SLAs
| Stream | Cadence | Owner |
|---|---|---|
| News refresh | Every 10 minutes (cron) | news pipeline |
| Entity profile review | Quarterly minimum | scope owner |
| Regulatory changes | Within 48 hours of source filing | policy desk |
| Exploits / incidents | Within 12 hours of confirmation | incidents desk |
| Funding rounds | Within 24 hours of announcement | company desk |
| Long-form re-review | 90 days | scope owner |
| Build pipeline | Daily; xref-integrity gate at every build | infra |
Data audits and reports
The internal audit pipeline runs on every build. It verifies xref integrity (currently 100% resolution, 0 broken), attribute-fill coverage, source-URL liveness and confidence-score distribution. Audit logs are retained.
External audit: as of 2026-05-03 no third-party data audit has been commissioned. We will commission an independent audit at 25,000 entities or end-2026 — whichever comes first. The full report will be linked here when it lands.
Open data policy
- Free: every one of the 11,787 entity profiles is browsable on the public web. The /api/ public endpoints are rate-limited but free for non-commercial use. CSV export of public categories is free.
- Pro ($49/mo): full graph export in Parquet/JSON, historical news archive, full long-form corpus, daily updates. See licensing.
- Enterprise (custom): Pro + on-prem mirroring, named SLAs, custom enrichment, AI training license terms.
- AI crawlers: welcomed under our public licence with citation. See licensing and /llms.txt.
Cross-links
Built by Round-9 Lane-13 (Trust pages). Last reviewed 2026-05-03.