Data Methodology

1. Coverage

The database covers AI startups, investors and people globally, plus four market-event types: funding rounds, IPOs, M&A, and fund raises — with a focus on the US and China.

Counts on the homepage and database pages are live structured counts that change with daily updates; the "cumulative tracked" and "community" figures on the membership page are historical / cross-channel aggregates. The two differ and are labeled in place.

2. Confirmation criteria

Funding events are either confirmed or rumored: rounds reported by media but not officially confirmed are marked Rumored and excluded from public feeds, the explorer and charts; they convert to confirmed upon official announcement or multi-source verification.

Amounts use officially disclosed values where available; ranges or local-currency descriptions are estimated to USD from public information. Where no reliable estimate exists we keep the original wording rather than invent a number. Missing valuations and exact dates are shown as undisclosed.

3. Sources & update cadence

Sources include official company and investor announcements, major tech / venture media, regulatory and exchange disclosures, and SVTR community applications and member feedback (used for cross-checking, never as sole confirmation).

Market events are scanned daily and reviewed by humans before entering the database; company and people profiles are continuously backfilled.

4. Known limitations

Some historical records lack stage, valuation or exact dates — these appear as "undisclosed" in charts rather than being guessed, with their share labeled next to each chart.

Structuring of company descriptions and people careers varies and is being backfilled; investor participation records reflect structured entries only, not yet exhaustive.

5. Rating methodology

Startups that pass review receive an S / A / B rating; investment firms receive T1 / T2 / T3. Startups are scored on five dimensions — team, market, product & technology, moat, and traction — with weights adjusted by funding stage (the earlier the stage, the more team weighs). Investment firms are scored on portfolio support, network, capital, track record, and brand signal. Ratings combine AI and human review, carry a rating date, and can be re-assessed as the company evolves.

Two kinds of subjects are rated. One: projects whose founders request a rating. Two: startups in the database — assessed independently along the funding flow from public information (funding announcements, official sites, products, public interviews), with no submission required; the rating reflects the company’s own differentiation, and is not raised just because a well-known firm led the round.

The full tier, one-line rationale, and five-dimension detail are available to members; non-members see only that “SVTR has evaluated” the company, and unlock the full judgment on subscribing.

Evidence first: every judgment needs support. Self-reported material that cannot be cross-verified earns no tier — it is marked "insufficient signal" and can re-enter rating once more material is provided; the top tier requires at least one publicly verifiable strong piece of evidence. Insufficient signal and falling short of the bar are distinct states, labeled separately.

Neutrality: ratings cannot be bought. Payment and partnership records live in an isolated layer the scoring system cannot read — the rating process does not know, and does not care, whether a company pays us. Where a company has a business relationship with SVTR, the rating card carries a plain disclosure of it, but the disclosure does not affect the tier.

Exact weights and thresholds are not published, to prevent gaming. To dispute a rating, email allen@svtr.ai and we will re-review.

6. Private signals

Some company pages and the Signals boards show signal strength as a 1–3 bar indicator, derived from three signals: content heat (mentions in published SVTR content over the past 180 days), activity (the company's applications/intros within the SVTR ecosystem over the past 90 days, combined with its recent public market activity such as funding raised or deals made), and interest heat (users who followed or requested an intro to the company in the past 90 days, deduplicated by person).

Only tiers are shown publicly — never raw counts. Tiers are computed daily by fixed thresholds and recalibrated quarterly. Membership and partnership records live in an isolated layer the signal system cannot read, and there is no manual override.

Signals reflect attention and activity within the SVTR ecosystem and are not investment advice.

7. Corrections & usage

Spotted an error or want to add information? Email allen@svtr.ai and we will verify and correct.

Data is for research reference only and is not investment advice; limited quotation with attribution is welcome, while bulk scraping and resale are prohibited (see Terms of Service).

8. How to cite SVTR Signals

SVTR Signals comprises four boards: Startup / Investor / People (private signal tiers) and Articles (reading heat). svtr.ai/signals shows each board’s live daily Top 10 by default (same source and boards as the home and section pages), and also archives a frozen snapshot every Monday for long-term citation and traceability, with past issues reachable via ?week=YYYY-MM-DD.

When citing in media, broker notes or research, please include the board name, issue date, source, and methodology link. Standard format:

SVTR Signals · Startup Signals (2026-07-13), SVTR.AI, https://svtr.ai/signals?week=2026-07-13 , methodology: https://svtr.ai/methodology

Cadence: a snapshot is generated automatically every Monday. Archived issues are immutable — even if an entity is later merged or unlisted, past issues stay as recorded, so citations remain verifiable over time. Positions show only ranks and tiers, never raw counts or weights.