Eight official sources, cross-referenced and scored — explained.

Every source is a primary, publicly-available official record — not a scrape of someone else's summary:

House-only note:congressional trade data covers the House of Representatives only. The Senate's equivalent public disclosure source is not currently structured enough to ingest reliably — this may be revisited later, but for now, any Congress signal you see here is House disclosures specifically.

Every signal is tagged with a ticker and a date. When two or more independent signal types land on the same ticker inside a compressed time window, that's convergence — the whole premise of the product. One signal is trivia; stacked, independent signals are conviction.

Direction conflicts — for example, insider selling alongside a fund adding a new position — are shown honestly as "mixed," never suppressed or hidden to make a ticker look cleaner than the data actually is.

Each ticker gets a 0–100 Convergence Score, computed from how many distinct signal types are active in a rolling 90-day window, weighted by rarity:

When bullish and bearish signal types are both active, the score is reduced (never hidden) to reflect conflicting conviction. Each ticker gets a plain-English badge — e.g. "3 independent bullish signals · 41 days."

A cluster is flagged when 3 or more distinct insiders file Form 4 transactions for the same ticker within any rolling 5-day window. Transaction floors filter out noise: trades under $5,000 are ignored entirely; VP/Director trades must exceed $10,000; CEO/CFO trades must exceed $25,000. The minimum aggregate cluster value before flagging is $50,000.

Each cluster then receives a weighted composite score and a grade: A (≥82), B (≥52), or C(<52). Six factors contribute:

Signals are rare — and that's the point

Most tickers show nothing at all on any given day. True insider buy clusters happen roughly a dozen times a year across the entire market. That scarcity isn't a flaw — it's the signal. A tool that flags something on every stock, every day, is measuring noise, not conviction. ClusterSignal is built to say nothing most of the time, and to say something worth reading when it does.

Every ticker with active signals has a Story page: a vertical timeline of every event across all 8 signal types, newest first, each linking to its primary source (EDGAR, USAspending, clinicaltrials.gov, USPTO). Insider clusters additionally have their own detail page — a 90-day price chart with the cluster window highlighted, the full insider trades table with SEC EDGAR links, score breakdown, company context, and signal flags (dilution risk, earnings proximity, volume spike, filing speed).

Signed-in users also see insider track records — prior 90-day returns for each insider on this ticker, direction-matched (prior buys for a buyer, prior sales for a seller) — and a generated signal summary.

Ticker Story and cluster detail pages include an AI analysis section that, on demand, generates a plain-English summary of what the pattern means in context. A chatbot can fetch any one of these feeds on request — it doesn't watch eight feeds continuously, cross-reference and score them, and alert the day the third signal lands. Persistent structured monitoring is the moat, not the data. Every analysis includes a disclaimer that this isn't financial advice.

This feature is temporarily disabled.