ClusterSignal

FAQ

Frequently asked questions

Answers to common questions about the 8 signal sources, the Convergence Score, insider cluster scoring, Short Radar, visitor access, and pricing.

What is a cluster?

A cluster is 3 or more insiders filing SEC Form 4 transactions for the same company within a rolling 5-day window. ClusterSignal requires at least 3 distinct insiders and a minimum $50,000 aggregate value before flagging a cluster. Individual trades below the noise floor ($5K general, $10K VP/Director, $25K CEO/CFO) are excluded.

What does BOUGHT vs SOLD mean?

BOUGHT means the cluster consists of open-market purchases (insiders buying shares with their own money). SOLD means insiders sold shares. These are informational labels — they describe what happened, not what you should do. Both types of clusters appear in the dashboard and have separate scoring logic. Not all sells are bearish; context matters.

What is Short Radar?

Short Radar is a special badge (shown in orange) reserved for sell clusters that pass additional filters: at least $500,000 in coordinated sales and 3+ distinct sellers — or $250,000+ if C-suite insiders (CEO, CFO) are among the sellers. C-suite participation lowers the threshold from $500K to $250K; the $250K floor applies regardless. Short Radar is not a sell recommendation — it flags activity worth investigating further.

What does the score mean?

The score is a weighted composite — no fixed ceiling — built from six factors: Insider Count (up to 60 pts — 15 per insider), Seniority (up to 30 pts — CEO/CFO score ~3× higher than VP or Director), Capital Deployed (up to 22 pts, scaled by $ amount), Win Rate (up to 22 pts, based on prior 90-day returns), Time Velocity (up to 15 pts — trades in 1 day score much higher than trades spread over 3 weeks), and Filing Speed (up to 12 pts — same-day filers get full credit). Grade A = ≥82, Grade B = ≥52, Grade C = <52.

Why are some small trades ignored?

Transaction floors filter out routine noise. Trades under $5,000 total value are excluded for all insiders. VP and Director trades under $10,000 are excluded. CEO and CFO trades under $25,000 are excluded. This ensures clusters reflect meaningful capital commitment, not token grants, dividend reinvestment, or trivially small purchases that would distort signal quality.

Why do CEO and CFO trades score so much higher?

CEOs and CFOs have the most complete and current view of the business. When they buy with their own money — particularly in amounts exceeding their noise floor — it is one of the strongest signals available in public data. Our seniority scoring reflects this: a CEO insider earns 30 points versus 10 for a VP and 7 for a Director, roughly a 3× multiplier.

What is a 10b5-1 plan and why does it matter?

A 10b5-1 plan is a pre-scheduled trading program an insider sets up in advance to sell shares at predetermined prices or dates, regardless of what's happening at the company when the sale actually executes. These trades are mechanical, not discretionary — the decision was made months earlier. ClusterSignal reads SEC's own 10b5-1 checkbox directly from each Form 4 (added to the form by the 2023 Rule 10b5-1 amendments), with footnote-text parsing as a fallback for filings that omit the tag. Any trade marked pursuant to a 10b5-1 plan is excluded from cluster formation and scoring entirely — not just from the Short Radar badge — so multiple executives' unrelated scheduled sales landing in the same week don't get mistaken for coordinated conviction. Separately, ClusterSignal also filters out same-day sell clusters that don't qualify as Short Radar — these almost always reflect mechanical RSU vesting events (cliff vesting, lockup expiry) rather than independent conviction.

What is the AI analysis?

Cluster detail pages include an AI analysis section that can generate a 2–4 paragraph plain-English summary covering: what the pattern means in context, what the insiders' seniority suggests about conviction, and what a prudent investor might watch for. It's generated once per cluster and cached, not regenerated on every page view.

Is this legal?

Yes. ClusterSignal uses only public SEC EDGAR disclosure data, which is freely available to anyone under federal law. The site is a research tool and does not collect or use insider-only information.

Is this financial advice?

No. ClusterSignal is not financial advice. The information is provided for educational and research purposes only. Always consult a licensed professional before making investment decisions.

Where does the data come from?

Eight official public sources, each ingested directly — never scraped from a third-party summary. Insider trades: SEC EDGAR Form 4 filings. Congressional trades: House of Representatives disclosures (House-only — the Senate isn't currently covered, see below). Trademark filings: USPTO. Activist stakes: SEC EDGAR Schedule 13D/13G. Fund positions: SEC EDGAR 13F-HR via the data.sec.gov submissions API. Government contract awards: USAspending.gov. FDA catalysts: clinicaltrials.gov. Aligned signals are derived on ingest by cross-referencing the other seven. Sector/industry classification and price context (market cap, P/E, current price, earnings dates) come from a mix of Yahoo Finance, Finnhub, and StockFit. Every displayed event links to its primary source for independent verification.

Why is House-only, not the full Congress?

House of Representatives disclosures are published in a structured, machine-readable format we can reliably ingest. The Senate's equivalent public disclosure source is not currently structured enough to sync automatically — every congressional trade signal on ClusterSignal today is House-only. This may be revisited via a paid data source if there's demand.

Why not just ask an AI?

A chatbot can fetch any one of these eight feeds if you ask it to, on request, once. It does not sit there continuously watching all eight, cross-referencing them by ticker and time window, and telling you the day a third independent signal lands on a stock nobody's talking about yet. That persistent, structured monitoring — not access to the underlying data, which is all public anyway — is what ClusterSignal actually does. The data was never the moat; watching it continuously is.

Why is some data delayed unless I register?

Visitors see the full shape of the product — every page, every ticker, real signals — but with a 30-day delay on current data (the Convergence page, Ticker Story pages, Dashboard, and Signals explorer all apply this), plus 2–3 auto-curated "current" ticker stories rotated weekly as a live preview. Registering (free, no credit card) removes the delay: current data, watchlists, and email alerts. The fence is recency and completeness, not features — nothing is hidden behind a paywall, it just isn't current until you have an account.

What does the API cost?

The REST API (all 8 signal types plus a per-ticker summary endpoint) is a $19/mo add-on for programmatic, rate-limited access to current data — built for the builder/quant persona who wants clean merged JSON rather than a web page. Docs are public at /docs/api regardless of whether you have a key. Everyone else — including all free accounts — gets full current access through the website itself at no cost.

Do signals predict returns?

We don't claim they do — ClusterSignal is detection and context, not prediction. What we do instead is publish an honest backtest at /backtest: historical forward returns (30/60/90 days) for both buy and sell clusters, broken out by grade, with sell-side clusters scored on the opposite polarity (a stock falling after coordinated insider selling counts as the signal being right, not wrong). Sample sizes are shown alongside every stat, including when they're small — a grade-A cluster type with a handful of historical instances is disclosed as exactly that, not dressed up as a robust result.

How often is it updated?

Each of the 8 sources syncs on its own cadence — insider trades every 30 minutes during market hours, most others daily. Signed-in users see everything as soon as it's detected; visitors see the 30-day-delayed view described above.