Cluster insider buying

The pattern

A cluster signal is what you have when three or more separate insiders at the same company file open-market buys (Code P) within a 60-day window. Catalyst tracks clusters on the buy side only — Code-S sells are noisier (taxes, 10b5-1 plans, diversification) and don't carry the same coordinated-conviction interpretation. The classic academic work on cluster buying dates back to Seyhun (1986) and has been replicated many times since: cluster buys outperform single-insider buys, and they outperform the broader market on a 30- to 90-day horizon.

Catalyst's threshold is three buyers because that's where the historical edge stops looking like noise. Two insiders buying could be a coincidence. Three is coordination of belief — they're all seeing the same thing, at the same time, and acting on it with their own money.

What makes a cluster strong

• Officer-heavy clusters. A cluster where two or three of the buyers are C-suite (especially CFO) carries more weight than one made up of directors.
• Compressed timing. Three buys in two weeks is louder than three buys spread across 55 days. Tight timing usually means the insiders are reacting to the same internal data point or board meeting.
• Substantial dollar size. Clusters where each buyer puts in real money (relative to their own net worth) are signal. Clusters where everyone bought the $5,000 minimum to satisfy a board-ownership rule are noise.
• Price near the 90-day low. Insiders buying near the bottom of a recent range is a strong setup. Insiders chasing momentum near a 90-day high is much weaker.

Common failure modes

• Forced ownership thresholds. Some boards have minimum-share-ownership requirements for new directors. New-director clusters in the first 90 days post-IPO or post-appointment often have no information content.
• 10b5-1 plan buys. Rare, but they happen. Always check the footnote.
• Mega-cap clusters. A $250K cluster buy in a $500B mega-cap means very little compared to the same cluster at a $500M mid-cap. Size relative to market cap is what makes the signal usable.

How Catalyst handles clusters

Whenever a new Code-P filing lands, we look back 60 days for other open-market buys on the same ticker. If the cluster crosses the three-buyer threshold, the trade is routed to deeper AI analysis (Sonnet) instead of the default Haiku tier. The cluster member count is shown alongside the signal on the homepage and the stock detail page.

If the cluster grows after the first analysis — a fourth buyer joins, or the same buyers buy more — we re-run the analysis automatically and flag the change. See the methodology page for the full pipeline.

Curious which tickers currently have an active cluster? The homepage surfaces them with a cluster badge.