The Cluster Buying Signal

A cluster buy is when two or more corporate insiders at the same company purchase shares on the open market within a short window - typically 48 hours to 30 days. Cluster buying is one of the strongest and most consistently documented anomalies in insider-trading research.

Why it works

A single insider purchase can be motivated by personal liquidity, a new-job share requirement, or simple optimism. But when several independent insiders - say a CEO, a CFO, and an outside director - all buy at roughly the same time, coincidence becomes an unconvincing explanation. The far more likely driver is that they are each responding to the same shared, asymmetric information about the company's near-term prospects: a turnaround taking hold, a contract about to close, or a valuation they collectively see as too cheap.

Empirical evidence

Cohen, Malloy & Pomorski (2012, "Decoding Inside Information") separated "opportunistic" insiders from "routine" ones and found opportunistic clustered buys generated value-weighted abnormal returns of roughly 8% per year. Lakonishok & Lee (2001) likewise found the predictive power of insider trading concentrates in active, multi-insider buying rather than isolated trades. Practitioner studies since have repeatedly confirmed that clusters outperform lone purchases.

How often do clusters actually happen?

Genuine clusters are uncommon, which is part of their value. To measure frequency we count a cluster week: a calendar week in which two or more distinct insiders at the same company report open-market purchases on Form 4. By that definition, InsiderAlpha's data contains 1,389 cluster weeks across 837 companies in the 12 months through late July 2026, out of the 6,114 companies in our August 3, 2026 corpus snapshot. That averages out to a few dozen companies per week showing multi-insider buying.

Two caveats. The weekly definition is an approximation: it can split one buying wave across two weeks or merge two nearby waves into one. And our live scoring uses a tighter 48-hour window than this measurement, so the set of clusters that actually earn the scoring bonus is smaller than the weekly count. Directionally, though, the message is clear: on any given week you have dozens of cluster candidates, not thousands, and quality filters decide which of them deserve attention.

Clusters form in slow motion

A cluster is rarely visible all at once. In our 2026 data the median open-market purchase reaches EDGAR 2.8 calendar days after the trade, and different insiders file at different speeds. So the typical sequence is: one buy appears, then a second filing from a different insider lands a day or two later and retroactively turns the first buy into a cluster. The signal strengthens after the fact. Practically, this means a single strong purchase at a company is worth putting on a watchlist precisely because the next filing may complete the pattern, and platforms that re-score on every new filing will catch that upgrade the moment it happens.

• Open-market purchases (code P) - not awards or option exercises.
• Distinct insiders - two trades by the same person don't make a cluster.
• A tight window - the closer together, the stronger; InsiderAlpha keys on a 48-hour core window.
• Meaningful size relative to each insider's pay and existing stake.

False clusters to avoid

Beware coordinated compensation events that look like clusters but aren't: a board-wide annual stock award, or several executives exercising options on the same vesting date. Pre-scheduled Rule 10b5-1 purchases that happen to overlap are also not genuine conviction clusters. We screen all of these out.

A worked example (hypothetical)

Consider a mid-cap industrial company whose stock has fallen 30% over six months. Two days after an earnings report, the CFO buys $250,000 of stock on the open market, code P, no 10b5-1 flag. Within the same week, two outside directors each add roughly $100,000 at similar prices. Three distinct insiders, discretionary timing, meaningful size against director-level pay, in a stock trading over a million shares a day. That is the archetypal cluster: the people closest to the numbers acted on the same information at the same moment, with their own cash.

Now change one detail. Suppose the same three trades were code A grants dated to the annual board meeting, or all three carried a 10b5-1 plan footnote. The surface pattern is identical, three insiders, one week, but the information content collapses to zero. The cluster label belongs to the first case only. This is why filtering on transaction code and plan flags comes before any excitement about the headline.

What to check before acting on a cluster

1. Codes: every leg should be an open-market purchase (code P). One real buy plus two grants is not a cluster.
2. Plan flags: discard legs carrying a Rule 10b5-1 marker.
3. Distinct people: confirm the buyers are separate insiders, not one person reporting across direct and indirect accounts.
4. Seniority mix: a cluster that includes the CEO or CFO is stronger than one made of peripheral insiders.
5. Size: each leg should be meaningful against that insider's pay. Three token $5,000 buys can be a coordinated show of confidence rather than real conviction.
6. Liquidity and freshness: the stock must be tradeable and the filings recent. A cluster from three weeks ago is history, not a signal.

Does cluster selling work in reverse?

Much less well. Selling has many innocent motives: diversification, taxes, expiring options, estate planning. And in the first quarter of 2026 roughly half of the insider sale transactions we tracked carried a 10b5-1 plan marker, meaning a calendar chose the date. Multi-insider selling is therefore both more common and less informative than multi-insider buying. It earns a look when it is discretionary, unusually large, and out of character for the insiders involved, but the asymmetry is real: clusters matter far more on the buy side. See insider selling: signal or noise for the full treatment.

Frequently asked questions

How many insiders make a cluster?

Two is the minimum and already meaningful; three or more distinct buyers is rarer and stronger. What matters is independence: two co-trustees reporting the same household purchase do not count as two buyers.

Does the window length matter?

Yes. The tighter the window, the more likely the buys share a single cause. InsiderAlpha's scoring keys on 48 hours; research definitions run out to a month. Anything beyond that stops being a cluster and becomes a trend.

Do clusters work in every kind of stock?

The evidence is strongest where the insiders' information advantage is largest: smaller companies, thinner analyst coverage, and businesses with lumpy, contract-driven revenue. In heavily covered mega-caps, insider buying is rarer and filings are priced within minutes, so the residual edge is smaller. Liquidity still sets the floor either way: a cluster in a stock you cannot exit is a curiosity, not a trade.

Are cluster buys a guarantee?

No. Insiders are early, wrong, or both often enough that position sizing and risk management still decide the outcome. Clusters shift the odds; they do not remove the risk.

How InsiderAlpha scores cluster buys

Our trading-plan generator adds a +15-point cluster bonus when two or more insiders purchase the same ticker within 48 hours, layered on top of role-weighted size and recency scoring. You can watch clusters form in real time on the live signal page.

See live cluster-buy signals → · Top insider buys this week →

This article is informational and is not investment advice.