What is Cluster Buying? How Multiple Insiders Buying Together Signals Conviction

A single insider buying their company's stock is interesting. Two or three buying within the same fortnight — independently, at similar prices — is a qualitatively different kind of signal. Here is what the research says and how to distinguish genuine cluster buys from noise.

The definition

Cluster buying refers to two or more distinct insiders at the same company executing purchase transactions within a compressed time window — typically 14 calendar days. The insiders need not coordinate: they may be unaware of each other's purchases at the time of filing. What makes the signal significant is precisely that independent agents with shared information access are reaching the same conclusion simultaneously.

This is different from a single insider making multiple purchases (that is repetitive buying) and different from a single large transaction. The cluster signal requires separate individuals — a CEO and a board member, or three non-executive directors — each independently deciding to deploy personal capital at roughly the same time.

Why it works: the coordination-without-communication problem

The most common explanation for why cluster buying predicts returns is straightforward: multiple insiders have each independently assessed that the stock is undervalued at the current price, and each chosen to act on that view with personal money.

The more nuanced version involves what you might call informal coordination without illegal communication. Directors and executives are in the same board meetings, see the same management presentations, and share the same reading of the business outlook. They are not permitted to share material non-public information with each other for trading purposes — but they can all reach the same legal inference from publicly available information that the stock is cheap relative to what they know about the company's direction.

The result is a natural experiment: when multiple informed agents with aligned (but legally obtained) information all make the same independent bet, the probability that they are all wrong simultaneously is meaningfully lower than for any single agent acting alone.

What the research shows

Academic literature on insider trading returns has consistently found that portfolio strategies tracking insider purchases generate abnormal returns, and that cluster purchases outperform single-insider purchases.

Jeng, Metrick & Zeckhauser (2003) — one of the most cited papers on the topic — constructed monthly-rebalanced portfolios of insider purchases and found they generated 11.2% abnormal annual returns using a four-factor model. While their paper focused on aggregate purchases rather than cluster specifically, subsequent replication studies and practitioner analyses have found that filtering for cluster buys substantially improves the signal quality.

Lakonishok & Lee (2001) found that purchase signals are strongest when multiple insiders are buying simultaneously and weakest (near zero predictive value) when it is a single insider with no corroborating activity.

A worked example

This is the kind of pattern that triggers the cluster signal. The companies and numbers below are illustrative of the type of filing sequence InsidersAlpha detects weekly across European markets.

Each of these three individuals filed separately with their national regulator. Each filing appeared in the InsidersAlpha database as it was processed. When the third filing came in, the cluster badge activated on all three transactions retrospectively.

The key observation: none of these amounts is individually remarkable for a board-level executive. But three board members buying in the same fortnight, when the stock has been flat for six weeks, is a concrete expression of collective conviction that the market is mispricing the company.

When the cluster signal is stronger

Not all cluster buys are equal. These conditions improve the signal quality:

• Higher-seniority insiders. A CEO + CFO cluster carries more weight than two non-executive directors, because the executives have deeper and more current information about business performance.
• Larger purchases relative to their compensation. An executive buying €200,000 worth of stock — roughly their annual bonus — is expressing significantly more conviction than a €5,000 symbolic purchase.
• Purchase following a drawdown. A cluster buy after a 15% stock decline is more informative than one at a 52-week high. InsidersAlpha shows the price dip badge alongside cluster when both conditions apply simultaneously.
• Tight time window. Three insiders buying on days 1, 3, and 5 is stronger than the same three buying on days 1, 8, and 13, simply because the convergence is tighter and less likely to be coincidental.
• No obvious mechanical explanation. Purchases made in a defined trading window that opens after earnings are expected and less informative than purchases made outside any scheduled window.

What cluster buying is not

Two caveats worth keeping in mind.

First, cluster buying does not predict the reason for conviction — only its existence. Three insiders buying does not tell you whether they believe revenues will beat expectations, a strategic deal is imminent, or simply that the sector sell-off has been overdone. The signal flags that informed people think the stock is cheap; the investor still needs to do their own fundamental work on why.

Second, cluster buying during a management buyout (MBO) process can produce a false positive. In an MBO, insiders accumulate shares as part of a coordinated private transaction, not as an expression of public market conviction. This is rarer in European listed companies than in private equity situations, but worth checking if a large cluster appears alongside elevated M&A speculation in the news.

How InsidersAlpha detects cluster buying

InsidersAlpha processes every new filing as it appears in the relevant regulatory database. When a new buy transaction is added, the system checks whether any other buy transactions from different named individuals at the same company appear within the preceding or following 14 days. If they do, all qualifying transactions receive the cluster badge.

The check runs on transaction_date (the date the trade executed, not the filing date). This matters because insiders have three business days to file — a transaction executed on a Monday might not appear in the database until Thursday. Using execution date rather than filing date gives a more accurate picture of when the buying actually happened.

The full technical methodology, including how we handle corporate entity transactions and the scoring model, is on the methodology page.