---
title: "The Cluster Buying Signal in Insider Trading"
description: "How to check an insider cluster buy: a worked counting example, date windows, shared-ownership checks, and the limits of plan flags and past research."
canonical: https://insideralpha.ai/learn/cluster-buying
date_published: 2026-05-12
date_updated: 2026-09-23
publisher: InsiderAlpha (https://insideralpha.ai)
---

Canonical URL: https://insideralpha.ai/learn/cluster-buying
Published 2026-05-12. Updated 2026-09-23.

# Cluster buying: verify the buyers before counting the pattern

Cluster buying usually means two or more insiders buying securities of the same company within a defined period. The SEC does not define the term, so the definition is yours to state. To make the label useful, state which securities qualify, how you identify distinct buyers, whether a plan-marked purchase is included and which dates define the window.

Start with the [SEC's Form 4 instructions](https://www.sec.gov/about/forms/form4.pdf). Code P includes private purchases and purchases of derivatives, so an ordinary-share cluster requires more than counting all P rows. Instruction 4 also allows separate or joint filings by people who beneficially own the same securities. Several documents can describe one economic purchase.

## Write the definition before inspecting a chart

For an ordinary-share study, a reproducible definition might use Table I common-stock P acquisitions by at least two distinct economic buyers, with transaction dates inside a specified rolling interval. That is an example research definition, not a claim about every InsiderAlpha screen or published paper. A researcher who includes preferred stock, indirect fund positions or plan-marked purchases should say so explicitly.

Choose transaction dates when describing when buying occurred. Filing dates instead describe when the information became public. Both are useful, but they answer different questions. The [Form 4 reporting deadline](https://insideralpha.ai/learn/sec-form-4) means transactions can become visible on different days even when they happened together. A historical screen also needs an information cutoff so that later filings are not silently treated as already known.

## Distinct reporting people are not always independent buyers

A fund, its adviser and a controlling individual can each have reporting obligations for a single holding. A person can also report direct shares and shares held through a trust. Compare ownership descriptions, shares, dates, prices and footnotes before treating those rows as separate commitments of capital. Different CIKs identify different reporting entities; they do not by themselves prove independent decisions.

Likewise, multiple executions by one executive do not create multiple insiders. If one purchase is split into three price bands, retain the rows for price reconciliation but count the buyer once. When a Form 4/A arrives, inspect what was corrected before adding another event. Instruction 9 specifically allows amendments containing only added or corrected lines, so a later amendment is not necessarily a complete replacement document.

## A worked counting example

**The following records are hypothetical** and illustrate the method rather than describe a current stock. Assume every P row below concerns the same issuer's common stock, and the footnotes identify the fund-related rows as the same economic purchase.

| Record | What the source shows | Counting treatment |
| --- | --- | --- |
| Director A, Monday | P acquisition of 1,000 shares | One buyer and one purchase |
| Director A, Monday | Form 4/A correcting that purchase's price | Update the price; do not add a buyer |
| Fund B, Tuesday | P acquisition of 5,000 shares | A second economic buyer |
| Manager of Fund B, Tuesday | The same 5,000 shares, disclosed indirectly | Retain the relationship; do not count a third purchase |
| Officer C, Wednesday | A award of 10,000 shares | Compensation acquisition; outside this purchase definition |

Five documents leave two economic buyers under the stated definition. The share total is 6,000 purchased shares, not 21,000 acquired shares or 11,000 purchased shares. If the underlying ownership relationship is unclear, the correct count is unresolved. A polished cluster headline is not a reason to decide that ambiguity in favor of a larger number.

## Window boundaries can change the result

A Friday purchase and a Monday purchase fall into different calendar weeks but can fall into the same rolling seven-day interval. Conversely, two purchases at opposite ends of one week can be far apart for a shorter window. Record whether the window uses calendar days, business days or exact timestamps. A date-only record cannot support a claim that two executions were within a precise number of hours.

Disclosure timing adds a second problem. Suppose the first buyer's filing arrives on Tuesday and the second buyer's on Thursday. The two-buyer pattern was not publicly established on Tuesday merely because both transaction dates later fall inside the same interval. A backtest must use the moment the qualifying information was available. Otherwise it gives the strategy information that a reader of EDGAR could not yet have had.

## Compensation and plan context belong beside the count

Annual grants, vesting and option exercises can put many executives on the same filing date. They are not ordinary-share purchases simply because holdings increased. Use the [code guide](https://insideralpha.ai/learn/form-4-transaction-codes) to distinguish A, M and F from P, and read the [two-table comparison](https://insideralpha.ai/learn/open-market-vs-derivative-transactions) before aggregating securities.

A disclosed [Rule 10b5-1 arrangement](https://insideralpha.ai/learn/rule-10b5-1) adds timing context. You can separately report all qualifying purchases and the subset with disclosed plan evidence, as long as the definitions are explicit. An unchecked or missing marker does not prove spontaneous buying. Excluding disclosed plans also does not make every remaining buyer independent or establish a shared view of the business.

## What a cluster does not prove

Several people buying close together does not establish that they possess the same nonpublic information. They may face a common post-earnings trading window, participate in the same disclosed capital raise, follow ownership requirements, or reach different conclusions from public information. These are questions to investigate, not motives to assign automatically.

Compare the issuer's recent public reports, the transaction footnotes and any relevant ownership policy. Size each buyer's purchase on a consistent holdings basis. A director's salary, an executive's reported total compensation and a fund's assets are different denominators. If a reliable comparison is unavailable, show the purchase amount and identify the missing context rather than calling it a token or high-conviction trade.

## Cluster buying and "opportunistic" trading are different definitions

[Cohen, Malloy and Pomorski's 2012 paper, Decoding Inside Information](https://doi.org/10.1111/j.1540-6261.2012.01740.x), distinguishes routine from opportunistic traders using historical trading patterns. That classification should not be relabeled as a test of every two-buyer cluster. Nor is it the same as the modern Form 4 plan checkbox. A published portfolio return is not an expected annual return for a current cluster.

When reviewing any cluster study, check its actual sample years, security filters, buyer definition, holding period and return benchmark. Ask whether amendments were handled, whether execution costs were included, and whether the selection used only information available at the time. Those details decide whether two results are comparable; the word "cluster" in both titles does not.

## Historical coverage is not a cluster-frequency measurement

Our [published Form 4 study](https://insideralpha.ai/learn/form-4-by-the-numbers) describes a frozen corpus covering 6,114 companies as of August 3, 2026. That coverage total is not the number of companies with qualifying clusters, and a filing count is not a distinct-purchase count. Measuring cluster frequency requires running a stated definition against that snapshot.

A useful comparison keeps the definition fixed across periods and shows both the number of qualifying companies and the number eligible to appear. A broader universe or longer observation window can produce more clusters without any change in underlying behavior. Avoid comparing this week's live screen with a historical total that uses different coverage or dates.

## Build a company-level comparison you can update

1. Write the interval and information cutoff, then identify the issuer by CIK.
2. List each candidate purchase with its accession, security, price, amount and ownership basis.
3. Resolve shared holdings, multiple execution rows and amendments before counting people or dollars.
4. Keep disclosed plans and unavailable plan evidence visible as separate observations.
5. Compare earlier purchases by the same people and the company's public disclosures during the interval.
6. Save unanswered questions so a later filing can revise the conclusion without rewriting the underlying facts.

Start with [the cluster-buying screen](https://insideralpha.ai/insiders/clusters), then open the company's original filings to apply these checks. Use [weekly purchase totals](https://insideralpha.ai/top-insider-buys-this-week) to compare amounts and [recent purchases](https://insideralpha.ai/insiders/recent-buys) to inspect the disclosure sequence. For the different problems involved in interpreting several sellers, continue with [the selling guide](https://insideralpha.ai/learn/insider-selling-signal-or-noise).

*This article is informational and is not investment advice.*