Cluster Buying: Why Multiple Insiders Buying Together Is the Strongest Signal

The Cluster Signal: When Insiders Agree

Corporate insiders—officers, directors, and 10% beneficial owners—possess the most granular, real-time knowledge of their companies' prospects. They see quarterly results weeks before public announcements, participate in strategic planning sessions, and understand product pipeline strength, competitive positioning, and margin dynamics at a level no external analyst can match. But individual insider purchases, while studied as positive signals, carry noise: personal liquidity needs, diversification goals, tax planning, and estate management often drive transactions that have nothing to do with company outlook.

Cluster buying is designed to filter out most of this noise. When three, four, or five insiders independently decide to purchase shares within the same narrow time window—typically 30 days or less—it becomes far less plausible that all of them are motivated by personal financial planning rather than a shared read on company prospects. Each additional insider is an independent data point, and the cluster pattern is best understood as consensus behavior among the people with the best information—not as a promise of any particular return.

Consider the behavioral logic: if you're the CFO and you know the upcoming quarter will miss estimates, you're unlikely to be buying shares—even if your personal financial plan says "diversify into your own stock." If the CEO, CFO, COO, and two board members are all buying in the same month, they're not coordinating on personal tax strategy. The most natural reading is shared conviction that the stock is undervalued relative to what they know.

The Academic Foundation

Lakonishok & Lee (2001) — The Benchmark Study

Josef Lakonishok and Inmoo Lee's paper in the Review of Financial Studies, "Are Insider Trades Informative?" (2001), analyzed insider trades across NYSE, AMEX, and Nasdaq firms from 1975–1995 and established a widely-cited framework for the topic. Its findings, stated as the paper reports them:

• Insider purchases carry predictive information; insider selling does not. The informativeness of insider activity comes almost entirely from buying.
• The effect is concentrated in smaller firms, where information asymmetry is highest—stocks bought by insiders in small-cap firms earned roughly 7.4% abnormal returns over the following twelve months (about 4.8% after adjusting for size and book-to-market).
• The predictive content of purchases increases when multiple insiders buy, which is the empirical basis for treating clusters as a stronger signal than solitary buys.

Note what the paper does not establish: it does not publish a specific "cluster of N insiders → X% return" table, and it does not rank C-suite versus board buyers by outperformance. Treat cluster strength as directional, not as a quantified forecast.

Jeng, Metrick & Zeckhauser (2003) — Sizing the Insider Edge

This study in The Review of Economics and Statistics built value-weighted "purchase" and "sale" portfolios of insider trades (1975–1996) and used performance-evaluation methods rather than event studies. Its headline findings:

• The purchase portfolio earned abnormal returns of more than 50 basis points per month (on the order of 6% per year); the sale portfolio earned no abnormal returns.
• Roughly one-quarter of the purchase abnormal return accrued within the first five days after the trade, and about one-half within the first month—so the informational value decays quickly, which is why recency matters when reading a signal.
• Importantly, the authors found that abnormal returns to insider trades in small firms were not significantly different from those in large firms, and that top executives did not earn higher abnormal returns than other insiders. Any C-suite or small-cap emphasis in practice is a heuristic, not a result this paper endorses.

Seyhun (1986) — Early Foundational Work

Nejat Seyhun's early research in the Journal of Financial Economics helped establish two durable regularities:

• Buying is more informative than selling—insiders sell for many non-informational reasons (taxes, diversification, liquidity), so sales are a weaker signal.
• Insider trading tends to be more informative in smaller firms, where the information gap between insiders and the public is wider.

Implication for signal design: when filtering for clusters, prioritize buying over selling and open market purchases over scheduled or mechanical transactions, and treat insider sales as weak/noisy signals unless they are unusually large or clustered.

Anatomy of a High-Conviction Cluster

Illustrative Pattern: The "C-Suite Consensus Buy"

Microsoft (MSFT) — February 2026 Example

• February 18, 2026: Director John W. Stanton purchased $1,986,750 in open market transactions
• This is a single large insider buy, not yet a cluster, but the size ($2M) and role (director with deep industry knowledge) make it notable
• Meridian's insider module surfaces it as a high-magnitude, recent open-market purchase—a candidate to watch, not a cluster

What would make this a cluster?
If within the next 30 days (by March 18, 2026):

• The CFO purchases $500K+ in shares
• The COO purchases $300K+ in shares
• Another board member purchases $200K+ in shares

That would be four unique insiders inside one month — a cluster by the literature's working definition, and a materially different read from the single purchase you started with. Four people acting separately is much harder to explain as personal financial planning than one.

Whether and how any given platform surfaces such a pattern is a product question, not a property of the filings. What the filings themselves support is the observation above: count the insiders, check the window, and read the size against the company rather than in absolute dollars.

Cluster Typology: Four Patterns

Type 1: C-Suite Consensus (Strongest Read)
CEO + CFO + COO buying within 30 days

• Why watched: all three see the full operating picture, so agreement is hard to dismiss as personal planning
• Typical scenario: upcoming product launch, major contract win, turnaround inflection point
• Example: Apple insiders buying during 2019 (ahead of the iPhone 11 cycle)

Type 2: Board Cluster (Strong Read)
3+ independent directors buying within 30 days

• Why watched: directors carry a fiduciary duty and rarely buy in size without conviction
• Typical scenario: stock trading below the board's view of intrinsic value, upcoming strategic announcement
• Example: Berkshire Hathaway directors buying during the March 2020 COVID crash

Type 3: Mixed Officer/Director Cluster (Moderate Read)
1-2 C-suite + 2-3 directors buying within 30 days

• Why watched: the most common cluster type, blending operating and oversight perspectives
• Typical scenario: general undervaluation, positive earnings outlook
• Example: Regional banks during the 2023 rate cycle

Type 4: Founder/Large Shareholder Cluster (Contextual Read)
Founder + family/affiliated entities buying

• Why watched, with caution: interpretation depends on whether buying is strategic accumulation or price support
• Typical scenario: stock under short attack, company defending valuation
• Example: Elon Musk + affiliated entities buying Tesla during the 2022 selloff

The four types are ordered by how cleanly the pattern removes noise—not by any expected return.

When Clusters Draw the Most Attention

Scenario 1: Market Stress / High VIX Environments

Insider cluster buying during broad market fear is watched especially closely, because it separates insiders' read of the business from the market's repricing of everything:

Why it draws attention:
During panic selling, insiders who buy are acting as contrarians—the implication being that the business fundamentals have not changed even as the market marks the stock down. This is the "Buffett principle" in spirit: "Be greedy when others are fearful." Clusters that formed during the March 2020 COVID crash, the December 2018 selloff, and the October 2022 inflation peak are frequently cited as examples of insiders buying into fear; the pattern is a behavioral tell, not a guaranteed rebound.

Scenario 2: Post-Earnings Selloffs (Overshooting Reactions)

When a stock drops sharply on earnings despite otherwise solid fundamentals, insider cluster buying in the following 2–4 weeks is a notable "insiders think the market overreacted" signal:

Pattern:

• Earnings miss or guidance reduction → stock sells off hard
• 2 weeks later: 3+ insiders buying at the depressed price
• Interpretation: insiders judge the selloff excessive relative to business reality

This does not tell you the stock will recover—only that the people closest to the business chose to add at lower prices.

Scenario 3: Small-Cap Value Stocks

Cluster buying is often considered more meaningful in small/mid-cap stocks ($500M–$5B market cap) because:

• Analyst coverage is sparse (less external information competition)
• Institutional ownership is lower (more room for accumulation)
• The information gap between insiders and the public is wider

This is consistent with the small-firm concentration Lakonishok & Lee documented for insider purchases. (Jeng, Metrick & Zeckhauser, using a different portfolio method, did not find a significant small-vs-large gap—so treat the small-cap emphasis as a reasonable heuristic on which the evidence is mixed, not a settled fact.)

Scenario 4: Multi-Source Confluence (The Meridian Edge)

The signal is strongest as corroboration when insider cluster buying coincides with:

• Congressional buying in the same stock
• Institutional 13F adds
• Dark pool accumulation
• A short interest spike (potential squeeze)

...because several independent groups pointing at the same ticker is harder to explain away than any single source. Meridian expresses this as a higher composite Smart Money Score, not as a return forecast.

Current Example: MSFT's insider activity combines with:

• Congressional buy (Sen. Mullin)
• Institutional holds (Citadel, ~$6.9B position)
• Superinvestor presence (16 funds)
• A short squeeze setup

→ These sources roll up into Meridian's composite Smart Money Score for the ticker (a multi-source confluence reading).

If MSFT insiders form a cluster in the next 30 days, this would add a further independent source to the confluence.

Red Flags: When Clusters Fail

False Signal 1: Scheduled Purchases (10b5-1 Plans)

Many insiders establish Rule 10b5-1 trading plans that execute purchases automatically on preset dates. These are not informational—they're mechanical.

How to detect:

• Check SEC Form 4 footnotes for "10b5-1" language
• Look for perfectly regular purchase patterns (same date every month/quarter)
• Quiver Quantitative and other platforms flag these automatically

Meridian filtering: Excludes 10b5-1 purchases from conviction scoring.

False Signal 2: Options Exercises (Non-Informative)

Insiders often exercise vested stock options and immediately sell shares to cover taxes. The Form 4 shows a "purchase" (exercise) but it's not a conviction signal.

How to detect:

• Transaction code "M" = options exercise
• Transaction code "P" = open market purchase (this is what you want)
• Check if exercise is paired with immediate sale ("cashless exercise")

Meridian filtering: Only counts code "P" open market purchases in cluster detection.

False Signal 3: Defensive/Support Buying

When a stock is under short attack or facing activist pressure, insiders sometimes buy to "support the stock" or "show confidence" rather than because they have new positive information.

How to detect:

• Check for recent short seller reports (Hindenburg, Muddy Waters, Citron)
• Look for unusual media coverage or activist campaigns
• Cluster happens immediately after a public attack (within 1-3 days)

Interpretation: Treat as neutral rather than bullish—insiders are defending, not initiating based on new data.

False Signal 4: Illiquid/Thinly Traded Stocks

In micro-caps with <$100M market cap and <100K daily volume, insider buying can move the price mechanically without reflecting genuine information:

Filter rule: before reading a cluster at all, check that the stock is liquid enough for the price to mean anything — a reasonable market capitalisation, daily volume deep enough that a single purchase cannot move it, and a spread tight enough that the quoted price is real. Where you draw those lines is a judgement about your own risk tolerance, not a property of the signal.

Practical Implementation: Cluster Detection Algorithm

How to Read a Cluster

The steps below are the ones the research supports, and they are the ones you can apply to any Form 4 feed yourself. Meridian's own implementation, including how it classifies and combines what it reads, is not published.

Start from open-market purchases only

• Form 4 transaction code "P" is the open-market purchase
• Set aside 10b5-1 plan transactions: they were scheduled in advance, so they carry no timing information
• Set aside options exercises (code "M") and gifts (code "G") for the same reason

Count insiders, not transactions

• A CEO buying three times in a week is one insider, not three
• Group by ticker over a window rather than by filing date, since Form 4 has a two-day reporting lag
• The literature's working definition of a cluster is at least three unique insiders

Read seniority as context, not as arithmetic
Lakonishok & Lee (2001) found the predictive content of insider trades sits with the aggregate of what insiders do, and Jeng, Metrick & Zeckhauser (2003) found top executives did not earn higher abnormal returns than other insiders. So treat a C-suite-heavy cluster as more legible, not as automatically stronger — the evidence for a seniority premium is weaker than practice usually assumes.

Recency matters most
Both papers point the same way: most of the informational value of an insider purchase shows up within the first weeks. A cluster that closed a month ago is telling you about a month ago.

A cluster is one source among several
An insider cluster is most interesting when it lines up with independent sources — congressional disclosure, institutional accumulation, superinvestor presence. Meridian expresses that agreement as a single reading on its own scale. It describes how strongly a pattern registers, not an expected return, and how it is assembled is not published.

API Usage Example

# Recent US insider transactions, newest first
curl "https://meridianfin.io/api/us/insiders?days=30"

Response (abridged; the route returns detected clusters alongside the trades):

{
  "data": [
    {"ticker": "EXAMPLE", "company": "Example Corp", "insider_name": "John Doe",
     "title": "CEO", "transaction_type": "Buy", "shares": 25000, "value": 1500000,
     "price_per_share": 60.0, "date": "2026-02-15", "filing_date": "2026-02-17",
     "is_cluster": true, "cluster_count": 4}
  ],
  "clusters": [
    {"ticker": "EXAMPLE", "trade_count": 4, "first_date": "2026-02-12",
     "last_date": "2026-02-20", "titles": ["CEO", "CFO", "COO", "Director"]}
  ],
  "metadata": {"total": 128, "filtered": 4, "buy_count": 4, "sell_count": 0,
               "cluster_count": 1, "has_more": false, "source": "openinsider"}
}

Key Takeaways

• Insider cluster buying (3+ insiders within 30 days) is a cleaner signal than a single insider purchase because it is far less likely to be driven by any one person's personal financial planning
• Academic evidence is strongest for a simple asymmetry: insider buying is informative and selling is not (Lakonishok & Lee 2001; Jeng, Metrick & Zeckhauser 2003; Seyhun 1986)
• The informational value of an insider purchase decays quickly—much of it within the first weeks—so recency matters when reading a cluster
• Watch context: clusters during market stress, after overreaction selloffs, or alongside other smart money sources are the ones practitioners weight most, though none of these guarantees a return
• Avoid false signals: exclude 10b5-1 scheduled purchases, options exercises, and defensive buying after short attacks
• Multi-source confluence (insider clusters + congressional/institutional buying) is valued as independent corroboration, not as a higher promised return

Further Reading

Academic Research:

• Lakonishok, J., & Lee, I. (2001). "Are Insider Trades Informative?" Review of Financial Studies, 14(1), 79-111.
• Jeng, L.A., Metrick, A., & Zeckhauser, R. (2003). "Estimating the Returns to Insider Trading: A Performance-Evaluation Perspective." Review of Economics and Statistics, 85(2), 453-471.
• Seyhun, H.N. (1986). "Insiders' Profits, Costs of Trading, and Market Efficiency." Journal of Financial Economics, 16(2), 189-212.

Related Meridian Articles:

• Insider Buying Signals — Foundational overview of corporate insider trading patterns
• Signal Confluence Explained — How insider clusters combine with other smart money sources