👁️🔥 Insider Cluster Buy Detector — 3+ Insiders Buying the Same Stock in 30 Days

The hedge-fund-grade insider-signal actor. Returns CLUSTERS of insider buying — events where 3+ distinct insiders (CEO, CFO, Directors, 10%-owners) buy the same stock inside a rolling 30/60/90-day window. Cluster signals are the highest-conviction insider trades on the planet — Lakonishok-Lee (2001) and Cohen-Malloy-Pomorski (2012) both documented that clusters of 3+ insiders outperform single-insider signals by 4-7% annually.

This is not another raw Form 4 firehose. Single insider buys are noisy — Directors get gifted shares, exercise options, accept comp grants, rebalance for divorce. A single CFO buy can be a personal-finance event. But when 3+ executives independently buy in the same 30-day window? That is a coordinated signal — they all see the same upside-asymmetry in their internal numbers and they are voting with their personal capital.

What you get (per cluster record)

Every dataset row is a fully-aggregated cluster — one row per stock, NOT one row per insider:

Field	Meaning
symbol, company_name, sector	Ticker, issuer name, OpenInsider industry classification
cluster_start_date / cluster_end_date	The first and last insider-buy trade dates inside the window
insider_count	Distinct insiders (≥ min_cluster_size)
insider_names	All insider names — list, deduped
insider_roles	All roles — list (CEO, CFO, Director, COO, 10% Owner, etc.)
total_shares_bought	Sum of shares across the cluster
total_value_usd	Sum of trade values in USD
current_stock_price	Last reference price (most recent cluster trade price)
stock_move_since_cluster_start_pct	Spot vs first-cluster-trade reference, percent
is_all_buy	true iff every trade in window is a P-Purchase, no sells/exercises mixed
cluster_strength_score	Composite — insider_count × role_weight_sum × log(total_value_usd). Higher = stronger conviction. CEO=5, CFO=4, COO/Pres=3, Director=2, 10%-Owner=4, Other=1.
top_insider	Highest-role buyer in the cluster (the "anchor")
data_source	openinsider.com cluster-buys + per-symbol drilldown

Why clusters > single insider signals

Academic finance has been on this for two decades:

• Lakonishok & Lee (2001) — Are Insider Trades Informative? Found cluster buys outperform by 4.8% in the year following.
• Cohen, Malloy & Pomorski (2012) — Decoding Inside Information. Distinguished "opportunistic" trades (cluster signals) from "routine" ones (calendar buys). Opportunistic outperformed the market by 8.2% annually.
• Jeng, Metrick & Zeckhauser (2003) — Documented that single Director buys carry almost no predictive content. The signal lives in the cluster.

Translation: a single 10K-share CEO buy is noise. Three insiders piling in over 30 days is alpha.

Inputs

• min_cluster_size — default 3 (academic threshold). 2 catches CEO+CFO duos. 5+ catches mega-clusters (whole-board buys).
• date_range — last_30d (canonical), last_60d (quarterly cycle), last_90d (cross-quarter accumulation).
• min_value_usd — default 25000. Filters out trivial gifts/exercises.
• exclude_sells — default true. Pure-buy clusters only — historically outperform mixed clusters by 2-3%.
• tickers — optional watchlist filter. Leave empty for full-universe scan.
• limit — max cluster records (default 25; tune by usage).
• include_industry — adds sector classification to each cluster.

Data source & method

Primary: OpenInsider (openinsider.com/latest-cluster-buys) — already pre-aggregates Form 4 filings into cluster format with the Ins (insider count) column. We pull the cluster page, drill into each stock via the per-symbol screener (/screener?s=TICKER&fd=N), extract every insider's name + role + trade date + share count + price, then build the composite cluster record.

Validation: SEC EDGAR Form 4 full-text (efts.sec.gov/LATEST/search-index?forms=4) — used as a cross-check when OpenInsider cluster data is stale or a ticker isn't covered. EDGAR is the authoritative source; OpenInsider is the aggregator on top.

Anti-bot risk: LOW on both — OpenInsider serves static HTML and welcomes scrapers. SEC EDGAR explicitly publishes scraping guidance and only asks for a descriptive User-Agent.

Cluster window: rolling — cluster_end_date - cluster_start_date ≤ date_range. If insiders bought on 5 separate days inside the window, the cluster insider_count = 5. We dedupe by insider name to avoid double-counting the same person twice in the window.

Role weights (for cluster_strength_score): CEO=5, CFO=4, COO/President=3, 10%-Owner=4, Director=2, Other officer=1. Scaled by log10(1 + total_value_usd) so a $5M cluster scores higher than a $50K one even at the same insider count.

Comparison vs the legacy stack

	Bloomberg Terminal	TipRanks Premium	OpenInsider Pro	Finviz Elite	This actor
Cost	$25K+/yr	$35/mo	$30/mo	$25/mo	$0.01 + $0.25/cluster
Cluster detection	✓ (function NIM)	Partial — single insiders only	✓ (manual screen)	Partial — calendar-only	✓ (3+ insider window, role-weighted)
Cluster strength score	✗	✗	✗	✗	✓ (composite, role × value)
Pure-buy filter (is_all_buy)	Manual	✗	Manual	Manual	✓ (one toggle)
Top-insider anchor	Manual	✗	Manual	✗	✓
API / programmatic access	Bloomberg API (locked)	Limited	✗	✗	✓ (Apify dataset/JSON/CSV)
Pay-as-you-go	✗	✗	✗	✗	✓
Stock move since cluster start	Manual	✗	✗	✗	✓
Sector classification	✓	✓	✓	✓	✓
No login / no monthly minimum	✗	✗	✗	✗	✓

If you are running an insider-signal portfolio at a $5-20M hedge fund or a quant retail-research shop, your monthly cluster-signal needs are 10-50 records — that is $2.50-$12.50 here vs $35-$300 elsewhere, with full programmatic access and a much cleaner schema.

When to fire this actor

• Weekly Sunday refresh — pull 25 clusters, sort by cluster_strength_score desc, work the top 5 as research priorities for Monday.
• Earnings-week scan — bump min_cluster_size to 4-5, scan last_60d to catch pre-print accumulation.
• Drawdown watch — when a name you own drops 20%+, fire this actor with the ticker as a one-element filter and min_cluster_size=2 to see if any insiders are stepping in.
• Sector wave detection — pull last_30d at min_cluster_size=3, group by sector — multiple clusters in the same sector inside a month is the highest-conviction sector signal in finance (regional banks 2023, biotech 2024).

Cross-link — sister actors in the NexGenData fleet

This actor is one of the cluster-signal layer in a 7-actor smart-money intelligence suite. Pair with:

• sec-form4-insider-tracker — the underlying Form 4 firehose. Every individual transaction, no cluster aggregation. Use when you want the raw stream; use THIS actor when you want the signal.
• 13f-holdings-delta-tracker — quarter-over-quarter institutional position changes (NEW / INCREASED / DECREASED / EXITED across Berkshire, Tiger Global, Bridgewater, Renaissance, Citadel). Pair cluster insider signals with institutional flows for double-confirmation alpha.
• sec-form-13f-holdings-tracker — full 13F portfolio snapshots (vs the delta tracker which only emits changes). Use when you need the absolute holdings list.
• short-interest-tracker — FINRA short interest bi-weekly. Cluster insider buys with high short interest = short-squeeze setup (the classic GME/AMC pattern).
• analyst-price-targets — Wall Street consensus. Use to spot divergence: insiders piling in but Street has Sell ratings = highest-conviction asymmetric setup.
• finance-mcp-server — Claude/MCP integration layer that orchestrates this actor and the other six for natural-language hedge-fund queries.

Pricing

• $0.01 actor start fee
• $0.25 per cluster record returned

A typical scan that yields 5 clusters costs $1.26. A heavy backtester pulling 200 clusters costs $50.01.

This is premium-tier PPE pricing — the cluster signal is hedge-fund-grade alpha and the data carries direct trading value. Compare to $30/mo for OpenInsider Pro (with no API), $35/mo for TipRanks (single-insider only), or $25K/yr for Bloomberg.

Output format

Apify default dataset — pull as JSON, JSONL, CSV, Excel, RSS, or HTML. Programmatic SDK access via the Apify client (Python, Node.js, PHP, Ruby) or direct REST. Webhook-able on completion for downstream pipeline triggers.

Author / Notes

Built by NexGenData. Direct OpenInsider + SEC EDGAR data — no intermediary aggregators, no licensing layer, no monthly fees, no API key. The cluster-detection logic is open and documented above so you can verify the score weights against your own academic-finance references. PRs and feature requests welcome.

Disclaimer: insider trading data is informational. Past insider behavior does not guarantee future stock performance. Insider transactions are reported to the SEC with a 2-business-day lag and may include hedged / planned (10b5-1) trades that we cannot always distinguish from open-market buys.

---

About NexGenData

NexGenData publishes 220+ buyer-intent actors covering SEC filings, YC alumni, Delaware DOC, global stock screeners across 30+ exchanges, IPO calendars, IP and patent intelligence, FDA approvals, B2B lead generation, and more. Every actor is pay-per-result with no seat licensing.

Apify affiliate program — free credits + 30% off

Sign up to Apify via our referral link and you'll get:

• Free starter credits to test this actor and the rest of our 220+ actor fleet
• 30% off platform fees for the life of your account

Browse the full NexGenData catalog and sign up here — same Apify, same actors, just cheaper for you.

Built and maintained by NexGenData.