An InsiderAlpha classification for multiple unique reporting persons making eligible Code P purchases in the same ticker and defined time window.
Definition
Cluster buying is a derived pattern rather than an SEC term. InsiderAlpha evaluates eligible Code P records by ticker, unique reporting owner, transaction date, and rolling 7-, 30-, or 90-day windows, while preserving transaction value, role, ownership nature, and filing lineage.
Cluster buying is an InsiderAlpha-derived pattern, not an SEC classification or a universal three-insider rule. Every published cluster must disclose its eligible code, unique-owner logic, time window, data coverage, and algorithm version.
Related Form 4 codes
Current Methodology
A cluster starts with eligible Type = P rows grouped by ticker and transaction date. Reporting persons are deduplicated by reporting_owner_cik where available, not by transaction-row count. InsiderAlpha stores rolling cluster_size_7d, cluster_size_30d, and cluster_size_90d features plus cluster_algorithm_version.
A person with several same-day purchase rows counts once for unique-insider breadth, while the eligible nominal values can still be aggregated. Code A grants, Code M exercises, and Code F withholding do not become cluster purchases.
Quality and Interpretation
Amendments, superseded rows, missing CIKs, private Code P purchases, indirect ownership, and incomplete historical coverage can change a cluster count. The published result should carry quality status and a link to every contributing accession.
A cluster describes multiple reported purchases; it does not prove coordination, consensus, undervaluation, or a forthcoming catalyst. Role and amount can be separate model features but should not silently change the unique-buyer count.
Common Analytical Errors
Error 1: counting rows instead of people. One filing can contain several rows.
Error 2: mixing A or M with P. Awards and exercises are different events.
Error 3: presenting a rolling feature without its as-of date. Cluster size changes as the window moves.
Error 4: claiming statistical strength without a published backtest. Pattern description and predictive evidence are separate.
Cluster Feature Map
| Field | Role | Validation |
|---|---|---|
| reporting_owner_cik | Unique buyer identity | Fall back cautiously when CIK is missing; names can vary. |
| cluster_size_7d/30d/90d | Rolling breadth | Publish window, as-of date, and eligible rows. |
| is_cluster_buy | Derived classification | Not an SEC field; tie it to algorithm version. |
| accession_number | Source lineage | Expose all contributing filings and amendments. |
Primary Sources & Filing References
- SEC Form 4 General Instructions
Primary source for Code P and the ownership fields underlying the derived cluster.
Live Insider Data
Track real-time buying, selling, and ownership-change activity on high-volume tickers:
Why it matters for Whale Tracking
Multiple independent purchase records can provide a broader ownership pattern than one transaction. The pattern does not prove coordination, undervaluation, or future performance and should be interpreted with coverage, role, amount, amendments, and common filing context.
Technical Nuance
A reproducible cluster must deduplicate amended or superseded rows, identify unique reporting-owner CIKs, avoid counting multiple rows from one person as multiple insiders, and disclose whether private Code P transactions and indirect ownership are eligible.
Track Cluster Buyings Live
Stop reading history. See what corporate insiders are buying right now in our real-time terminal.
Real-World Example
"If three distinct reporting-owner CIKs file eligible Code P transactions for one ticker inside 30 days, the 30-day cluster count is three even when one filing contains multiple transaction rows."
Inspect before purchasing
Preview 10,000 real rows and all 64 fields
Validate the schema and research workflow before buying the 1,633,946-row release. No email or account is required for the sample.
- Reproduce a published report
- Run the included queries with DuckDB
- CSV, Parquet and DuckDB in the full release
- Three months of API access included
One-time payment · instant download · regular price $199