TRANSPARENCY
How We Analyze
Every number in our articles comes from a verifiable source. Here's exactly how our research process works to ensure institutional-grade precision.
Our Data Sources
SEC Edgar Filings
Real-time monitoring of all Form 4, 13F, and Schedule 13D filings with automated parsing and classification of transaction types.
Primary source for all insider trading data
Financial Datasets API
Comprehensive financial data including income statements, balance sheets, cash flow, key ratios, and 10-year historical price data.
financialdatasets.ai — trusted by hedge funds
Market Intelligence
Earnings call transcripts, analyst consensus estimates, and real-time news feeds cross-referenced with insider filing timestamps.
Verified citations with page numbers
Our Analysis Process
STEP 01
Data Collection
Automated pipelines continuously monitor SEC EDGAR for new Form 4 filings. Each filing is parsed, validated, and enriched with company fundamentals within 2–4 minutes.
STEP 02
AI Research Agent
Our Dexter research agent aggregates financial data, insider history, price action, and sector context into a structured intelligence package for analysis.
STEP 03
Article Generation
Claude Sonnet 4.6 generates institutional-grade analysis. Every sentence must contain a verifiable fact. A 14-point quality gate ensures publication standards.
STEP 04
Quality Gate
Automated checks verify title length, meta descriptions, banned phrase detection, keyword placement, and verdict validation before any article is published.
We Use AI — And We're Transparent About It
At EarlyInsider, we leverage advanced Large Language Models to synthesize vast amounts of financial data into readable, actionable insights. However, we maintain a human-in-the-loop oversight to ensure the highest standards of accuracy.
Crucially, the AI is not “hallucinating” data points. Every ticker symbol, purchase price, and percentage change is pulled directly from official SEC EDGAR filings via our proprietary data pipeline. The AI serves as a high-speed research analyst, identifying correlations that might take a human hours to uncover.
Limitations
Reporting delays: Insiders have 2 business days to file Form 4. Some transactions are reported late. We flag late filings but cannot detect unreported trades.
10b5-1 Plans: Pre-scheduled trading plans create noise. Our model downweights routine plan executions but cannot always distinguish conviction from mechanical selling.
AI limitations: Our analysis is probabilistic, not deterministic. Conviction scores reflect historical patterns but cannot predict future outcomes with certainty.
Past performance: Historical backtest results do not guarantee future performance. Markets evolve and insider signals can be noisy during regime changes.
Verify Our Data
Every data point in our articles can be independently verified.