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
Noise Filtering
80% of Form 4 filings are routine: scheduled 10b5-1 plan executions, small option exercises, and tax withholding. The filtering engine classifies each transaction against 14 criteria to isolate the 20% that reflect discretionary conviction.
STEP 03
Conviction Scoring
Each filing that passes the noise filter receives a conviction score from 0 to 100. The scoring model weighs 7 factors including trade size, executive track record, cluster activity, and sector context. 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
The analysis engine uses large language models to generate plain-English filing context, research report narratives, and conviction score explanations. The models operate exclusively on structured data extracted from SEC filings and verified financial datasets. No model has access to material nonpublic information.
Every AI-generated narrative is constrained by the underlying structured data. If the filing data says the CEO purchased 50,000 shares at $34.12, the narrative cannot state a different figure. Conviction scores are computed algorithmically from 7 weighted factors; the language model explains the score but does not determine it.
Limitations
Mandatory reporting lag. SEC Form 4 filings must be submitted within 2 business days of the transaction. EarlyInsider detects filings within seconds of SEC EDGAR publication, but the underlying trade may have occurred up to 48 hours earlier.
10b5-1 pre-planned trades. Insiders who file trades under Rule 10b5-1 plans execute on a predetermined schedule, not in response to current information. The noise filter flags known 10b5-1 transactions, but not all plans are publicly disclosed.
AI analysis scope. The analysis engine operates on public data only: SEC filings, market prices, historical fundamentals, and corporate event calendars. It cannot assess management sentiment from private communications or pending regulatory actions.
Historical performance is not predictive. Academic research demonstrates that insider purchases have historically outperformed market benchmarks. Past patterns do not guarantee future results for any individual filing or trade.
Verify Our Data
Every data point in our articles can be independently verified.