Methodology
Every CEO purchase we surface is scored the same way, by the same process, every day. Here is exactly how a filing becomes a signal — and where the method can be wrong.
1. What qualifies
We only consider open-market purchases by a chief executive, disclosed on SEC Form 4. We deliberately exclude:
- Option exercises and grants (compensation, not conviction)
- Automatic / 10b5-1 scheduled buys (no discretionary signal)
- Sales of any kind (selling has little predictive value — why)
- OTC and penny stocks (data quality and liquidity)
2. The six factors
Each qualifying purchase gets a weighted composite score from 0–100. The weights:
| Factor | Weight | What it measures |
|---|---|---|
| Fundamental | 25% | Business quality, valuation, growth — is the underlying company sound? |
| Financial | 25% | Balance sheet, margins, cash flow, debt — financial health and durability. |
| Technical | 20% | Price trend, momentum, and where the buy sits relative to the range. |
| Conviction | 10% | Purchase size relative to the insider's compensation and prior buying. |
| CEO history | 10% | Track record of this insider's past open-market purchases. |
| Cluster | 10% | Whether multiple executives bought around the same time (strongest variant). |
3. The opinion scale
The composite score maps to an opinion: STRONG BUY > BUY > WATCH > AVOID > STRONG AVOID. Only BUY and STRONG BUY enter the tracked portfolio.
4. How signals are tracked
Each BUY / STRONG BUY opens a hypothetical $100 position at the signal's entry price, with a matching $100 SPY position bought the same day. Mechanical exits:
- +10% → close at target
- -15% → close at stop loss
- Otherwise → stay open, marked to the latest price
Alpha is our signal's return minus SPY's return over the same window. Every position is public on the track record and aggregated in the 2026 performance report.
5. Where this can be wrong
- The analysis layer uses AI and can misread data, miss context, or hallucinate. We surface the reasoning so you can check it.
- Insider buying is probabilistic. It is documented to predict positive abnormal returns on average and over time — not on any single name.
- Mechanical exits are blunt. A +10%/-15% rule is simple and testable, but it will sell winners early and ride some losers to the stop.
- The portfolio is hypothetical — no slippage, fees, or taxes. Real results would differ.
The academic basis for tracking purchases (Lakonishok & Lee; Jeng, Metrick & Zeckhauser; Cziraki) is reviewed in this post. Not financial advice.