Why a systematic framework beats intuition
The greatest investors in history didn't pick stocks on instinct. They built repeatable systems — specific criteria that a business had to meet before a single dollar was committed. The system removes emotion from the equation.
Buffett has described his investment process as running through a mental checklist. Graham codified his criteria so precisely that he could train students to apply them. The edge isn't in finding hidden information — it's in consistently applying a rational framework when most investors are acting emotionally.
What follows is our implementation of that framework — the exact filters our AI scanner runs on every company in the SEC database.
Stage 1 — The quantitative screen
Every company must pass all of these filters before advancing. These aren't guidelines — they're hard cutoffs. A company that fails any single filter is eliminated, regardless of how attractive it looks on other metrics.
| Metric | Threshold | Why it matters |
|---|---|---|
| Return on Invested Capital (ROIC) | > 10% | Measures how efficiently the business creates value. Below 10% means the business is destroying economic value at the cost of capital. |
| Debt-to-Equity Ratio | < 2.0 | High leverage amplifies losses in downturns. Graham was adamant: a leveraged company has no margin of safety. |
| Free Cash Flow (TTM) | Positive | Earnings can be manipulated. Cash flow is harder to fake. We require positive FCF over the trailing twelve months. |
| Margin of Safety vs. DCF | 28–38% (sector-adjusted) | The price must be at least 28% below our intrinsic value estimate. Thresholds vary by sector volatility. |
| Exchange | NYSE, NASDAQ, AMEX | OTC and pink sheet stocks are excluded. Liquidity and disclosure requirements matter for fair valuation. |
| Beta | Logged, not filtered | We track beta for context but don't eliminate on it. High beta stocks trading at deep value can offer exceptional risk-adjusted returns. |
Stage 2 — The DCF valuation
Companies that pass the quantitative screen receive a full discounted cash flow valuation. This is where we calculate intrinsic value — the number we compare against the current market price to determine the margin of safety.
How we build the DCF
We pull financial data from multiple sources: SEC EDGAR filings (the primary source, as reported directly to regulators), Yahoo Finance for market data, and Financial Modeling Prep for supplementary historical data. The DCF inputs are:
- Free Cash Flow (base): Trailing twelve months FCF, normalized for one-time items where identifiable
- Weighted Average Cost of Capital (WACC): Calculated using the CAPM model with sector-specific equity risk premiums
- Terminal Growth Rate: Conservative — typically 2-3%, reflecting long-run GDP growth
- Net Debt adjustment: Enterprise value minus net debt divided by shares outstanding gives per-share intrinsic value
We don't use a single point estimate. We stress-test across three scenarios (bear/base/bull) and use the conservative case as our working intrinsic value. The margin of safety must hold even in the bear scenario.
Stage 3 — The LLM qualitative analysis
Numbers tell you what happened. SEC filings tell you why — and what management thinks will happen next. For every company that passes quantitative screening, our AI reads the most recent 10-K or 10-Q filing from EDGAR and evaluates:
What a real signal looks like
This is a representative example of a DEEP VALUE signal from our scanner — anonymized but based on a real output:
What the framework doesn't do
Transparency matters. Here's where our approach has known limitations:
- It doesn't predict timing. A stock can be deeply undervalued for 18 months before the market recognizes it. Value investing requires patience — often more than feels comfortable.
- It doesn't catch all fraud. Financial fraud that hasn't surfaced in filings won't be caught by any systematic screen. Position sizing and diversification are your protection here.
- The DCF assumes ongoing concern. If a company is in terminal decline, our valuation model will overestimate intrinsic value. The LLM red flag scan is designed to catch this — but isn't infallible.
- Our backtest beat-SPY rate is 45.6%. We publish this honestly. Over a full market cycle we expect it to improve, but we won't cherry-pick periods to make it look better than it is.
"Knowing what you don't know is more useful than being brilliant." — Charlie Munger
How to use the signals
The signals are a starting point for your own research, not a replacement for it. When you receive a DEEP VALUE signal, we recommend:
- Read the most recent 10-K yourself — particularly the risk factors and MD&A sections
- Understand the industry dynamics and who the main competitors are
- Verify that you understand why the stock is trading at a discount (there's always a reason — is it temporary or structural?)
- Size your position appropriately — even high-conviction signals warrant position sizing discipline
The scanner surfaces the opportunity. The judgment call is still yours.