Signal Engine
How CPI turns SEC filings into climate-disclosure signals.
The signal engine is a dual-oscillator system that reads every 10-K and 10-Q filed with SEC EDGAR, extracts climate-related language, and converts it into structured, comparable signals.
Data source
All signals are derived from SEC EDGAR filings — specifically the full text of 10-K (annual) and 10-Q (quarterly) reports. These are the same documents companies file with the Securities and Exchange Commission. No third-party scores, no surveys.
The engine parses four key sections of each filing:
| Section | Label | What it covers |
|---|---|---|
| Item 1 | Business | Company overview, operations, strategy |
| Item 1A | Risk | Risk factors including climate-related risks |
| Item 7 | MD&A | Management discussion and analysis |
| Item 8 | Financials | Financial statements and notes |
The breathing machine
The core system is called CASLO — a dual-oscillator model where each company maintains two independent signal processors:
- Fast oscillator — reacts to each new filing, measuring the immediate signal from climate-related language
- Slow oscillator — tracks the baseline expectation, adapting gradually to the company's disclosure norm
The disclosure intensity is the divergence between these two oscillators. When a company's climate language deviates from its own baseline, the intensity rises. A company that has always disclosed heavily will have a different intensity than one that just started.
Observation channels
Each filing observation produces five channel measurements:
| Channel | What it measures |
|---|---|
| Volume | Total amount of climate-related language |
| Change | How much the language changed from the previous filing |
| Spread | How broadly the language is distributed across filing sections |
| Materiality | Concentration in financially material sections (Risk, MD&A) |
| Novelty | New climate terms or phrases not seen in prior filings |
All channels are normalized to 0–1. Higher values indicate stronger signal on that dimension.
From signals to patterns
The engine classifies each company into a disclosure pattern based on its intensity and filing history:
| Pattern | Meaning |
|---|---|
| Baseline | Minimal climate disclosure, close to the company's norm |
| Emerging | Growing climate language that deviates from baseline |
| Established | Sustained, significant climate disclosure |
Each company also has a trend direction:
- Intensifying — disclosure intensity is increasing
- Steady — disclosure intensity is stable
- Easing — disclosure intensity is decreasing
For a practical guide to interpreting these signals, see Reading the Signals.
Update cadence
Signals are recomputed whenever new filings appear on SEC EDGAR. In practice this means daily updates during peak filing seasons (February–March for annual 10-Ks, and after each fiscal quarter for 10-Qs).