TraceAlpha Methodology
Understanding our Alternative Data Extraction
How does the daily alternative data pipeline work?
TraceAlpha operates an automated, scheduled data ingestion pipeline that executes daily across global records:
- Discovery: Scheduled scripts query international registry registries, customs logs, and patent databases to check for new documents associated with our trackable entities, immediately filtering out previously indexed sources using hashed URLs.
- Ingestion: Discovered documents are fetched, converted to clean plain-text structures, and stored locally for auditability.
- AI Extraction: Our advanced processing engine utilizes the latest models to analyze the documents under strict JSON schemas. The AI maps raw events to correct entity FIGIs, computes sentiment, assigns materiality, and generates a structured investment thesis.
How does the Sentiment Score work?
Our sentiment score is a continuous value ranging from -1.00 (Highly Negative) to +1.00 (Highly Positive).
Unlike traditional sentiment analysis that simply counts "happy" or "sad" words, our scoring is strategically weighted by our AI engine. It evaluates the actual macroeconomic or business impact of an event. For example, a "Supply Chain Disruption" that halts production will score heavily negative, whereas a "Strategic Partnership" that opens a new market will score highly positive. A score near 0.0 indicates a neutral or developing situation.
What is the Materiality Score & Weighted Sentiment?
To filter noise from critical information, we introduce two key investment metrics:
1. Materiality Score (1 to 5): This score measures the financial or operational scale of an event relative to the company's market cap or asset size. A score of 1 indicates a minor event with minimal balance-sheet impact (e.g., standard localized hiring). A score of 5 represents a highly material event (e.g., multi-billion dollar mergers, critical factory closures, or broad patents transforming the company's product line).
2. Weighted Sentiment: Traditional sentiment models compute a simple average of sentiment scores, which can easily be skewed by a high volume of low-impact press releases. TraceAlpha calculates a Weighted Sentiment by multiplying each signal's sentiment score by its materiality score. This prioritizes highly material signals, giving quant researchers a high-fidelity representation of genuine company momentum.
What are Market Outlook & Investment Thesis?
Rather than forcing analysts to read through hundreds of pages of raw shipping manifest details or regulatory permits, TraceAlpha extracts two key structured features:
- Market Outlook: Bounded to Bullish, Bearish, or Neutral, this field classifies the directional asset-price implication of the extracted signal over a 1 to 3 quarter horizon.
- Investment Thesis: A concise analyst thesis explaining how and why this event impacts the company's valuation, supply-chain resilience, or competitive advantage.
What is the methodology behind the Signal Text?
TraceAlpha does not rely on manual human data entry. Instead, our automated pipeline continuously ingests unstructured documents from across the globe.
These documents are processed through our AI node utilizing the latest models. We enforce a strict extraction schema using Pydantic, instructing the AI to act as a quantitative analyst. The AI isolates the core event, categorizes it into bounded event types (e.g., Patent Filing, Talent Pivot, Earnings Report), and synthesizes the "Impact Summary"—a concise, 1-2 sentence analytical breakdown of the strategic value hidden in the text.
What is a FIGI and why do we use it?
FIGI stands for the Financial Instrument Global Identifier. It is an open-source, globally recognized standard for identifying financial instruments.
Ticker symbols are notoriously messy—they change when a company rebrands (e.g., Facebook changing from FB to META) and the same ticker can overlap across different global exchanges. A FIGI, however, is a persistent, globally unique 12-character alphanumeric string that never changes.
How we lean into it: TraceAlpha explicitly resolves every target company to its Composite FIGI using the OpenFIGI API before we even begin extracting data. When you query our API or use our Terminal, you can search directly by FIGI to perform perfectly deterministic joins against your own internal quantitative databases, completely eliminating ticker mapping errors.
Why purchase a Premium API Key?
The public TraceAlpha Terminal provides a delayed, freemium view of our data. Signals extracted within the last 48 hours have their impact summaries, sentiment scores, market outlooks, and investment theses redacted.
- Real-Time Access: Instantly unlock signals the moment they are extracted from our AI node.
- Programmatic Integration: Feed quantitative signals directly into your trading algorithms, risk models, or internal dashboards.
- Full Historical Pagination: Query our entire historical database without UI restrictions.
Usage & Data Accuracy
TraceAlpha is designed as a powerful discovery and research tool. Our pipeline automates the ingestion of unstructured global records to synthesize high-velocity alternative data signals, significantly accelerating your quantitative analysis workflow.
While our AI extraction engine (utilizing latest models) uses highly precise structured schemas, natural language processing models can occasionally misinterpret context or generate errors. These signals are built to augment your research and should be integrated as part of a broader quantitative framework rather than relied upon as sole investment advice.
We prioritize transparency and auditability: every signal contains a direct link to the original source document. We encourage users to verify critical data points by reviewing the original source context.