Research Signal

Research Signals exist only in analysts' heads and informal conversations. A senior analyst tells a PM 'I think industrials are about to break out based on my channel checks' during a hallway chat. There is no written signal, no confidence level, no time horizon. When the PM acts on it and loses money, nobody can reconstruct what the signal actually said or what evidence supported it.

None — AI cannot process Research Signals because no signal records exist. Systematic signal evaluation, backtesting, and performance tracking are impossible without documented signal definitions.

Research Signals are documented in analyst notes and morning meeting summaries. A note might read 'Strong buy signal on TSLA — technicals showing golden cross, fundamentals support with delivery numbers above consensus.' But signals are narrative prose — there is no standard format for signal type, strength, confidence interval, or expected holding period. Every analyst describes signals differently.

AI could attempt NLP extraction of signal direction and target security from analyst notes, but cannot reliably parse confidence levels, time horizons, or supporting evidence from free-text narratives. Systematic signal aggregation across analysts is unreliable.

Research Signals are entered into a research management system with standard fields — signal type, direction, strength (1-10 scale), confidence percentage, time horizon (days/weeks/months), and target security. Analysts submit signals through a form that enforces required fields. But supporting evidence is a free-text field with no structure. Two analysts rate the same stock differently with no way to compare their analytical basis. The signal record says 'strong buy, 85% confidence' without machine-readable justification.

AI can aggregate signals by type, rank by confidence, and track signal-to-outcome accuracy over time. Cannot evaluate signal quality or identify conflicting analytical bases because supporting evidence is unstructured narrative rather than decomposed factors.

Research Signal records contain structured signal parameters and decomposed supporting evidence. Each signal links to categorized factors: earnings revision magnitude, relative strength index values, options flow sentiment readings, credit spread movements. Factor data points include source, observation date, and statistical significance. A quant can query 'show me all fundamental signals where the primary factor was earnings revision exceeding two standard deviations and the time horizon was under 30 days' and get a precise result set.

AI can evaluate signal quality by comparing factor decompositions against historical outcomes, identify which factor combinations produce the highest hit rates, and flag signals with contradictory factor evidence. Systematic signal-weighted portfolio construction becomes feasible.

Research Signals are schema-driven entities with formal relationships to factor models, alternative data pipelines, and portfolio construction frameworks. Each signal encodes its alpha source, expected decay profile, correlation with existing portfolio exposures, and capacity constraints. An AI agent can ask 'which active Research Signals have alpha decay horizons under 5 days, are uncorrelated with our current momentum tilt, and have sufficient capacity for a $50M position?' and get a precise, actionable answer.

AI can autonomously construct signal-weighted portfolios, optimize position sizing based on alpha decay and capacity, and dynamically adjust signal weights as market conditions change. Full autonomous signal-to-portfolio translation for systematic strategies.

Research Signals are living entities that generate and recalibrate themselves continuously from streaming data. Market microstructure patterns, earnings transcript sentiment, satellite imagery, credit card transaction flows, and social media signals all feed into real-time signal generation engines. Each signal carries a continuously updated alpha estimate, decay curve, capacity limit, and factor exposure profile. The Research Signal is not something an analyst 'creates' — it emerges from the data and presents itself to the investment team with full supporting evidence.

Fully autonomous Research Signal intelligence. AI generates, evaluates, combines, and acts on signals end-to-end. The signal pipeline from raw data through alpha estimation to portfolio construction operates without human intervention for systematic strategies.

Ceiling of the CMC framework for this dimension.