Intelligence your competitors
do not have yet

Every major research team starts the day with the same data from the same terminals. ContextQuant adds an automated alert layer that flags companies whose filings changed in ways that historically predict underperformance, before the street has reacted. When a company's risk language becomes materially more specific than prior quarters and more specific than its peers, the system fires an alert with the historical evidence for what happens next.

When our accruals screen flags a company AND the filing language is cautious, the next quarter is particularly predictable (IC=-0.128 OOS, 3x the base signal, walk-forward confirmed 6/8 years). The forward-looking ratio is a contrarian early warning: when management shifts to more future-oriented language, current fundamentals are weakening (IC=-0.282 financials, -0.367 healthcare).

The composite signal achieves an Information Ratio of 1.41 across ten years of out-of-sample testing with zero sign flips. Each component is uncorrelated with traditional quantitative factors, adding genuine diversification to existing signal libraries. Four Haiku features (confidence, net sentiment, risk tone, forward ratio) provide factor-like dimensions from the same filing.

When a Regulation FD disclosure co-occurs with a filing signal, earnings prediction strength increases 10x (IC=-0.138 vs -0.013, N=1,411 vs 3,218). This defines a timing condition: enter positions when Reg FD + cautious NLP co-occur, not on NLP alone. The joint alignment filter (negative LM + downward revision) adds a second condition: 95.6% beat rate (N=135 OOS).

Filing language predicts ROA changes one quarter before the balance sheet shows them. The AI confidence score achieves IC=+0.398 OOS in financials (p<0.001, N=428), surviving all macro controls at t=+4.77. Higher management confidence predicts ROA improvement. This is the simplest, strongest credit signal in the program: one number from one filing predicts balance sheet direction.

When our signal fires negative and the company misses earnings, ROA deteriorates 50% of the time versus 21% when the signal is positive and the company beats. That is a 29 percentage point spread. The NLP signal predicts ROA most strongly for companies whose management guidance is least reliable (IC=+0.376 for slight-miss companies versus +0.005 for volatile beaters).

Management credibility classification plus AI confidence score provide plain-language quality metrics backed by five years of out-of-sample validation. The signal works best where management guidance is least reliable (IC=+0.376 for slight-miss companies versus +0.005 for volatile beaters). When your client asks "is this company getting better or worse," the confidence score answers with validated evidence.

The confidence score is a single number from 0 to 1 that predicts next-quarter profitability direction. Higher is better. IC=+0.398 OOS in financials (p<0.001, N=428). The forward-looking ratio adds a contrarian early warning: "This company's management is spending more time talking about the future than usual. Historically, this predicts the current quarter's fundamentals are weakening." Accessible, evidence-based, and counterintuitive enough to demonstrate analytical value.

Signal weights adjust automatically based on the current macro regime. During tariff escalation or rate transitions, risk specificity signals are upweighted. During calm markets, sentiment drift takes priority. The CRO's dashboard reflects the current environment, not a static model.

When multiple signals fire together on the same company (elevated risk specificity, negative outlook divergence, widening transcript-filing gap), the alert is high-confidence. When none fire, the system stays quiet. That asymmetry is a feature, not a bug.