$ prism-api --info

Behavioral Signal
Infrastructure

Quarterly behavioral signals covering 17,500+ companies across the US and Japan, derived from 57 psychometric dimensions scored on regulatory filings. Four headline signals. One raw panel. Delivered via CSV, API, or S3.

Flagship
Conduct Risk Index
+21.72% alpha | Sharpe 1.50 | t-stat 5.66 | 11/12 years positive | defensive beta
$ prism-api --list-products

Signal Suite

Conduct Risk Index R3K
+21.72% walk-forward alpha
Quarterly L/S composite. Sharpe 1.50, t=5.66, defensive beta. Validated walk-forward on Russell 3000, formation years 2013–2023.
Distinctiveness Index TSE
Behavioral uniqueness signal
Measures how behaviorally distinct a company is from its sector peers across 57 dimensions. Available as a raw feature in the quarterly panel.
Enforcement Risk Index
AUC 0.94 · ~2 quarter lead
Predicts regulatory enforcement actions. Nested cross-validation on S&P 500. 94.7% of signal variance unexplained by sector or size.
Prosecution Risk Score
Firm-level probability
Continuous enforcement probability per company. The signal is behavioral drift — firms losing prosocial tone in public communications.
Tier 2 — Raw Behavioral Panel
Full 57-dimension quarterly panel. 17,500+ companies, 2001–2026. Build your own signals.
$ prism-api --query conduct-risk

Conduct Risk Index

CARHART_ALPHA .. +21.72%Net of Carhart 4 factors. Walk-forward, not in-sample. annualized, walk-forward
SHARPE .......... 1.50Top decile among quantitative signals globally. 132 months
T_STAT .......... +5.66p < 0.001. Survives every standard correction. p < 0.001
MAX_DRAWDOWN ... -9.37%Worst peak-to-trough. Shallow drawdown relative to alpha magnitude.
MKT_BETA ....... -0.333Negative beta = the signal is defensive. Gains when markets fall. defensive
YEARS_POS ...... 11/1211 of 12 formation years positive. Worst year: -7.58% (2014). 92%
SMB_BETA ....... -0.130Not a small-cap effect. Signal works across market cap spectrum. not size-driven
UMD_BETA ....... -0.062Not a momentum effect. Statistically insignificant. not momentum

Walk-forward Carhart 4-factor regression. MiniLM embedder on Russell 3000 universe. Formation years 2013–2023.

Conduct Risk — Annual Returns

+25%
+3%
+13%
+4%
+14%
+37%
+23%
+9%
+40%
+26%
+6%
+9%
+11%
+48%
+16%
+10%
+27%
+19%
+22%
'06
'07
'08
'09
'10
'11
'12
'13
'14
'15
'16
'17
'18
'19
'20
'21
'22
'23
'24
'25
-10%
'06+25%
'07+3%
'08+13%
'09+4%
'10+14%
'11+37%
'12-10%
'13+23%
'14+9%
'15+40%
'16+26%
'17+6%
'18+9%
'19+11%
'20+48%
'21+16%
'22+10%
'23+27%
'24+19%
'25+22%

Full 20-year backtest range shown. Validated walk-forward window (formation years 2013–2023) delivers +21.72% net of Carhart 4 factors, Sharpe 1.50, 11 of 12 years positive.

17,500+
Companies Scored
400K+
Documents Analyzed
57
Behavioral Dimensions
2001–2026
US + Japan · Quarterly
$ prism-api --query enforcement-risk

Enforcement Risk

PROSECUTION_AUC . 0.94Nested cross-validation on S&P 500. 495 firms, 24 prosecuted. Honest out-of-fold scores. nested CV, S&P 500
LEAD_TIME ...... ~2 quartersSignal detectable before DOJ enforcement action. before enforcement
SIGNAL_TYPE .... Behavioral driftFirms losing prosocial tone over time, not cross-sectional level. temporal, not cross-sectional
ORTHOGONALITY .. 94.7%Signal variance NOT explained by sector + log-assets. Strongly idiosyncratic. unexplained by sector + size
FEATURES ....... 15Comparable AUC with 29 total features vs 242 in full model (1/8 the feature count). parsimonious composite
BOOTSTRAP_CI .. [0.91, 0.97] 95% CI, N=1000

The signal is drift, not level. Firms whose public communication loses its prosocial and moral tone over time are the enforcement-prone ones. Nested CV holds out each ticker before feature selection — no information leakage.

$ prism-api --methodology

How It Works

SIGNAL Behavioral scores from regulatory filings and earnings transcripts
METHOD Proprietary scoring methodology
UNIVERSE US Russell 3000+ and Japan TSE (17,500+ companies)
REBALANCE Annual
DIMENSIONS 57 proprietary behavioral constructs per company
SOURCES_US SEC EDGAR (10-K, 10-Q, DEF 14A) + earnings transcripts
SOURCES_JP TDnet/EDINET regulatory disclosures

Prism operationalizes "tone from the top" into a quantitative, cross-sectional measure. Conduct risk is a regulator-defined category with $41B+ in US enforcement fines since 2019. All inputs are public documents. No web scraping, no PII, no MNPI.

$ prism-api --pricing

Access

Evaluation
Sample panel, case study, coverage map, and full validation deck.
Evaluation under NDA
Tier 1 — Headline Signals
Conduct Risk Index, Distinctiveness Index, Enforcement Risk Index, Prosecution Risk Score. Quarterly, via API, CSV, or S3.
Annual institutional license
Tier 2 — Raw Panel
Full 57-dimension quarterly panel. 17,500+ companies, US + Japan. Build your own signals.
Premium license
Listed on Neudata  ·  Eagle Alpha

Request Evaluation

Sample panel, case study, coverage map, and full validation deck. Free under NDA.

prism@stronglight.io