Narrative Drift Score · Institutional Investor Edition · 2026
VelesAnalytics NDS quantifies systematic language drift in SEC 10-K and 10-Q disclosures — the early warning that precedes adverse events, automated across your entire portfolio.
Selective onboarding · Institutional clients only
The Problem
A holding in your portfolio filed its 10-K last night. Management softened forward guidance, introduced three new risk factors, and buried the shift across 94,000 words of structured disclosure.
Your analysts are busy. By the time the news breaks, the stock has already moved. You will not catch it in time — not manually, not at scale.
Cohen, Malloy, and Nguyen (2020) documented in the Journal of Finance that firms with minimal textual changes in their filings underperform by approximately 1.5% per month. The signal exists. The tools to extract it systematically do not.
Until now.
The Product
The Narrative Drift Score (NDS) is a per-company composite metric quantifying how a firm's disclosure language evolves over time. Each component is grounded in peer-reviewed academic research and produces an actionable output requiring no interpretation of raw NLP results.
Elevated drift detected across linguistic complexity, hedging density, and omission patterns. Recommend full filing review prior to next earnings event.
Z-score changes in Fog Index and sentence length relative to the firm's own rolling five-year baseline. Increasing complexity signals deliberate obfuscation.
Percentage changes in uncertainty terminology per 1,000 words. Flags increasing qualification of forward statements using the Loughran-McDonald dictionary.
Decline in quantified projections — monetary targets, percentages, and dates — versus prior filings. Vagueness precedes disappointment.
Identification of newly introduced risk categories versus boilerplate repetition from prior disclosures. New risks rarely appear without cause.
Topics and entities present in prior filings that silently disappear in current reports. What management stops saying is as informative as what it says.
Jaccard similarity detecting copy-paste disclosure patterns — a signal of management complacency or deliberate concealment of material change.
Direct EDGAR API integration · FinBERT transformer models · Loughran-McDonald financial sentiment dictionaries · Rolling sector-normalised baselines · Backtested against earnings surprises, restatements, and bankruptcy events · Sub-second API response via Redis caching
Who We Serve
Systematic coverage across 50–200 names. Early warning on existing positions before earnings confirm what the 10-K was already signalling. NDS is additive to Bloomberg and FactSet.
Fundamental research teams expanding coverage without adding headcount. NDS as systematic first-pass triage across the investable universe — not manual file review at 8 hours per filing.
Corporate bond risk assessment. Disclosure deterioration as a leading indicator for credit events — documented, repeatable, and defensible to risk committees and LPs.
Institutional-grade narrative monitoring without institutional-grade headcount. Show LPs a documented, systematic risk process that goes beyond standard financial data.
Pricing
Annual commitment required. Three-month pilot programs available for qualified institutions. · All prices in CHF. USD and EUR invoicing available.
Early Access Programme
NDS is currently in late development. We are building a small cohort of pilot institutions — funds and research teams who will be the first to integrate NDS into live workflows and shape the product before general availability.
Pilot participants receive locked early-access pricing for 12 months and priority support during onboarding. Submit your application below — we review each profile individually and respond within 2–5 business days.