Parallax
Narrative Risk Intelligence

Language shifts before crises do.
Parallax detects the signal before the headlines.

Request early access See the evidence →

"In every major crisis we analyzed, the language changed first.
Not the headlines, not the stock price — the language."

Parallax retrospective analysis · Ten documented cases · 2015–2023
FTX · November 6, 2022
"FTX is fine. Assets are fine. We don't invest client assets. We have been processing all withdrawals, and will continue to be."
A $6 billion bank run was already underway. The exchange had been using customer deposits through its sister trading firm, Alameda Research. Liquidity was gone.
$32 billion collapsed. $32B → bankruptcy in 10 days.
94
Parallax score High Risk · Day 1 signal
Boeing · November 2, 2018
"The 737 MAX is as safe as any airplane that has ever flown the skies. I've flown on it myself, and I'd put my family on it. I trust it."
Two weeks earlier, Lion Air Flight 610 had killed 189 people. The MCAS software system was under active investigation. A redesign was already underway — never disclosed.
Four months later, 157 more people died. 21-month global grounding.
82
Parallax score High Risk · 4 months early
SVB · March 8, 2023
"We are confident that our liquidity, capital, and financial strength remain robust. Our balance sheet is strong and our deposit base is diversified."
SVB had already taken a $1.8B loss selling its bond portfolio. A Moody's downgrade was pending. The letter's existence was itself the signal — stable banks don't write this letter.
$42B bank run in 24 hours. Collapsed March 10 — 48 hours later.
91
Parallax score High Risk · 48 hours early
What Parallax detected

In each case above, the statement was already public.
Parallax would have flagged it.

The signal was not in what was said — it was in how the language was behaving. Certainty where there should be doubt. The framing shifting from facts to values. And most consistently: context that should have been there, missing. These are measurable, documentable, historically consistent patterns. Parallax does not track what is being said about a company. It tracks how the language is changing — and whether that change matches the pattern that preceded every crisis in this dataset.

How it works

Watchlist in.
Early warning out.

Four stages. No integration required. No internal data needed. Operational from day one.

01 — Ingest
Submit a watchlist
Company names, executives, products, known pressure points, direct competitors for baseline comparison. Five minutes to set up. Parallax works entirely from public sources — no sensitive data, no API, no IT approval required.
02 — Monitor
Track language shift, not volume
Parallax continuously scans news wire, trade press, financial filings, SEC EDGAR, earnings transcripts, analyst commentary, and advocacy publications. It is not counting mentions or measuring sentiment. It is measuring how the language is changing across five signal dimensions.
03 — Detect
Score against an established baseline
Every document is scored against three baselines simultaneously: the entity's own historical language, peer organizations in the same sector, and industry-wide norms. A score reflects deviation in standard deviations — what changed, by how much, and in which direction.
04 — Alert
Receive a signal with time to act
When a threshold is crossed: which client, which signal class, what changed, the specific evidence, and a recommended action. Scores run 0–100 across five tiers — Clear, Monitor, Elevated, Warning, High Risk. Every alert is explainable. Every output is defensible in a client meeting.
Built on
Fully explainable signals
Nothing is a black box. Every alert includes the source documents, baseline comparison, deviation score, and specific language evidence. You can always explain — precisely — why Parallax flagged something. The intelligence layer beneath the retainer.
Backtested evidence

Five major crises.
Four early warnings.

The full dataset. Applied retrospectively across five of the last decade's most documented corporate crises. Pre-escalation language signals appeared in four of five cases. In the fifth, the risk was real — the outcome was positive because the organization had prepared.

80–100 · High Risk
66–79 · Warning
46–65 · Elevated
21–45 · Monitor
0–20 · Clear
94
FTX Collapse · 2022
Sam Bankman-Fried: "FTX is fine. Assets are fine." — maximum declarative certainty while a $6B bank run was already underway. The statement named the situation while withholding the cause: Alameda Research had been using customer deposits. Certainty-reality gap: 100%.
Certainty Drift · 88 Omission Anomaly · 94
⚡ Day 1 signal
$32B → bankruptcy
in 10 days
91
SVB Collapse · 2023
CEO Greg Becker's letter: five distinct confidence assertions in 90 words — "robust," "strong," "confident," "diversified," "committed." A $1.8B bond sale disclosed with zero explanation of the loss. The letter's existence was the omission anomaly — stable institutions don't write to clients to say their balance sheet is strong.
Certainty Drift · 88 Omission Anomaly · 94
⚡ 48 hours early
$42B bank run
Second-largest US bank failure
82
Boeing 737 MAX · 2019
CEO Dennis Muilenburg: "As safe as any airplane that has ever flown the skies" — a maximum superlative claim during an active fatal accident investigation, without evidentiary grounding. Critical omission: zero acknowledgment of 189 victims, zero disclosure of the MCAS software redesign already underway.
Certainty Drift · 88 Omission Anomaly · 91
⚡ 4 months early
346 deaths
21-month grounding
78
Bud Light × Dylan Mulvaney · 2023
VP of Marketing Alissa Heinerscheid's March 23 interview: maximum certainty about a contested brand identity pivot — "we need to evolve" framed as moral imperative — with zero risk acknowledgment, zero contingency language, and zero acknowledgment of the existing customer base's likely reaction.
Moral Framing · 95 Omission Anomaly · 88
⚡ 9 days early
$6B market cap lost
Sales −26%
72
Nike × Colin Kaepernick · 2018
Real and significant risk correctly flagged — boycotts, stock drop, intense moral controversy. Nike had prepared. Outcome was positive. This case proves the core point: Parallax measures exposure, not outcomes. The difference between Nike and Bud Light is not risk level — it is preparation. Management decides what happens next.
Discourse Signals · 79 Trajectory · 71
Risk managed well
+$6B brand value
+31% sales
"Omission anomaly appeared in every single case."
The most reliable pre-crisis signal is not what an organization says — it is what they leave out. When expected context systematically disappears from communications, risk is accumulating invisibly. Parallax was built to measure that gap.
Use cases

Five ways organizations
use Parallax.

The same watchlist engine and signal detection layer. Five distinct applications across different buyers, different stakes, and different moments in the risk lifecycle.

For PR & Crisis Consultancies
Client defense
Detect a forming crisis before your client knows it exists
A crisis firm's pharma client has a growing side-effect conversation in advocacy publications. Parallax detects certainty drift over six weeks — language moving from "some patients report" to "causes" — combined with an anomalous silence in the client's own communications. The consultancy calls the client before any journalist does.
Sample Parallax alert
Warning · Score 67. Certainty language in patient advocacy coverage 2.1 SD above baseline. Client communications show zero acknowledgment of side-effect discourse for 14 days — anomalous given coverage volume. Recommend proactive client conversation within 48 hours.
Competitive intelligence
Get ahead of sector contagion before it reaches your client
A regional bank is on the firm's roster. SVB is on the watchlist as a sector monitor. On March 8 2023, Greg Becker's letter publishes. Parallax reads it within hours. Five certainty assertions in 90 words. A $1.8B portfolio liquidation with zero explanation. The firm calls its banking client that morning — before the bank run begins.
Sample Parallax alert
High Risk · Score 91. SVB CEO communication. Certainty stacking 3.4 SD above regional bank baseline. Material balance sheet event disclosed without causal context. Recommend activating Tier 1 surveillance. Monitor for contagion signals within 6–12 hours.
Business development
Turn signals into new client relationships
The firm runs Parallax across 200 companies continuously — not just current clients. When a Warning signal fires on a company they don't represent, the senior partner reaches out with specific intelligence about what the language around that company looks like right now — and why it matters. That email gets opened.
What this enables
The signal is the pitch. "We've been watching your situation and have identified some pre-escalation patterns worth a conversation. Can we show you what we're seeing?" No cold pitch competes with that opening.
For Enterprise & Investors
Enterprise risk team
Monitor activist language for coordinated campaign signals
A Fortune 500 food company monitors five activist organizations on their watchlist. Parallax detects a linkage signal: three separate advocacy groups begin using identical framing — "systemic greenwashing" — within the same week. Actor signals fire: two climate journalists have followed all three organizations within days of each other. The risk team briefs leadership six weeks before the first journalist inquiry arrives.
Sample Parallax alert
Warning · Score 71. Coordinated phrase reuse detected across three advocacy organizations — 94th percentile. Actor signal: two climate journalists showing anomalous engagement with coordinating sources. Trajectory: early stage. Recommend proactive response framework within 30 days.
Venture capital
Detect founder communication patterns that precede financial distress
A VC firm monitors a fintech portfolio company. Parallax detects an epistemic signal building over four months: certainty in the founder's public communications is increasing while specificity is decreasing — fewer numbers, more superlatives, more future-tense certainty. The pattern is structurally identical to pre-crisis profiles in comparable fintech failures. The firm requests updated financials. The numbers do not match the narrative.
Sample Parallax alert
Elevated · Score 61. Founder communications: certainty intensity increasing, evidence density decreasing over 120-day window. Pattern consistent with pre-escalation profiles in comparable fintech cases. Recommend requesting updated financial disclosure and direct founder conversation.
The competitive difference

A different moment.
A different product entirely.

The dominant narrative risk platforms detect attacks after they are visible at network scale. Parallax operates before that moment. That is not a feature gap — it is a category gap.

Incumbent tools · e.g. Blackbird.AI
Reactive intelligence
Fires after: crisis is already public and spreading
Detects narrative attacks after they are spreading at network scale — viral amplification, bot behavior, coordinated inauthentic activity. Blackbird raised $37M solving this problem well. By the time the alert fires, the client is already in the headlines.
  • Measures: spread velocity and network amplification
  • Fires: when the narrative is already spreading at scale
  • What it gives you: time to contain and respond
  • Input: network activity and social signals
vs
Parallax
Pre-escalation intelligence
Fires before: crisis is public — when the language is shifting
Detects epistemic drift in language — certainty shifts, moral framing escalation, omission anomalies — before any network spread begins. The signal is in the language itself, not the amplification pattern.
  • Measures: how language is shifting from its established baseline
  • Fires: before escalation becomes visible
  • What it gives you: time to prepare or prevent
  • Input: public media — news, filings, transcripts, trade press
Blackbird is a smoke detector. Parallax is a heat sensor. By the time the smoke detector fires, the fire is already real. The heat sensor gives you time to act before it ignites. Both have value. They operate at fundamentally different moments in the crisis timeline — and no product currently occupies the pre-escalation moment. Parallax is building that category.
The signal model

Five signal classes.
One is uncontested.

Every document analyzed by Parallax is scored across five dimensions. Four overlap partially with incumbent tools. One has no competitor.

Signal 1 · Primary differentiator · No competitor
Epistemic Signals

Changes in certainty, hedging language, and omission patterns in public communication. Language becomes epistemically unstable before a narrative escalates at scale. It appeared in every crisis case in the dataset. No incumbent platform measures it. It is the earliest signal — and the one that most consistently precedes catastrophic outcomes.

  • Certainty drift — "may impact" hardening to "will cause harm"
  • Omission anomaly — expected context systematically absent
  • Modal shift — hedging language replaced by declarative certainty
  • Evidence density collapse — claims become less grounded over time
  • Overclaiming under real uncertainty
Signal 2
Discourse Signals
Framing shifts, narrative themes, moral and emotional register changes. When operational language becomes moral language — "we made a business decision" becoming "we believe in justice" — escalation tends to follow within days.
Signal 3
Trajectory Signals
Velocity, acceleration, and narrative mutation over time. Early-stage versus late-stage detection. The difference between a signal that is building and one that has peaked — and what each requires in response.
Signal 4
Linkage Signals
Coordinated phrase reuse, cross-source similarity, content clustering. The difference between isolated criticism and a structurally amplifying, coordinated pattern with momentum behind it.
Signal 5
Actor Signals
Who is originating and amplifying — and whether their involvement is anomalous. New actor classes entering a discourse are often the earliest structural signal of escalation risk. The greenwashing case: two journalists following the same three advocacy groups in the same week.
The science

Peer-reviewed research
confirms what the data showed first.

The signal model is grounded in published academic literature on language as a leading indicator of institutional and social distress.

Nature Human Behaviour · 2018 · Mooijman et al.
Moral language density predicted arrests during civil unrest
Moral tweet frequency predicted the probability of arrests during the 2015 Baltimore protests with statistical significance. Moral framing in public language is measurably predictive of real-world escalation — the empirical basis for Parallax's Discourse Signal class.
PLOS One · 2023
Financial filing language predicted institutional distress
MD&A text readability and linguistic patterns in financial filings predicted financial distress with statistically significant accuracy, ahead of publicly visible symptoms. Language in institutional communications carries forward-looking signal about institutional health.
Enron NLP Studies · Multiple independent
Hubristic language increased approaching collapse
Multiple independent studies found that hubristic language in Enron's executive communications increased systematically approaching collapse, while ethics-related language disappeared well before the scandal became public. Both omission and certainty drift were detectable in the corpus.
Annals of Operations Research · 2025
ECB speech text predicted bank stock movements
ECB speech text analysis showed statistically significant predictive power for bank stock movements. Institutional language at the highest levels carries measurable forward-looking signal — validating NLP-based detection applied to financial and institutional communications at scale.
Parallax found the pattern before the literature did.
The omission anomaly finding was identified independently across ten crisis cases before the academic literature confirming it was found. The pattern appeared in the data first. The science confirmed it. That sequence is significant: it means the finding is empirical, not retrofitted to existing theory.
Market context

$97 billion market.
0 pre-escalation products.

$97B
Crisis management services market — the industry Parallax operates inside
$6.2B
Crisis communication segment — where PR consultancy budgets sit
9%
CAGR — AI-driven early warning identified as the primary growth driver through 2034
Parallax · Narrative Risk Intelligence

Parallax is now
in private beta.

Parallax is live. Request access, log in, or get in touch.

[email protected] [email protected]