LinguaFrame Analytics™
Language Intelligence

Words are the outer
clothing of ideas.
We read both.

Behind every customer comment is a way of thinking. Standard analysis reads the surface — themes, sentiment, counts. LinguaFrame Analytics™ reads what's underneath: the motivational, cognitive and emotional signals that predict what someone will do next.

See it in action ↓ Explore the architecture →
"Poetry is when an emotion has found its thought and the thought has found words." — Robert Frost  ·  LinguaFrame™ reverses the journey.
Live signal detection
"We've been members for twelve years and I'm starting to wonder if it's still worth it. The recent changes weren't explained properly and we feel like loyal members are being taken for granted."
Traditional analysis: Negative sentiment · Theme: member value · Theme: communication
The proposition
Standard analysis stops
at Layer 2.

Every piece of customer text contains three distinct layers of information. Most tools read one. Better tools read two. LinguaFrame Analytics™ reads all three — and it's the third layer that changes what you'd do next.

Layer 01 · Everyone reads this
The Words
Topics, themes, keyword frequency. What subjects came up. The raw content of what was said.
Output: "Members mentioned value, communication, and loyalty"
Action implied: unclear. Which of these matters most? What do you actually do?
Layer 02 · Some tools read this
The Feeling
Sentiment and emotional intensity. How positive or negative, and how strongly felt. The emotional charge on the content.
Output: "Negative sentiment, medium-high intensity. Value topic most charged."
Better — but still no action. What do you say? To whom? When?
Layer 03 · Only LinguaFrame™
The Thinking
How this person is processing the world right now. Their motivational orientation, cognitive style, current state. The machinery behind the feeling.
Output: "Protection-oriented · Past-anchored · Undecided · Collective identity · High persuadability window"
Action: acknowledge 12-year loyalty specifically, provide concrete rationale, reach out before next billing cycle.
The intelligence layer
Four questions.
Every communication answered.

Before any skilled communicator writes a single word, they need four things answered. LinguaFrame Analytics™ extracts all four from your verbatim data — at any scale.

01 · Motivational
Motivational signals?
Are they chasing an upside or avoiding a downside? Acting by choice or under pressure? This determines the emotional register your message needs to operate in.
motivational orientation internally vs externally motivated
02 · Cognitive
How are they processing?
Are they thinking in the big picture or in operational detail? Rationally or emotionally? Deeply or superficially? This determines the format, length, and content type your message needs.
broad vs detailed thinking rational vs emotional processing Depth of engagement
03 · Perspective
Where do they stand?
Are they owning the experience or pointing outward? Thinking individually or collectively? This determines the tone and attribution framing your message needs to take.
internal vs external attribution individual vs collective framing relational warmth signal
04 · State
What state are they in?
Is this urgent or background noise? Have they already decided? Are they living in a past grievance or anxious about the future? This determines timing and whether any intervention is worth making.
Then / Now / Next decision certainty urgency signal
Interactive demo · Yelp restaurant reviews
Same score.
Completely different story.

Five real restaurant reviews. All rated 3 stars. Traditional analysis calls them all "mixed sentiment." Click each review to see what LinguaFrame Analytics™ reads underneath — and what that means for how you'd respond.

Select a review above to see the LinguaFrame™ intelligence layer.
Structural intelligence · Two steps
From landscape
to architecture.

Once LinguaFrame Analytics™ scores a corpus of verbatims, the analytical work moves through two stages. First, map the terrain — discover which topics cluster, how intensely they're felt, and where they sit relative to each other. Then build the structural model — trace the causal pathways from topics through thinking styles to business outcomes.

Step 01 — Map the landscape
Signal Network
Force-directed · Nodes = topics · Size = intensity · Edges = co-occurrence
How to read it: This is the terrain — the raw landscape of your verbatim data. Larger nodes = topic mentioned with higher emotional intensity. Thicker edges = topics that co-occur. Colour = which signal cluster the topic belongs to. The shape of this landscape tells you where attention and emotion are concentrated.
Step 02 — Build the model
Signal Pathways
Structural model · Topics → Signals → NPS outcome
How to read it: This is the architecture — built from the landscape in Step 01. Topics (left) flow through signal clusters (centre) to the NPS outcome (right). Arrow weight = strength of relationship. This is the text-native structural model: causal pathways from what people said, through how they were thinking, to what they then did.
Higher-dimensional space
Your entire verbatim dataset.
Mapped by how people think.

Each dot is one verbatim, positioned in signal space — not by what words were used, but by the underlying thinking profile. Verbatims cluster naturally by motivational orientation and cognitive state. Hover any cluster to see the thinking profile it represents.

Signal Galaxy — 847 verbatims (Yelp + membership data)
UMAP projection · 2D from 13-dimensional signal space
Move your cursor over a cluster to see the thinking profile and recommended action.
Tipping points in language · PDP/ICE from text
How strongly something
is felt predicts outcome
non-linearly.

Once topics have sentiment intensity scores, LinguaFrame Analytics™ builds partial dependence curves — showing exactly where the intensity of a topic feeling tips into NPS impact. The non-linear thresholds that flat averages completely miss.

Topic
Segment
NPS impact at max intensity
−34 pts
Low intensityMediumHigh intensity
Population average (PDP) Individual members (ICE) Danger zone
Low intensity (0–0.3)
Topic mentioned in passing. Minimal NPS impact. No intervention needed — this is background noise.
Medium intensity (0.3–0.65)
Actively felt but manageable. Linear NPS relationship. Standard improvement actions apply.
High intensity (0.65+) · Danger zone
Non-linear cliff. NPS collapses disproportionately. Immediate, personalised intervention required before next touchpoint.
The composite measure
Involvement isn't
how often they show up.
It's how deeply they're in.

Behavioural involvement — visit frequency, purchase rate, activity level — tells you what people do. The Signal Depth Score tells you how cognitively and emotionally invested they are in what they do. These are completely different signals. And the second one predicts the future better.

73
Signal Depth Score
out of 100
Composite score across four signal clusters. Reflects linguistic investment — how much cognitive and emotional resource went into this communication.
Motivational signal
82
Cognitive depth
71
Perspective clarity
68
State urgency
74
High Signal Depth — cognitively and emotionally invested
High Signal Depth (70–100) · Your most valuable signal
Deeply invested, cognitively engaged, emotionally activated. High Signal Depth members who shift from positive to negative are your most urgent retention risk — and your most persuadable window. High Signal Depth members are your untapped advocates.
Medium Signal Depth (40–70) · Monitor and nurture
Engaged but not deeply invested. Susceptible to drift — positive experiences can move them up, negative experiences can accelerate disengagement. Standard retention approaches apply.
Low Signal Depth (0–40) · Background noise
Cognitively disengaged. Comments are transactional. Not your focus for personalised intervention — but a cohort to watch if Signal Depth begins declining across the segment.
The policy and membership insight: In policy settings, Signal Depth identifies the voices that matter most in a consultation — not the loudest, but the most cognitively invested. In membership organisations, a Signal Depth drop in NPS is an early warning that the most loyal members are approaching a critical decision point.
Segment profiles · Membership organisations
One message to every member.
Or the right message to each one.

Membership organisations communicate to everyone as though they're the same person. But their members have completely different thinking styles. The same renewal campaign — the same words — lands completely differently depending on how each member processes information. LinguaFrame Analytics™ tells you which is which.

70%
Churn risk
82
Signal Depth
74
Persuadability
36
NPS score
Example verbatim — segment theme
"We've been loyal members for over a decade. The recent changes feel like the organisation has forgotten what made it special. I'm not sure renewal is automatic this year."
LinguaFrame™ signals detected
Protection-oriented motivated by protecting what was
Past-anchored "twelve years" · "what made it special"
We-focused collective identity at risk
Undecided "not sure" — window still open
Recommended action
Acknowledge 12-year loyalty. Validate the loss. Provide one concrete reason to stay.
This member is Protection-oriented and undecided — they can be retained. But a generic renewal campaign will feel tone-deaf. They need to feel seen first.
⚡ High persuadability window · Act within 14 days
The key insight
All three members scored the same NPS. Traditional analysis says "run a retention campaign." LinguaFrame Analytics™ says three completely different interventions — and one of them is no intervention at all.
The frontier · Language as geometry
Words are the surface.
The fingerprint is
what's underneath.

Recent advances in AI reveal something remarkable: when a language model reads a sentence, it creates a mathematical fingerprint — 768 numbers — that encodes not just what was said, but how the mind was working when it said it. The motivational orientations and thinking styles that determine how someone will behave leave distinct geometric traces in that fingerprint space.

Two sentences. Different fingerprints.
BERT embedding space · 768 dimensions → 2D projection
Abstract · Opportunity-oriented
"I want a bank I can genuinely trust with my future."
abstract language processing opportunity-seeking orientation future-anchored
Concrete · Protection-oriented
"The app crashed twice this week and they charged me $12 without warning."
concrete language processing protection-seeking orientation present-anchored
These two sentences end up in completely different regions of the fingerprint space — not because a rule was applied, but because the language model learned this distinction from billions of sentences. The orientations and thinking styles that LinguaFrame™ reads are encoded as geometric structure in this space. The dimensions are real — they're discoverable from language itself.
What researchers have found

A 2024 study demonstrated that LLMs encode meaningful differences in how people think — not just what they say — with statistically significant effects on task performance depending on the cognitive mode active in the text. The fingerprint reflects the thinking style, not just the words.

What recent research shows

A 2025 study confirmed that the way people use language reliably predicts their underlying motivational orientation — even when they have no intention of disclosing it. The signal is in the structure of the language, not in what is explicitly said.

The LinguaFrame™ research programme

We are testing whether the LinguaFrame™ signal dimensions are geometrically separable in the fingerprint space of customer verbatims — meaning they exist as real structure in the data, not as categories imposed from outside. If confirmed, the dimensions LinguaFrame™ reads aren't a framework we applied. They're a map the language drew.

That would mean LinguaFrame™ is reading a map that language drew — not one we drew for it.
Get access
Your verbatims are already
telling you how people think.

LinguaFrame Analytics™ is in active development. If you work with open-ended survey data and want to see what the thinking layer reveals — let's talk.

  • Start the conversation → ← Research Foundry