What's an LQM?
A model that reads
behavior, not text.
A large quantitative model learns from structured signals about what happened in the world — where, when, in what order — and returns what is likely to happen next.
The short version
Language models learn from what was written. Quantitative models learn from what was done.
Both find relationships. The difference is the material. A language model relates words to other words. A quantitative model relates events to other events — a visit, a purchase, a session, a place, a time — across a very large number of them.
- Input. Structured, de-identified signals rather than documents.
- Structure. Sequence and proximity carry meaning, not grammar.
- Output. A probable next action with a confidence value attached.
Large Language Model
LLM
Finds relationships across words and documents.
- Generates and summarizes content
- Answers questions about text
- Optimized for language
Large Quantitative Model
LQM
Finds relationships across structured signals at scale.
- Learns patterns in real-world behavior
- Predicts what is likely to happen next
- Optimized for quantitative signals
- Signal Interchange
- Privacy-first by design
- Owned GPU infrastructure
- Living Behavioral Model
How an LQM learns
From signal to probable next action.
-
01
Inputs
Permissioned, de-identified signals: web and CTV activity, transactions, and real-world places, each contributed under agreement.
-
02
Relationships and sequence
The model learns which events tend to follow which — and how strongly — across an enormous number of anonymous sequences.
-
03
Outputs
A prediction with a confidence value: audiences, propensity, timing, or an answer to a question you asked directly.
Where an LQM helps
Questions worth pointing it at.
01
Who is about to need this?
Propensity built from behavior in the world, not from declared interest.
02
When does demand arrive?
Timing patterns that show up in sequence before they show up in sales.
03
Where should we be?
Relationships between places, visits and outcomes across mapped locations.
04
What is this audience actually like?
Behavioral texture instead of a demographic label.
05
What changed?
Shifts in relationships surface before they show up in a reporting cycle.
06
What is this worth?
A confidence value you can weigh against the cost of being wrong.
Common questions
The parts people ask about twice.
No. They do different work and are often used together — a language model to express an answer, a quantitative model to determine it. Intuizi builds the quantitative side.
Signals are permissioned and de-identified before they enter the system. The model learns relationships between anonymous events; it does not hold personal identifiers.
The Living Behavioral Model is the structure an LQM reads. Signals become relationships between anonymous events, and the model learns the patterns those relationships form over time.
A probability attached to a specific outcome for an anonymous cohort — for example, likelihood of a category action within a window — not a certainty and not a person.
No. Engagements are scoped so the model works alongside your existing stack. What that looks like in practice depends on your environment, so we walk through it in the demo.
Still abstract? Bring a real question.
The fastest way to understand an LQM is to watch one answer something you care about.
Ready to unlock the value of quantitative AI?
Bring one question your current data cannot answer. We'll show you what an LQM can return.
Get a DemoPrefer to read first? Start with what an LQM is.