Platform

The infrastructure beneath every decision.

One real-time system for user data: five services that turn raw signal into action, a control plane and a data plane that keep policy and processing separate, and an Activity Graph that understands what a user is trying to do. This page is the canonical definition of how the platform works.

Services

The service spine.

Every pipeline on the platform runs the same five services. Four of them run in order; Sourcing is invoked by Orchestrating when a record needs data that is not yet yours. This is the canonical definition — every other page references these terms rather than redefining them.

  • Ingesting — real-time collection of user actions at the edge. Included, never metered.
  • Graphing — structuring activity and identity into the graph: who the user is, what they did, what they're trying to do. Suppression and dedupe. Included, never metered.
  • Orchestrating — the rules governing a record in flight, including whether to source. Included, never metered.
  • Activating — egress under a use case license, where the result creates value: audiences, messages, system actions. Metered, one charge per use case.
  • Sourcing — acquiring data that is not yet yours. Metered, one charge per query, on the disposition that answers it.
↓ Signals in
Site & app events · CDP / CRM & 1st-party · Inbound calls · Email engagement · Impressions · CTV exposure

Ingesting

Collect at the edge
Cross-device · Cross-channel
Deterministic · Probabilistic
→

Graphing

Resolve & structure
Identity · Activity
Audience · Portfolio
→

Orchestrating

AI / ML engine
Context · Rules · Models
Real-time triggers
→

Activating

Route & act
Audiences · Messages
System actions
Actions out ↓
Audience activation · Lookalike seeding · Frequency & sequencing · Dynamic creative · Channel routing

Ingesting, Graphing and Orchestrating are included. Value is charged on egress and on data; everything that makes them possible is included.

* Ingesting, Graphing, Orchestrating and Activating are the IaaS services. Sourcing is the DaaS service, metered per query on the disposition that answers it. See how each service maps to a billable Usage on the taxonomy.

Taxonomy & token specification
How each service maps to Solution → Product → Feature → Usage
Read the spec
Architecture

Two planes, cleanly separated.

Data planes process events. Control planes control outcomes.

The platform splits into two planes that run independently. The control plane decides; the data plane processes. Between them sits the User Activity Graph — the unified user memory that both planes read and write. Keeping the planes separate is what lets policy, models, and orchestration evolve without touching the fabric that handles every event — and lets each scale on its own.

Directives & Rules (Many Modes — One Platform)
User Control PlaneRealtime Orchestration
  • Governance
  • Context
  • Decisioning
  • Experience
  • Outcome
User Activity GraphUnified User Memory
  • User Action
  • Identity Resolution
  • Data Sourcing
  • System Action
  • Feedback Signal
User Data PlaneRealtime State & Activity Fabric

The two planes scale and evolve independently — new models and rules ship on the control plane without disturbing the data fabric, and the data plane grows with volume without waiting on policy. The Activity Graph is what they share.

Activity Graph

Structured intent, not static segments.

The Activity Graph organizes everything known about a user into five layers — from who they are down to what they just did. Each layer answers a different question, and together they express intent rather than guess at it.

Identity → Vectors → Missions → Sessions → Actions
IdentityWho the user isIdentity
VectorsWhat they care aboutIntent
MissionsWhat they're trying to doIntent
SessionsHow they're interactingBehavior
ActionsWhat they didBehavior

Marketing platforms segment people. The Activity Graph understands what they are trying to do.

Closed loop

Every pass makes the next one smarter.

The platform runs a single loop end to end. Three steps map directly onto the service spine; the last two close the loop by feeding every outcome back into the graph.

Observe
≈ Ingesting
Capture the action the moment it happens.
Decide
≈ Orchestrating
Evaluate context, rules, and models to choose the action.
Act
≈ Activating
Deliver the result where it creates value.
Measure
closes the loop
Attribute the outcome back to the action.
Learn
closes the loop
Write what happened back into the graph.

Every action feeds the graph. The graph improves the next decision.