Platform Convert & Engage Data Platform

Your first-party data, finally working as one.

CRM, CDP, loyalty, commerce and warehouse data unified into a living context graph of customers, products and outcomes. It refreshes on every run, and it stays on your infrastructure.

  • Bring your own Customer 360
  • Warehouse-native, zero-copy where possible
  • Consent checked on every use
Cubes and cylinders connected by tubes to one glowing sphere at the centre
Graph refreshed14 sources · 38 seconds ago

Audience freshness

  1. Cart abandoners2 min
  2. Lapsing high-value5 min
  3. New to category1 hr
Revenue per customer Every decision made on the full picture
Audience freshness Minutes from event to usable audience
Wasted spend Existing customers suppressed from acquisition
Data requests Marketers self-serve, engineers stop fielding tickets

Why change

You don't have a data problem. You have a context problem.

The data exists. It is spread across systems that each know one thing about the customer and nothing about why it matters.

History is fragmented

Paid, commerce, care and loyalty each work from a partial view of the same person.

Audiences go stale

By the time the segment refreshes, the moment that made it valuable has passed.

Every tool wants its own copy

More pipelines, more cost, more places for the numbers to disagree.

Consent is a spreadsheet

Permitted use, lineage and retention are documented but not enforced.

What it does

Connect once. Useful everywhere.

A data layer built for agents and marketers alike: governed enough for the CDO, fast enough for real-time decisions, simple enough to query in a sentence.

01 The living context graph

More than a profile table. A model of your business.

Customers, households, products, offers, channels and outcomes are held as connected entities. Every run of every agent reads from it and writes back what happened, so context keeps improving instead of drifting.

  • Entities and relationships, not just rows: who bought what, after seeing which offer, through which channel.
  • Computed traits for lifecycle, propensity and value, refreshed as events arrive.
  • A semantic layer so "active customer" means one thing to every team and every agent.
Context graph · healthLive
8.4Mcustomer nodes
212traits computed
38ssince last refresh
Customer → Household → DevicesIdentity resolved with evidenceFresh
Product → Category → Margin bandSynced from catalogue and financeFresh
Decision → Exposure → OutcomeWritten back by every runFresh
Campaign → Creative → CostAd platform sync delayed 40 minLagging
Lineage available for every fieldIllustrative

02 Pipelines and quality

Data you can trust, because something is watching it.

Connectors for web, app, CRM, commerce, loyalty and warehouse feed a standard event contract. Agents watch schema, freshness and volume, explain what broke, and hold bad data before it reaches a decision.

  • Warehouse-native: query in place and avoid copies wherever the source allows.
  • Data contracts with freshness, provenance and SLAs per source.
  • Batch, streaming and reverse ETL, on one set of definitions.
Sources1 needs attention
SourceModeFreshnessQuality
Commerce ordersStream4 s99.8%
CRM contactsZero-copyLive99.1%
Loyalty ledgerStream9 s100%
Store POSBatch26 hLate
Warehouse martsZero-copyLive98.7%
Store POS: file missing since yesterday · owner notifiedIllustrative

03 Audiences and activation

Describe the audience. Get it built, sized and delivered.

Marketers build behavioural, lifecycle and predictive audiences by clicking or by asking. Each audience is evaluated continuously and published to the channels and tools that need it, with consent applied on the way out.

  • Plain-language audience builder, with the generated logic shown for review.
  • Profile API, audience API, webhooks, for anything that needs to read in real time.
  • Clean-room collaboration, to match with partners without handing over raw data.
New audienceDraft

High-value customers who haven't bought in 60 days but opened the app this week

Built: top 20% lifetime value, last order over 60 days ago, app session in last 7 days, marketing consent = yes.41,280 people · 3,912 excluded for consent

MessagingEvent-driven syncReady
Ad platforms · suppression listHashed, dailyReady
DecisioningReal-time lookupReady
Evaluated every 5 minutesIllustrative

Agents at work

A data team that watches every pipeline.

Data agents handle the unglamorous work: mapping schemas, watching freshness, explaining anomalies and answering questions. Anything that changes what a customer is, or who can see them, stays with a person.

  • Schema mapping suggested, reviewed by your data owner before it is applied.
  • Anomalies explained, with the likely cause and affected audiences listed.
  • Ask in plain language, and get an answer with the query and the sources attached.
Agent activity  Live
Store POS batch missingExpected 02:00 · 14 audiences affectedDetected
Affected audiences frozen at last good stateDecisions continue on yesterday's POS dataAutomated
Owner paged with cause and impactSFTP credential expiredAutomated
New field mapping proposed: "loyalty_tier_v2"Replaces deprecated field in 6 traitsNeeds approval
Export of raw emails to partner toolPurpose not covered by consentBlocked
Every step logged with inputs, reason and outcomeIllustrative

You decide what the agents may do here.

AutomatedKeeping data healthy

Ingestion checks, freshness monitoring, trait computation, anomaly triage and documentation.

Human-approvedChanging meaning or access

New sources, schema changes, new traits used in decisions and any new export destination.

RestrictedHard stops

Use beyond consented purpose, raw personal data leaving your environment and retention rules being bypassed.

What's inside

The data layer an AI-native stack needs.

Use ours out of the box, or put the same interface in front of the CDP and warehouse you already own.

Source connectors

Web, app, CRM, commerce, loyalty, service and warehouse.

Standard event contract

Schema mapping and validation at the point of ingestion.

Context graph

Customers, products, offers and outcomes as connected entities.

Computed traits

Lifecycle, propensity and value, in batch and real time.

Audience builder

Behavioural, lifecycle and predictive, by click or by sentence.

Real-time evaluation

Batch, event-driven and near-real-time on the same logic.

Activation

Profile and audience APIs, webhooks and reverse ETL.

Warehouse-native

Federated queries and zero-copy access where the source supports it.

Clean rooms

Privacy-safe matching and measurement with partners.

Lineage and quality scorecards

Field-level provenance and freshness for every source.

Consent and retention

Purpose, suppression and deletion enforced, not documented.

Agent access

A permissioned MCP server so your own assistants can query the graph safely.

How it runs

From raw events to governed activation.

Each step is a named, logged stage. Nothing autonomous is switched on until its data, permissions, monitoring and rollback have been checked.

  1. Connect consented sources
  2. Validate schema and quality
  3. Resolve identity safely
  4. Build the context graph
  5. Compute traits and audiences
  6. Activate through governed APIs
  7. Monitor freshness and match

How you'll measure it

Measured like infrastructure, valued like revenue.

We agree the numerator, denominator, cohort and success gate with you before anything goes live, and compare against a baseline or a holdout, never against a hunch.

  • Audience freshness (p95)Time from an event happening to the audience reflecting it.
  • Profile completenessRequired attributes present, divided by required attributes audited, for the scoped use case.
  • Activation success rateRecords accepted by the destination, divided by eligible records sent.
  • Consent propagationSends consistent with the latest valid consent. The only acceptable figure is all of them.
  • Data reconciliationSource totals matched to graph totals for orders, revenue and members.
  • Time to new audienceFrom a marketer's request to an audience live in a channel.

Questions

Things people ask before they start.

We already have a CDP. Do we need this?

Keep it. A standard adapter puts your existing Customer 360 behind the same interface the rest of the platform uses, so nothing is migrated. Use our data platform only where you have a gap.

Where does the data physically live?

In your cloud or data centre. The platform is deployed inside your environment and reads warehouse data in place where it can, rather than copying it out.

What makes a context graph different from a customer table?

A table tells you who the customer is. A graph also holds how things relate: which offer they saw, what they bought next, what it cost and what margin it made. That is what agents need to make a good decision.

How is consent enforced?

Each attribute and audience carries its permitted purposes. Every read by an agent, a campaign or an export is checked against them, and withdrawals propagate to downstream systems.

Composable by design

Useful alone. Sharper when connected.

The data platform is the centre everything else reads from and writes back to.

Keeps your existing stack

Reads from the warehouse, CDP, CRM and commerce systems you run today, through one standard adapter.

Runs where your data lives

Small language models trained on marketing data are deployed inside your cloud or data centre, so customer data and the models reasoning over it stay under your control.

Getting started

Start with one outcome, two sources and one destination.

Choose a lifecycle outcome, connect the two sources it depends on, build two audiences and activate them in one channel.

  1. Choose sources, define contracts and consent
  2. Ingest, link profiles and check quality
  3. Build two audiences, activate in one channel
  4. Add sources and set production SLAs