Approach
A shared, software-readable description of the objects your organisation runs on, the actions taken against them and the models that improve those actions. Every MetaTeam division builds against it.
Why a model
Three layers
Semantic layer
Objects, their properties and the links between them, unified from every source into a single resolved representation with history. Customers, accounts, assets, vehicles, orders, campaigns, content, segments.
- Object types and properties
- Links and cardinality
- Identity resolution
- History and lineage
Kinetic layer
The actions that change object state and the systems that carry them out: applications, integrations, autonomous processes, campaigns and lifecycle programmes. Every action is defined, permissioned and logged.
- Actions and permissions
- Workflows and processes
- Integrations and APIs
- Audit and write-back
Adaptive layer
Models, measurement and experimentation that read outcomes and improve the next decision: prediction, optimisation, attribution, anomaly detection, with monitoring so improvement is verified rather than assumed.
- Predictive and generative models
- Metrics and attribution
- Experiments and holdouts
- Monitoring and feedback
Foundation layer
Beneath the three layers sits the infrastructure that hosts them: public, private, hybrid or on-premises environments and the pipelines that move data between them. The model is environment-agnostic; the foundation is chosen to meet your residency, latency and cost constraints.
Anatomy of an object
- PropertiesThe attributes that describe it, with source, type and freshness known.
- LinksIts relationships to other objects, so context is one traversal away.
- ActionsWhat can be done to it, by whom, through which system.
- SignalsThe measures and model outputs attached to it, updated as events arrive.
Divisions in the model
- 01 · IntelligencePolarisPopulates the semantic layer at scale and owns most of the adaptive layer: models, measurement and autonomous decision logic.
- 02 · Software EngineeringOrbitBuilds the kinetic layer: applications, integrations and platforms that carry actions to real systems and write results back.
- 03 · InfrastructureTerraProvides the foundation layer and the pipelines that connect it to the semantic layer, wherever the data must live.
- 04 · DesignLumenShapes how people meet the kinetic layer, with research that keeps the model honest about how work is actually done.
- 05 · GrowthVectorExtends the model to the market: brand and audience objects, campaign actions and attribution signals.
Method
Decision inventory
- 01
Decision inventory
We start with the decisions that matter most and work backwards to the objects, actions and signals they need. Weeks, not months.
- 02
Object definition
Workshops with domain owners define each object once, with source systems, properties, links and permitted actions written down.
- 03
Implementation
Pipelines, services, interfaces and models are built against the definitions. The model itself lives in version control alongside the code.
- 04
Stewardship
A named owner for each object, a change process and monitoring keep the model accurate as the organisation changes.
Principles
- 01
Define once, use everywhere
An object has one definition. Systems adapt to it, not the reverse.
- 02
Every action is explainable
Automated or manual, each state change records who, what, why and through which system.
- 03
Measurement is part of delivery
Nothing ships without the signals that show whether it is working.
- 04
Your data stays yours
Models and measurement run inside your environment on your data, independent of any platform's reporting.
- 05
Smallest sufficient system
We choose the simplest architecture that meets the requirement, then earn complexity with evidence.
Contact
A decision inventory takes two weeks and produces a first draft of the model.
info@metateam.devReply within two working days.