01
First-party collection
Our own beacon and collector. No third-party analytics tag, so the data belongs to the operator and the privacy position is simple to state.
Data and analytics
Measurement built as a product, not as a report.
A business that cannot measure its own traffic cannot price anything. So the analytics layer is engineered, versioned and licensed like any other system we build.
- Collection
- First party, no third-party tag
- Store
- Columnar, designed for billions of rows
- Attribution
- Resolved server side at landing
- Access
- Multi-tenant with enforced isolation
01 / Pipeline
From event to answer.
Each stage is separable and independently licensable. An operator can take the collector alone, or the whole path.
Reporting reads pre-aggregated rollups rather than raw events. That is what keeps a dashboard responsive when the raw table holds billions of rows, and it is a design decision rather than an optimisation added later.
L5Reporting interfacesMulti-tenant · RBAC
L4AggregationRollups · materialised views
L3AttributionSessions · windows · dedupe
L2Columnar storePartitioned · TTL
L1CollectionBeacon · collector · spool
02 / Capability
What the layer does.
02
Columnar analytics
A columnar store with partitioning, retention and materialised aggregates, sized for continuous high-volume ingest.
03
Server-side attribution
Source, campaign and click identifiers resolved at the landing event and carried through the session, rather than inferred afterwards.
04
Deduplication
Configurable windows so a repeat visit inside a defined period counts once, with the rule written down rather than implied.
05
Reconciliation
Two independent counts shown side by side with the capture rate between them, which turns a commercial dispute into a data question.
06
Multi-tenant reporting
Role-based access with tenant isolation enforced in the application and again by row-level security in the database.
03 / Data minimisation
Designed to hold as little as possible.
The systems are built to answer commercial questions about traffic and monetisation, not to profile individuals. Events are recorded against a session identifier rather than an identified person, and reporting is aggregate.
Full network addresses are not retained beyond the point where they are operationally needed. Geography and device are resolved at collection time and only the resolved classification is stored. Where an identifier is not required to answer the question, it is not collected.
This is an engineering position as much as a legal one: a system that holds less is cheaper to run, simpler to license and easier to defend.