Skip to content
Linear Horizon

05 / Products

Query/Agentic Product Analytics

See what users do. Understand what to improve.

Query is a privacy-conscious analytics system for products and websites. It measures how people actually behave, and lets an agent investigate why the numbers moved — through the same tools and evidence a good analyst would use.

Measurement

Events become metrics, and a question turns those metrics into an investigation.

Analytics calculate. Agents investigate.

Query is product and website analytics built for investigation, not just reporting. It instruments the events that matter, calculates the metrics you rely on — visitors, sessions, activation, retention — and keeps them in a model an agent can question.

Analytics calculate. Agents investigate. A dashboard can tell you activation fell; it will not tell you where, for whom, or what changed just before. Query exposes its metrics through a query interface and an MCP server, so an agent can pursue that question step by step.

It is designed to be privacy-conscious by default: measure behaviour in aggregate, keep the data you actually need, and avoid the invasive tracking that most analytics quietly assume.

A dashboard is a fixed set of questions someone decided to ask in advance. Query keeps the same measures, but leaves the follow-up open: when a number moves, an agent can pursue it through the data rather than waiting for a new chart to be built.

From events to an answer

Instrument once. Measure continuously. Investigate on demand.

Query collects product and website events, calculates the metrics your team relies on, and exposes them through a query interface and an MCP server. The metrics answer “what”. The agent is how you get to “why”.

  1. 01

    Instrument

  2. 02

    Ingest events

  3. 03

    Calculate metrics

  4. 04

    Ask a question

  5. 05

    Investigate

  6. 06

    Explain

What Query measures and investigates

  1. 01

    Privacy-conscious tracking

    Measure behaviour in aggregate with a light instrumentation footprint, rather than assuming invasive per-person tracking.

  2. 02

    Product and web analytics

    Visitors, sessions, page views, activation and retention — the core measures teams actually make decisions on.

  3. 03

    Event ingestion

    A write key and ingest pipeline collect product and website events into a single, queryable model.

  4. 04

    Funnels and retention

    Follow how users move from first visit to activation and return, and see where the drop-off actually happens.

  5. 05

    Segmentation and cohorts

    Separate the groups that moved from the averages that hide them — by source, cohort, behaviour or period.

  6. 06

    Agentic investigation

    Expose the metrics through a query interface and MCP server so an agent can test hypotheses about what changed.

Use cases

Where Query is used.

  1. 01

    Product analytics

    The problem

    Product teams can see that activation or engagement moved, but the dashboard does not run the follow-up: which step, which cohort, which change.

    How it is addressed

    Query measures the funnel and lets an agent investigate the movement — comparing cohorts and periods until there is an explanation worth acting on.

  2. 02

    Website analytics

    The problem

    Marketing and web teams want to understand traffic and conversion without deploying invasive tracking or exporting visitors to a third party.

    How it is addressed

    Query instruments the site with a light footprint, calculates the standard measures, and keeps the data in a model you control.

  3. 03

    Activation and retention

    The problem

    A drop in activation or retention has several plausible causes, and reading tiles one at a time rarely isolates the one that matters.

    How it is addressed

    Ask Query what changed before the decline; review the segments and periods the agent compared and the reading the evidence supports.

  4. 04

    Growth experiments

    The problem

    Teams ship changes and then argue about whether the numbers moved because of the change or because of everything else.

    How it is addressed

    Query holds the before-and-after measures in one place, so the investigation is grounded in the same events rather than competing exports.

Limits worth being clear about

Measurement and investigation, not a verdict on intent.

  1. 01

    Query measures behaviour; it does not read intent. A metric tells you what happened, not why someone did it.

  2. 02

    Privacy-conscious means deliberate collection. Query is not designed to reconstruct detailed individual profiles.

  3. 03

    An investigation is grounded in the events you instrument. Gaps in instrumentation are gaps in the evidence.

  4. 04

    Analytics support judgement. Deciding what to build or change remains a human decision.

Query

Have a product where the dashboard shows the what, but never the why?

Enterprise deployment

These products are delivered as engagements rather than self-serve licences. We start with your problem, your data, and the systems you already run.