Gavin Somers | Reporting Trust Discuss a Contract

Selected production delivery

Analytics engineering case studies

Evidence from dbt, Snowflake, BigQuery, Looker and production reporting systems.

These examples show how I have approached real transformation estates, event-data pipelines, semantic layers and multi-source reporting platforms. They focus on the delivery problem, the work completed and the practical result.

Discuss Similar Work View Services

01 · dbt and Snowflake

Refactoring a legacy transformation estate

At a global lead-generation business, production analytics had accumulated legacy transformation logic. The delivery need was to improve structure and reliability without interrupting reporting used across the organisation.

What I delivered

  • Managed production pipelines in dbt Cloud while using dbt Core for local development on Snowflake.
  • Refactored legacy models into a cleaner, more modular transformation architecture.
  • Strengthened automated testing, Git review and CI/CD release practices.
  • Designed reusable LookML modelling patterns so consistent logic could be consumed through Looker.

Delivery outcome

The resulting estate was easier to review, test and maintain, with less avoidable warehouse compute and cleaner foundations for dashboards and self-service reporting.

  • dbt Cloud
  • dbt Core
  • Snowflake
  • Looker
  • LookML
  • SQL
  • Git
  • CI/CD

02 · Event-data modelling

Turning nested GA4 events into reporting-ready models

GA4 event data arrived in BigQuery as complex nested records, while analysts needed consistent and performant reporting through Snowflake and Looker. Raw event exports were not a useful business-facing model on their own.

What I delivered

  • Transformed complex nested GA4 event data from BigQuery through production dbt pipelines.
  • Established explicit dimensional grains instead of exposing raw event structures directly to reporting users.
  • Built normalised reporting models in Snowflake for downstream analysis.
  • Exposed reusable business fields and metric logic through the Looker and LookML semantic layer.

Delivery outcome

The work converted raw event data into reusable dimensional models that could support high-performance Looker reporting with clearer and more consistent analytical meaning.

  • GA4
  • BigQuery
  • dbt
  • Snowflake
  • Looker
  • LookML
  • Dimensional modelling

03 · Multi-source reporting

Replacing fragmented reporting with an eight-source warehouse

A digital marketing agency relied on fragmented reporting processes across multiple operational and performance-data sources. The objective was to create one scalable reporting foundation rather than continue reconciling disconnected outputs.

What I delivered

  • Integrated eight source systems into a shared cloud reporting architecture.
  • Built the warehouse in BigQuery with dbt-based SQL transformations.
  • Introduced version control, deployment, testing and documentation around the transformation layer.
  • Created the reporting foundation used to power live performance dashboards.

Delivery outcome

Recurring reporting moved onto a unified and maintainable warehouse architecture, reducing dependence on disconnected manual processes and giving dashboards a shared technical foundation.

  • BigQuery
  • dbt
  • SQL
  • Python
  • Git
  • Automated testing
  • Live dashboards

Available for new work

Have a similar delivery problem?

Share the stack, the reporting or modelling constraint, the timescale and what a useful outcome would look like. I can help as a contract, interim or fixed-scope senior analytics engineer.

Discuss an Engagement View Full Experience