Analytics Warehouse Design & Build for Litebox's Growth Platform

I led the design and implementation of Litebox's first Data Platform. Starting with technical discovery, ran the vendor evaluation, and designed and implemented the resulting stack. The discovery covered warehouse, ingestion, orchestration, transformation, and BI, and produced an initial decision framework the client could interrogate rather than a recommendation they had to trust: a cost model whose assumptions were editable inputs, and options tested against their real constraints, including a cloud-provider decision still ahead of them, which the warehouse choice was deliberately designed not to preempt. The result is a stack sized to what Litebox operates today, with the conditions for scaling it up documented in advance rather than discovered under pressure.

The build is a multi-tenant dimensional model across multiple custom, parametrized PostHog endpoints, Google Search Console API, and Ahrefs for Domain Rating, shipped as a Dataform workflows project and BigQuery datasets. Sources land isolated per client and unify at a single staging boundary, so cross-client comparison (the point of Litebox's product) is a property of the model rather than a query written carefully every single time. Conformed dimensions, a fact layer built on the Litebox's Sr BI Analyst's desired mart framework, and an experiment registry that accommodates new experiments without schema changes mean that the model is able to absorb growth instead of being rebuilt by it. Per-client integration lives in version-controlled configuration, so onboarding a client or reassigning a taxonomy is a config change, not a deploy.