Star schema vs snowflake schema — which is right for us?
Star schemas (denormalized dimensions) are faster and simpler for BI tools and are our default for analytics marts. Snowflake schemas suit very large dimensions with maintenance-heavy hierarchies. In modern columnar warehouses, the performance gap has narrowed — clarity for analysts usually decides it.
Do dimensional models still matter in the dbt era?
More than ever. dbt makes transformation easy, which makes unmanaged sprawl easy too — we have rescued platforms with 500+ ungoverned models. Kimball-style facts and conformed dimensions give dbt projects a stable backbone that scales past hundreds of models.
How do you handle slowly changing dimensions?
We automate Type-2 SCDs with dbt snapshots or custom merge logic, with effective-dating, current-flag conventions, and late-arriving record handling. Choice of SCD type follows the business question: do users need history as-was, as-is, or both?
Can you fix our existing messy warehouse models?
Yes — model rationalization is a core engagement. For one bank we restructured 560+ dbt models, cutting runtime from 6.5 hours to 87 minutes and total cost by 63%, without breaking a single downstream dashboard.
What deliverables come with a modeling engagement?
Bus matrix, ERDs, grain and SCD decisions log, dbt project with tests and docs, semantic/metrics layer, and a style guide your team can enforce in CI. Knowledge transfer sessions are included by default.