Star vs Snowflake in Retail Analytics

Context

You’ve joined the data platform team at a large omnichannel retailer (web + 2,000 physical stores) processing ~10M order line items/day. The company is migrating from ad-hoc operational reporting to a governed analytics warehouse that powers executive dashboards (gross margin, promo performance), machine learning features (demand forecasting), and self-serve BI for hundreds of analysts.

A recurring design decision is whether to model the warehouse using a star schema (denormalized dimensions around a central fact table) or a snowflake schema (normalized dimensions broken into multiple related tables). The choice affects query simplicity, storage, data quality, and performance—especially for join-heavy BI queries and for maintaining slowly changing dimensions.

Core Question

Explain the key differences between star schema and snowflake schema in the context of an analytics warehouse.

In your answer, cover:

1. Structural differences: how dimensions are represented (denormalized vs normalized) and how that changes join paths.
2. Querying implications: typical BI query patterns (GROUP BYs, rollups) and how each schema impacts SQL complexity and correctness.
3. Performance trade-offs: join cost, predicate pushdown, clustering/partitioning considerations, and how columnar warehouses (Snowflake/BigQuery/Redshift) may influence the decision.
4. Data governance & maintenance: handling conformed dimensions, slowly changing dimensions (SCD), and avoiding update anomalies.
5. When to choose which: give at least two concrete scenarios where star is a better fit and two where snowflake is a better fit.

Scope Guidance (What the interviewer expects)

Discuss trade-offs rather than claiming one is always superior. Use a realistic example (e.g., fact_sales joined to dim_product, dim_store, dim_date) and explain how the model affects a common metric query like “weekly revenue by brand and region.” Mention how the decision changes with scale (billions of fact rows), team maturity, and downstream tools (semantic layers, dbt models, BI extract engines).