Snowflake's advantage is that storage and compute scale independently, so a heavy transformation does not slow down the analysts querying the same tables. Its risk is the same thing: it is trivially easy to leave warehouses running, clone datasets casually, and receive a bill nobody can attribute.
Overview
Warehouses sized to the work
Most Snowflake overspend comes from a small number of habits: oversized warehouses left on, no auto-suspend, queries scanning far more than they need, and clones nobody deletes. We design warehouse tiers against real query patterns, set suspension aggressively, and put resource monitors in place before the first month's bill rather than after it.
What we deliver
What we build in it
The platform is only half the job. The other half is the modelling and the governance around it — which roles can see what, how personal data is masked, and how a table's definition is agreed and versioned.
- Warehouse tiering and auto-suspend policy, with resource monitors
- Layered modelling: raw, staged, modelled, serving
- Role hierarchy and row-level or column-level masking
- Streams and tasks for incremental processing
- Zero-copy cloning used deliberately, with a lifecycle
Why it matters
Why teams pick it
Snowflake suits organisations with mixed workloads and several teams querying the same data at once, because those teams can be given their own compute without copying the data. It is less compelling where the workload is small and steady, or where the estate is already committed to a single cloud's native stack.
Workflow
How we work in it.
- 01
Assessment
- Query and concurrency profile
- Current spend and where it is going
- Data sources, volumes and update patterns
- Governance and compliance obligations
- 02
Account & Warehouse Design
- Database, schema and role hierarchy
- Warehouse tiers matched to workload
- Auto-suspend, auto-resume and resource monitors
- Environment separation
- 03
Modelling & Loading
- Ingestion via Snowpipe or batch, with replay
- Layered models under version control
- Streams and tasks for incremental work
- Tests on the tables that matter
- 04
Operate
- Cost attribution by team and workload
- Query performance review and pruning
- Access review cadence
- Documentation and handover
Related in Warehouse & lakehouse
