Warehouse Cost Baseline (Onboarding)

Warehouse Cost Baseline (Onboarding)

1. 01

   Pull 90 days of warehouse billing and attribute spend to client, pipeline, and query source

   Snowflake, BigQuery, Databricks, and Redshift all expose account-level usage views; split storage from compute before quoting any retainer line.

2. 02

   Rank the top 20 queries or jobs by credit burn and flag any that run on a schedule nobody owns

   Orphaned dbt models and stale Airflow DAGs are the usual culprits; a single unpartitioned full-table scan can exceed a month of expected spend.

3. 03

   Record ingestion volume per connector so the client sees what each source actually costs to move

   Fivetran and Hevo Data bill on monthly active rows, so a connector syncing a 40-column table for two useful fields is a visible, fixable line item.

4. 04

   Document the transformation layer's materialization choices and their refresh cadence

   In dbt, a table rebuild every 15 minutes costs far more than an incremental model on the same data; capture the setting, not just the model name.

5. 05

   Set a monthly spend ceiling per client with an alert threshold at 80 percent

   Put the ceiling in the statement of work so overages trigger a conversation rather than a surprise invoice at renewal.

6. 06

   Write the baseline into a one-page cost sheet the client signs before pipeline work starts

   Include the assumed refresh frequency and row growth rate; both are the variables most likely to move the number in quarter two.

7. 07

   Schedule a 30-day recheck against the baseline and log variance reasons

   Variance without a written cause becomes an argument; variance with a cause becomes a scoping decision.