Data Warehouse & Pipeline Cost Audit

What's included

What's flowing in

• Inventory of every pipeline writing into the warehouse, including GA4 exports, ad platform connectors, CDP syncs, and custom jobs
• Streaming and batch decisions with what each one costs you
• Event volume by source, with the share that has never appeared in a query
• Schema review covering nested fields, unused columns, and tables written but never read
• Duplicate ingestion, which happens more often than anyone expects once two tools sync the same source

Query and storage cost

• Cost breakdown by user, job, and scheduled query
• Partitioning and clustering on your largest tables
• Full table scans running hourly for dashboards nobody opens
• Materialized views and aggregate tables that would replace repeated scans
• Storage class and retention policy per dataset
• Slot or warehouse sizing if you're on a capacity model

PostHog and product data

• Event volume against what you query, since most PostHog bills are made of events nobody looks at
• Session replay sampling, retention, and masking settings
• Warehouse source joins and whether they behave the way you think
• Batch exports and destinations, plus filtering in the pipeline so you stop paying twice for the same rows

Governance

• Access control and who can run expensive queries
• Cost attribution by team or client if you run multi-tenant
• Alerting on cost anomalies
• Documentation of the pipeline end to end

Deliverable

• Cost breakdown showing where the money goes
• Ranked list of changes with estimated monthly savings
• Critical changes implemented

Why it matters

Warehouse bills grow quietly because nothing breaks when they do. A connector stays on for a project that ended, a scheduled query scans a full table every hour for a dashboard nobody opens, and the invoice creeps up until someone in finance asks about it.

Most of the savings are unglamorous. Partition the big tables, kill the exports nobody queries, sample replay properly, and replace the hourly full scans with an aggregate. The work takes a few days and the result lands on a line item your client already sees every month, which makes it one of the easier things to get signed off.