American Golf: Mapping a 35–40% data-platform cost reduction & a defensible target architecture.
We audited American Golf’s data platform — Snowflake, the surrounding ETL, and the vendor-managed pipelines around it — mapped every dependency across their retail estate, and produced an evidence-based recommendation the business now owns: a costed, sequenced plan to bring recurring platform spend under control without disrupting the storefront or the custom-fitting business.
The business problem
Years of layered integrations had left American Golf’s data estate carrying a growing recurring bill, spread across a mix of platforms, pipelines and vendor arrangements — with no single view of what was being paid for or why. Snowflake credit consumption and downstream pipeline costs were climbing month on month, without a clear line of sight into which workloads were driving them. Customer, product and transaction data spanned multiple systems with no agreed source of truth for downstream teams to rely on. And platforms that had already been replaced were still being paid for, with legacy pipelines continuing to run without an active business consumer.
What we delivered
- An independent systems inventory — every platform, integration and job catalogued with owner and purpose, built ground-up from the running code rather than the org chart.
- A full code audit of every application in the estate — 12 front-end, back-end and integration services, each reverse-engineered to trace what it reads, writes and depends on.
- Data-flow mapping across 501 Snowflake base tables, 269 stored procedures and 308 views, with 1.85 billion rows profiled across five schemas — every domain traced to source and consumer.
- A workload-level cost profile that isolates the biggest and most avoidable spend, quantified across five distinct cost levers (non-production warehouse controls, retiring legacy pipelines, incremental CDC ingestion, schema clean-up, and vendor & licence consolidation).
- A defensible target-state architecture with a sequencing plan the business signs off — what to keep, what to retire, what to migrate, and in what order.
Results
35–40% recurring cost reduction identified
Across ingestion, storage, licensing and managed-service arrangements — quantified across five distinct cost levers, each with a proposed implementation approach and a rollback path.
1.85 bn rows profiled
Across 501 Snowflake base tables, 269 stored procedures and 308 views, spanning five schemas — every domain traced end-to-end so the recommendation is defensible under scrutiny.
12 applications audited end-to-end
Every application in the estate reverse-engineered from the running code, not the org chart — front-end, back-end, and integration services alike.
A costed, sequenced target architecture
What to keep, what to retire, what to migrate, and in what order — signed off by the business as the direction of travel for the subsequent build phase.
Fix the data warehouse. Then build AI on top.
Twenty minutes to talk through your Snowflake environment or your AI plans — whichever is holding you up. No pitch, no pressure.