Bank DWH: Build vs Buy

The "build or buy" question for a DWH is one of the most expensive architectural decisions a bank will make. Getting it wrong costs $1–5M and 2–4 years.

When to Buy a Ready-Made Solution

Small and mid-sized banks (up to $1B in assets)

Build costs are hard to justify. Off-the-shelf solutions (Teradata, Snowflake, Oracle DWH) deliver 80% of the functionality you need without the risk.

When there is no internal data team

A DWH built on custom architecture without a team to own it becomes technical debt. If you don't have a data architect and 3–5 engineers — buy.

Tight deadlines

A DWH pilot on a ready-made solution takes 3–6 months. Custom development takes 12–24 months.

When to Build

Large banks with unique requirements

Specific regulatory reporting, complex integration with legacy systems, performance requirements that off-the-shelf solutions can't meet.

Import substitution strategy

Migration to domestic databases (Postgres Pro, ClickHouse) requires custom development.

When data is a core competitive advantage

If analytics and ML are the business core, not a support function.

The Hybrid Approach (What I Recommend Most Often)

1. Operational DWH — a ready-made solution for regulatory reporting

2. Analytical Layer — custom build on ClickHouse/BigQuery for ML

3. Data Catalog — open-source OpenMetadata/DataHub

This approach gives you speed to launch plus flexibility for analytical workloads.

Common Mistakes

• Underestimating ETL costs (usually 40–60% of the project budget)
• Choosing a vendor based on features, not 5-year TCO
• Ignoring data governance from the start
• No internal ownership — the DWH becomes a data dump

I've been involved in dozens of DWH projects. If you'd like an outside perspective on your situation — get in touch.