Financial Services · Snowflake
Global bank consolidates 14 data marts into a governed Snowflake foundation
60% faster reporting · $22M in legacy-platform retirement
At a glance
Key metrics
Challenge
The situation
Fourteen independently governed data marts — assembled through two decades of M&A — produced conflicting numbers on the same regulatory line items. Finance, risk, and treasury each reconciled manually every close. Cloud hyperscaler bills were rising 22% year over year, and the group CRO had committed to the regulator that a single source of risk-weighted-asset data would be in production within 18 months.
Approach
How we delivered
- 01 Ran a 6-week discovery to classify every downstream report by its true data lineage, not the system it happened to live in.
- 02 Designed a Snowflake-centric target state with Unity-style governance: one account per region, replicated through secure data sharing, policy-based masking for PII and client-confidential columns.
- 03 Stood up a migration factory — pattern-based ETL conversion, automated reconciliation harness, and a "parallel-run" control plane that compared legacy vs. new numbers nightly for 90 days.
- 04 Retired systems in waves. Each wave had a named executive sponsor, a reconciliation threshold, and a hard cutover date. Nothing retired until the parallel-run variance was under 0.01% for ten consecutive business days.
Architecture
Solution architecture
Ingestion through Snowpipe and Fivetran into a bronze layer per source system. Silver layer built with dbt, tested via Great Expectations, with row-level lineage captured in Atlan. Gold layer exposed via Snowflake Secure Views and Cortex Analyst for self-serve. Governance ran on Collibra integrated with Snowflake tag-based masking; sensitive columns were masked at query time based on role + purpose. Observability via Bigeye with Slack alerting into the Apptad managed-services runbook.
Outcomes
Measured results
- 60% reduction in regulatory reporting cycle time — from 11 days to 4.3 days end-to-end.
- $22M annual run-rate saved by retiring four mainframe marts and three on-prem Hadoop clusters.
- 0.007% variance against legacy at final cutover — well under the CRO-committed threshold.
- 4,200+ business users self-serving from a governed semantic layer within 9 months of go-live.
- Audit-ready lineage from regulatory field back to source system, trimming prep time for each examination by an estimated 38%.
Technology
Tech stack
Data platform
- Snowflake (multi-region)
- Snowpipe
- Snowflake Cortex Analyst
Ingestion & transformation
- Fivetran
- dbt Core
- Great Expectations
Governance & catalog
- Collibra
- Atlan (lineage)
- Snowflake tag-based masking
Observability
- Bigeye
- PagerDuty
Orchestration
- Apache Airflow on AWS MWAA
Apptad delivery
- Migration factory accelerator
- Parallel-run reconciliation harness
“Apptad didn't just move our data to Snowflake — they gave us a governance posture we can defend in front of any regulator in the world.”
Related
More outcomes
Chasing a similar outcome?
Tell us your target — cycle time, cost, risk, adoption. We'll walk you through what we've delivered closest to it.