Delivering Snowflake migration Programme for a major UK bank

The client, a major UK Bank undertaking a significant asset sale in support of their strategic market goals, reached out to the Version 1 team to assist with the transition from a data analytics perspective. This included data gathering, transformation, consolidation, and segmentations.   
  
In addition, the Version 1 team also assisted with the overarching bank-wide technology transformation initiative – to modernise the bank’s data analytics platform as part of an enterprise move to Snowflake.  View the Snowflake Services Select partner page.

There were two main challenges in this project. Firstly, winding down the business in one of the bank’s locations, and secondly, seamlessly integrating the migration with the bank’s wider technology transformation programme.  

Stats from the business transformation perspective include:  

• 1.4 million customer and internal accounts  
• €42 Billion total balances (absolute assets + absolute liabilities)  
• 600+ customer and product segmentations
   

Stats from the technology transformation perspective include:  

• 100+ Data ETL (Extract-Transform-Load) scripts to rewrite  
• 17 account-holding sources  
• Wide range of tools and platforms encompassing Cloudera Hadoop, Impala, Teradata, Oracle, Snowflake, AWS (Amazon Web Services), Python, SAS, SQL, and more
   

Delivering enterprise technology migration plans meant that the Version 1 team were also called upon to maintain the delivery of operational and regulatory requirements through the modernisation of the Bank’s data and analytic landscape.  

This progression towards cloud-based architecture, leveraging the power of Snowflake to support analytic and reporting workloads, gave the Version 1 team on the ground the opportunity to demonstrate their ability to deliver meaningful technological change. They were able to do this without disrupting key business processes and politically sensitive programme objectives.  

Solution

Despite the challenges, the Version 1 team delivered two major milestones for the client at pace, within the agreed timeframe:

• The Perimeter Reporting Framework (PRF): this was a new data and reporting process that supports the client’s wind-down activities. It encompasses the end-to-end data extraction (from 17 source systems), transformation and segmentation (using business logic), consolidation (via standardise table structure and schema), and reporting (for ease of downstream consumption). This new process provided the client with (a) the ability to obtain the required in-scope population at suitable granularities for downstream planning and (b) actions for the business wind-down activities.
• The Snowflake Migration Programme: this was part of the wider technology transformation programme to modernise analytics infrastructure. Version 1 successfully developed, tested, and migrated the new Perimeter Reporting Framework (PRF) process from the original legacy analytics infrastructure (Cloudera Hadoop, Teradata, SAS) to the bank’s strategic infrastructure (Snowflake, AWS, Python). This enabled business continuity as the new process continued to function regardless of the decommissioning of legacy infrastructure. In addition, the newer Snowflake architecture offered a more streamlined analytics experience for the client.

Version 1 was able to achieve two milestones – solving the business and technology transformation programme, as well as winding down the bank’s presence in a different country – due to several factors: 

Informed decision-making supporting by relationship building

Our teams built a positive relationship with the client by setting realistic expectations whilst maintaining a flexible approach to problem-solving . Version 1 ensured that the client’s needs and contexts were sufficiently documented, reviewed, and signed off – prior to the initiation of extensive downstream development work. 

Continual learning, rapid prototyping, and creating a feedback loop

Technology moves at pace and, at Version 1, the team are encouraged to learn new technologies and obtain any required training and certification as needed. As Snowflake was (at the time) a relatively modern technology domain for the project team and on-site engineers, they quickly adapted by completing training courses (such as the Snowflake University workshop training series), as well as certifications (such as Snowflake SnowPro Core Certification).  
 

Version 1 built an initial POC (proof-of-concept) solution – often in its most basic form for longer-term enhancements, such as automation. The flexible and pragmatic approach enabled the team to make incremental progress at pace, obtain feedback from the client quickly, and more importantly earn trust.  

 Data quality, testing, and assurance  

Instead of decommissioning the old process and migrating to the new Snowflake process overnight in one step, the Version 1 team took the more robust multi-step approach – running both the new and old processes in parallel for a period (of at least 6 months) before the full migration. 
 
This provided a window for the team to compare results from both old and new processes, identify bugs and potential improvement opportunities, and implement fixes and enhancements.   

Outcomes

• Delivered a strategic migration to the new Snowflake / S3 solution.  
• Decommissioned legacy SAS, Teradata, and Cloudera Hadoop  
• Migrated raw data to target the Snowflake environment.  
• Empowered the customer by closing the customer’s knowledge gaps around migrating raw data into Snowflake  
• Built a framework to pull data from various sources to import and load into Snowflake, accounting for data differences and complexities  
• Completed extensive documentation around the framework allowing it to be deployed and used by other departments bank-wide  
• Removed manual intervention from the bank’s data engineering teams, saving cost and time  
• Delivered documentation and training to the bank, empowering the client to become self-sufficient. The Version 1 team also successfully trained the client team on Snowflake fundamentals and provided supplementary template scripts for additional experimentation and learning