Cutting reporting time from days to an hour using Amazon Redshift with Deutsche Kreditbank AG (DKB)

For almost two decades, DKB's central raw data warehouse (Rohdatenhaushalt) ran most of the bank's analytical workload. Operational since 2005, it held about 75 percent of all data on the bank's central platform. Each new requirement added another layer of complexity, and the environment had reached its physical limits. Scaling up meant ordering hardware that could take weeks or months to arrive. Yet cost and stability were not the only concerns. The existing foundation was not built for what came next: reusable data products, self-service analytics, and AI. DKB faced a clear choice: keep patching or rebuild while it still had time to do it.

Every day, millions of people check balances, pay with their cards, and manage loans at Deutsche Kreditbank AG (DKB). Behind each of those moments sits data. Roughly 20 core-banking subject areas, from customer and account data to cards and loans, feed regulatory reporting, risk and liquidity management, and the processes that serve customers.

When Sascha Kloss, SVP Business Intelligence at DKB, looks back, the scale is what stands out. "We built a platform on premises over a very long period, with many consumers and many consuming applications," he says. "And what grew historically, we then successfully migrated within 2.5 years." Since 2005, most of that data had flowed through one place: the Rohdatenhaushalt. Because operational and analytical workloads shared the same system, the two competed for resources. A demanding analysis could slow down the daily data load, and the load could in turn delay analytical queries.

The system could no longer keep pace with the bank's growing ambitions and moving the bank's differentiating applications into the cloud was already central to DKB's strategy. The decision was to rebuild, and that became project NEXUS.

DKB was one of the first banks to move a process this central to banking into the cloud, so meeting strict regulatory requirements was non-negotiable. For a German bank of DKB's size, data residency, security, and compliance were decisive. "AWS was simply the quickest to understand the requirements we face in the financial industry and build them into its services, which gave us the confidence to move," says Kloss.

DKB built its Cloud Data Platform on Amazon Redshift Serverless, which runs and scales analytics without the need to manage data warehouse infrastructure. Data lands in Amazon Simple Storage Service (Amazon S3), and the AWS Glue Data Catalog keeps it organized and easy to find. With the serverless model, the team could start quickly and gain experience, bringing data in, step by step. “Redshift scales up and down and makes the resources available when we need them,” says Stephan Richter, Tech Lead for the Cloud Data Platform (CDP) migration at DKB, pointing to the separation of compute and storage, and the ability to stand up the service quickly using infrastructure as code.

Instead of replacing one large system with another, DKB chose a decentralized architecture with a dedicated Amazon Redshift consumer instance per business domain. Each domain accesses the platform data it needs via Amazon Redshift data sharing, without copying it, and can also hold its own data. "Ownership no longer sits with a central team but with the domains and data producers, and that early architecture decision is what made it possible," says Richter. See Figure 1: Cloud Data Platform (CDP) Architecture Overview.

Each team brought its own strengths during the migration. DKB delivered the project together with AWS Professional Services and 23DATA, an AWS Partner that combines deep expertise in banking data architecture and data engineering with hands-on migration experience. 23DATA data architects shaped the target architecture and data model and traced existing data flows to help DKB decide what to migrate, restructure, or retire. Its engineers rebuilt the data pipelines. In the bank steering domain, a dedicated 23DATA team moved every application to the new platform and ran a synchronization process that kept data consistent across the old and new environments, so critical steering and reporting continued running through cutover. AWS contributed a resident architect during the design phase, ran on-site Amazon Redshift deep dive workshops in Berlin, and guided the team on data sharing, access control, query performance, and operational monitoring, followed by an operational readiness review ahead of go-live.

For Florian Lindemann, Domain Lead for the CDP and Fraud Solutions at DKB, the technology was only part of the story; what made the difference was how the change was managed. "The decisive factor and the key to success was creating transparency in order to achieve shared responsibility," he says, so that technical teams and business units always knew what had to switch over, when, and how. Giving domains ownership of their data fit a genuine demand, he adds: "Actually everyone wants to work more with data."

The Finance team shows the change best. Each quarter, it prepares an industry portfolio report that feeds into a broader report for the group's parent company. Preparing the data used to take two to three days. The analysis had eight to ten parts stitched together through intermediate tables, some steps ran for hours, and at year end the year-to-date figures made the work even longer.

Today, the same preparation takes about one hour. Analysts query large volumes of data directly, follow new questions as they come up, and finish the work in one focused session. For Sebastian Sawatzky, Business Analyst in One Finance at DKB, the day-to-day feel is completely different. On the previous system, a query could run for hours, so he would start it, switch to other tasks, and return later, hoping it had finished. This stop-and-start process sometimes fragmented his analytical workflow. Now he stays with the task from start to end. “I can generate a large table with 200 million rows in 3 seconds, make adjustments, drop it and generate it again,” he says. The work has shifted from keeping queries alive toward actual analysis. As he sums it up: “I can start my task and finish it the same day, answer the arising questions with my colleagues, and really concentrate on the functional side.”

The quarterly report is only one example. Across all its analyses, the Finance team estimates it saves more than 20 percent of its time, which analysts now spend on the actual business questions. Behind those results, the platform holds more than 1 petabyte of data, adds about 5 terabytes every day, and serves around 30,000 queries a day from technical users, supporting more than 200 productive applications and over 500 reports and processes across roughly 20 business domains.

DKB is already extending this foundation. The bank plans to migrate its bank steering data warehouse to the cloud on the same model, is building a data marketplace so domains can discover and use data products on their own, and wants to open up self-service analytics in natural language so more business users can answer their own questions. For Kloss, that foundation is the point. “Only a scalable and high-performing data platform gives us the foundation to use AI in our bank at all. The quality-assured data we provide is what our data and AI initiatives build on,” he says.