![](https://d15shllkswkct0.cloudfront.net/wp-content/blogs.dir/1/files/2019/07/capitalonedata.jpg)
![](https://d15shllkswkct0.cloudfront.net/wp-content/blogs.dir/1/files/2019/07/capitalonedata.jpg)
From General Electric Corp. and Xerox Corp. to Radioshack Corp., PerkinElmer Inc. and Sears Holding Corp., the history of computing is littered with companies that were very good at using technology but not as good at making and selling it.
Capital One Financial Corp. intends to buck that trend. Cloud co-development, it believes, has altered the landscape.
With last month’s launch of Capital One Software, the financial services giant is aiming to turn the expertise it gained from a four-year-long migration and expansion on Snowflake Inc.’s cloud data warehousing platform into a revenue source.
Equally important is that Capital One is using the opportunity to collaborate with Snowflake on new features while they’re still in the labs, such as the cloud company’s recently announced Native Application Framework. It lets developers deploy applications to the Snowflake Marketplace that customers can run directly in their Snowflake instances without moving data.
In fact, Capital One Software’s introductory video for its first data management product, called Slingshot, cites one of the company’s goals as being to accelerate Snowflake adoption more widely.
Using a cloud data warehouse has been transformative for Capital One, said Salim Syed, the company’s vice president of engineering. “We have eliminated over 55,000 hours of manual changeover and cut the cost of queries by 43%,” he said.
Like many large enterprises, Capital One used a Teradata Corp. data warehouse for years, but the limitations of on-premises infrastructure severely limited who could use the resource and for what purposes.
“We had limitations on storage and concurrency and upgrades were a six-month process,” Syed said. “Teradata had to ship us the hardware and install it in our data center with maybe a week or two of an outage. We’re in a different world with Snowflake. There’s no outage, it’s very transparent and we got cross-region resiliency, which is critical to our business.”
The migration to Snowflake took more than a year and required the firm to run both data warehouses in parallel for a while. But there were collateral benefits, such as an opportunity to rationalize data and find efficiencies. For example, only 6,000 of the 18,000 tables in Teradata needed to be migrated, Syed said.
More important is that moving to the cloud has enabled Capital One to expand the size of its data warehouse 250-fold from 200 terabytes to 50 petabytes with improved predictability and consistency, Syed said. It has also dramatically grown the population of users who can access analytical data.
“Since we migrated, we have onboarded over 450 new use cases because lines of business saw that their costs would be consistent,” he said.
Slingshot was developed internally to help Capital One better manage its own cloud warehouse use. The software includes pre-configured templates, custom workflows and a variety of controls and reports that constantly monitor warehouse performance and serve up intelligence on potential cost savings.
“We had a small team supporting 6,000 to 7,000 users who were running millions of queries a day,” Syed said. “We didn’t want to be a bottleneck. We wanted to empower business people to manage their own infrastructure but with best practices and guardrails built-in.”
While asserting that “we feel we have a very good product,” Syed said there’s plenty more to come. “Slingshot today is limited; it gives you insight into cost drivers and recommendations on improving efficiency, but it’s very command-line oriented,” he said. “We’re excited to build a new [user interface] as the Streamlit piece comes in.”
Streamlit is the set of frameworks for data scientists and engineers that Snowflake picked up with its purchase of Streamlit Inc. earlier this year. They enable the rapid development of analytics applications and machine learning models directly in Snowflake.
Working with software that’s still in the development phase has its pitfalls, said Syed, noting that working with the pre-release Native Application Framework was “like building a plane while taking off. We had to really be close to Snowflake, pushing and pulling and learning new ways of deploying software.”
But the experience has been worth it for a financial services company that’s also determined to be a technology leader. “We’ve been on a decade-long journey of transformation,” Syed said. “It’s challenging to bring new software to the world, but our teams have perfected ways to build it with security, resiliency, performance and scale.”
THANK YOU