Entercard Accelerates Credit Risk Modeling with Anaconda and Snowflake
How a leading Nordic credit card company reduced model development time by 25% and slashed documentation from weeks to days
Entercard stands as one of the Nordic region’s leading credit market companies, serving over 1.7 million customers across Sweden, Norway, Denmark, and Finland. Founded in 2005 as a joint venture between Swedbank and Barclays Principal Investments, this Stockholm-based financial institution employs over 450 professionals representing more than 40 nationalities.
When Nicholas Munford joined Entercard seven years ago, the company was relying on traditional statistical software for their critical credit risk modeling. As the foundation of their lending decisions, these models needed to be both highly accurate and rapidly deployable—but their existing toolchain was becoming a bottleneck.
“We wanted to start using more sophisticated machine learning techniques like gradient boosted decision trees, which are very useful in our industry,” explains Munford, senior decision science analyst who builds the statistical and machine learning models that assess credit worthiness for Entercard’s applicants. “The support for those kinds of techniques in our previous platform was pretty patchy.”
Finding the Right Foundation
Entercard’s decision science team faced a classic enterprise dilemma. While they needed access to cutting-edge open source packages for advanced analytics, their information security team wasn’t comfortable with unrestricted access to open repositories.
“We encountered some initial hesitation from our information security departments,” Munford recalls. “They were happy to give us a Python installation, but understandably cautious about allowing us to install whatever open source packages we wanted to. There were potential security considerations from just installing anything from an open repository with no controls.”
The alternative—seeking approval for packages individually through a formal process—would have significantly slowed down a team that needed to move quickly in the fast-paced credit market.
After evaluating various options, Entercard chose the Anaconda AI Platform specifically for its trusted repository of curated packages that could satisfy their security requirements while providing the flexibility their analysts needed.
“The main reason we went with Anaconda was that it gave us access to this curated package repository that we could implement through our information security process,” Munford explains. “Then analysts could install whatever they wanted from that repository—we didn’t have to seek approval for packages one by one.”
This approach proved both cost-effective and practical. While other providers offered comprehensive cloud-based analytics environments, they came with significantly higher price tags and more complexity than Entercard needed for their focused use case.
Powering Critical Credit Risk Decisions
With their new platform in place, Entercard’s team could focus on building sophisticated credit risk models that determine which loan and credit card applicants are likely to repay their debts. This business-critical analysis directly impacts the company’s profitability and risk exposure.
“When someone applies for a loan or a credit card, there’s a risk that might not be able to repay,” Munford explains. “Our goal is to create compliant models and scores that try and predict as best as possible who’s going to repay their loans and their credit cards so that we can give responsible offers to those applicants and potentially decline the ones where it may not be in anyone’s best interest.”
The team analyzes diverse data sources including credit bureau information, existing debt obligations, income data, and internal payment history. Using techniques from traditional logistic regression to advanced machine learning models like gradient boosted decision trees, they can build predictive scores that feed into both automated decision-making systems and manual review processes.
The business impact has been substantial. These improved models have enabled Entercard to realize material cost savings by more accurately identifying creditworthy customers while reducing defaults and risk exposure. At the same time, they can stay in compliance with their organization’s information security policies.
Transforming Development Efficiency
Beyond enabling better business outcomes, the move to Anaconda delivered immediate operational benefits. Entercard’s team achieved a 25% reduction in overall model development time by implementing modern machine learning techniques using Python packages from Anaconda’s trusted distribution. But the most dramatic improvement came in an unexpected area: compliance and auditing documentation.
Credit risk modeling is heavily regulated, requiring extensive documentation for compliance and audit purposes. Previously, this process was largely manual, involving copying graphs, metrics, and tables into static documents.
“We were able to standardize that process to a greater degree,” Munford notes. “We had a huge reduction in the time it takes to document—from maybe several weeks to a month down to a few days. That’s a really dramatic improvement.”
With the time saved in the documentation process, analysts on the team built custom Python libraries that automatically generate the standard metrics and graphs they produce as part of every model development process. This templated approach eliminated the tedious copy-and-paste work that previously consumed weeks of analyst time.
Scaling with Snowflake Integration
When Entercard decided to modernize their data warehouse infrastructure, they selected Snowflake as their platform of choice. The seamless integration between Anaconda and Snowflake proved to be a perfect fit for their evolving needs.
Working with Snowflake notebooks enabled analysts to pull from the same Anaconda repository as their local installations, ensuring compatibility and eliminating packaging conflicts when deploying Python code to production.
This unified approach addresses a key challenge Entercard previously faced: the gap between development and production environments. While they could build sophisticated models locally, implementing them in their previous data warehouse required time-consuming translation work.
“Before enabling Snowflake, anything that was implemented in production had to be ultimately re-coded. But with Snowflake you can create a stored procedure directly in Python—you can have things all in the same place,” Munford says.
Advanced Analytics for Competitive Advantage
Beyond traditional credit scoring, Entercard is now exploring more sophisticated use cases that leverage machine learning’s ability to model complex interactions and non-linear relationships.
“We’re moving into more advanced analytics on marginal impacts,” Munford explains. “If you want to model the impact of changing someone’s price by 1% or the impact of calling a customer that’s behind on their payment… These sorts of marginal impacts can be done with machine learning models because they’re nonlinear, because they can pick up the interaction of different features.”
This ability to model nuanced scenarios and optimize decisions at the margin offers a meaningful competitive edge in the credit industry, where even small improvements in risk assessment can drive substantial business value—while also enabling more tailored, responsible outcomes for customers.
The Unified Platform Advantage
What sets Anaconda apart for Entercard is its comprehensive approach to the entire AI development lifecycle. Rather than cobbling together multiple tools, they have a single platform that addresses their needs from development through deployment.
“Getting access to those packages, which is not something we’d have done otherwise, quite significantly improved our productivity,” Munford reflects. The combination of trusted distribution, secure governance, and actionable insights has eliminated the fragmentation issues that typically plague AI development in enterprise environments.
Looking Forward
As Entercard continues their migration to Snowflake, they’re positioned to unlock even greater efficiencies. The ability to productionize Python models directly in their data warehouse will eliminate the translation step that previously slowed implementation cycles.
“We’re hoping to be able to move more and more over to the cloud environment,” Munford says, anticipating further reductions in their end-to-end development timeline, providing significant operational efficiencies and enabling their team to focus their time and skills on high-impact analysis.
For Entercard, the Anaconda AI Platform has delivered exactly what they needed: the confidence to innovate with open source at enterprise scale, backed by the security and governance their business requires.
Ready to accelerate your AI initiatives with trusted open source? Discover how the Anaconda AI Platform can transform your data science workflows while maintaining enterprise-grade security and governance. Contact our team for a personalized demonstration.
