Banco Hipotecario, a leading Argentinian commercial bank, is on a mission to leverage machine learning to deliver new insights and services that will delight customers and create upsell opportunities. With a legacy analytics and data warehousing system that was rigid and complex to scale, they turned to Databricks to unify data science, engineering, and analytics. As a result, they were able to significantly increase customer acquisition and cross-sells while lowering the cost for acquisition, greatly impacting overall customer retention and profitability.
Banco Hipotecario set forth to increase customer acquisition by reducing risk and improving the customer experience. With data analytics and machine learning anchoring their strategy, they hoped to influence a range of use cases from fraud detection and risk analysis to serving product recommendations to drive, up-sell and cross-sell opportunities and forecasting sales.
Banco Hipotecario faced a number of the challenges that often come along with outdated technology and processes: disorganized or inaccurate data; poor cross-team collaboration; the inability to innovate and scale; resource intensive workflows, — the list goes on.
“In order to execute on our data analytics strategy, new technologies were needed in order to improve data engineering and boost data science productivity,” said Daniel Sanchez, Enterprise Data Architect at Banco Hipotecario.”The first steps we took were to move to a cloud-based data lake which led us to Azure Databricks and Delta Lake.”
Banco Hipotecario turned to Databricks to modernize their data warehouse environment, improve cross-team collaboration, and drive data science innovation. Fully managed in Microsoft Azure, they were able to easily and reliably ingest massive volumes of data, spinning up their whole infrastructure in ninety days. With Databricks’ automated cluster management capabilities, they are able to scale clusters on-demand to support large workloads.
Delta Lake has been especially useful in bringing reliability and performance to Banco Hipotecario’s data lake environment. With Delta Lake, they are now able to build reliable and performant ETL pipelines like never before.
Meanwhile, performing SQL analytics on Databricks has helped them to do data exploration, cleansing, and generation of datasets in order to create models, enabling the team to deploy their first model within the first 3 months, and the second model generated was off to the races in just two weeks.
At the same time, Data scientists were finally able to collaborate thanks to interactive notebooks, meaning faster builds, training, and deployment. And MLFlow streamlined the ML lifecycle and removed the overreliance on data engineering.
“Databricks gives our data scientists the means to easily create our own experiments and deploy them to production in weeks, rather than months,” said Miguel Villalba, Head of Data Engineering and Data Science.
Since moving to Databricks, the data team at Banco Hipotecario could not be happier as it has unified them across functions in an integrated fashion.
The results of data unification, and markedly improved collaboration and autonomy cannot be overstated. Since deploying Databricks, Banco Hipotecario has increased their cross-sell into new products by a whopping 90%, while machine learning has reduced the cost of customer acquisition by 35%.
The best part is that we calculate that by using Databricks, we’ve saved around 90% of the cost of having a Spark cluster in our datacenter.”
– Matias J. Stanislavsky, Head of BI and Advanced Analytics, Banco Hipotecario