Skip to main content

Driving customer retention with data and machine learning


Increase in inquiries resolved for some customers


Improved customer satisfaction

directly hh header image color
SOLUTION: Customer support

"Databricks has helped Directly democratize data across the company, empowering our data teams to create state-of-the-art machine learning and data pipelines to automate and engage with our clients and users."

— Sinan Ozdemir, Director of Data Science, Directly

Directly’s AI solutions help customer support leaders resolve issues with the right mix of automation and human interaction, but they struggled to ensure data reliability and accuracy for downstream analytics. With Databricks as a unified platform for data and AI, Directly has been able to build a real-time data pipeline that scales, fostered collaboration across its data teams, and streamlined the ML lifecycle to ensure models are optimized for performance. Today, everyone is happier, with the insights that help resolve customer issues faster than ever before.

Ensuring the latest data is accessible 24×7

Customer service teams are critical to ensuring the satisfaction and success of a customer. But as any customer support professional knows, that’s easier said than done. In order to help solve end-users’ inquiries while providing valuable insights, Directly’s data engineers needed on-demand access to the latest customer data available. Unfortunately, infrastructure complexity made it difficult to access the data required to answer customer questions. It was difficult to maintain, and in turn difficult to ensure they were able to build pipelines that could scale to meet their needs. The team was forced to work in cycles of just a few hours in order to access and use data for ML, and these access limitations were topped with the scalability issues that come along with using single laptops and data sampling.

“I’m always thinking about whether we are optimizing our AI and machine learning to not only have the best model performance but also perform optimally for our customers’ KPIs,” said Sinan Ozdemir, Director of Data Science at Directly. “We needed a better solution to democratize data and analytics across the organization.”

In their search for a new solution, Directly prioritized the ability to ensure data access to both technical and non-technical end-users. They want to provide various business teams (operations, network operations, sales, marketing, etc.) with important insights to show exactly how the efforts of their data teams are impacting the business. Databricks met all of their requirements with a unified approach that democratized data across their organization while increasing cross-team productivity.

Unlocking real-time data for downstream BI and ML at scale

Since implementing Databricks, Delta Lake has become Directly’s primary lake for production. Their infrastructure, which is now running 24/7 on AWS, enables Directly to stream real-time data, which is then used to power both business intelligence dashboards through Looker for their sales and marketing teams to better understand their customers; and improve data science productivity which has accelerated machine learning innovation.

“Delta lake has become the source of truth and research for several of our teams,” added Sinan. “With accurate data at our fingertips, we can now make smarter decisions that benefit our customers.”

With unification in place, Directly has been able to democratize data, making it available for whoever might need it, in a format they can understand. “Databricks is providing a singular, unified source of data that has multiple access points for engineering and non-engineering teams alike,” explained Sinan.

Their data science team is also able to now leverage distributed computing to explore data at scale and train models at a level of pace and accuracy that wasn’t possible before. Today, they are able to collaborate and train models within interactive notebooks, and use MLflow to track and monitor the life cycles of all their machine learning pipelines.

Faster pipelines, faster model training, happier customers

After moving to the Databricks Data Intelligence Platform and Delta Lake, Directly has been able to build real-time data pipelines that are both reliable, accurate, and fast. Now they are able to ingest massive volumes of streaming data in seconds rather than hours. This has led to faster business insights for various non-engineering teams and innovation for the data science team.

And their customers have noticed. Directly’s ability to resolve customer inquiries has shot up by 80% in certain areas, reflected in a 20% improvement of their customer satisfaction score in these areas.

Looking ahead, Databricks will continue to serve as their core data analytics platform as they continue to grow and expand their endeavors into machine-learning-based fraud detection, conversational AI, RPA, and robotic process automation.

“Directly is focused on helping customer service organizations move into the era of intelligent automation through the ML-powered interpretation of dynamic customer needs in real time,” concluded Sinan. “This critical feedback loop is and will continue to be powered almost entirely by Databricks.”