Product descriptions:
Insulet Corporation is a medical device company specializing in innovative insulin delivery solutions. Their flagship product, the Omnipod® Insulin Management System (Omnipod), offers a wearable insulin pump to simplify life for individuals with insulin-dependent diabetes. Internally, the company had major roadblocks with siloed data, slow processing times and limited real-time insights. This made it difficult to monitor Omnipod’s performance, optimize manufacturing workflows and ensure regulatory compliance. Additionally, outdated infrastructure, manual ETL processes and fragmented systems slowed data access, prevented cross-functional collaboration and forced reliance on costly third-party tools. By adopting Databricks, the brand achieved 12x faster data processing while reducing total cost of ownership by 97%.
Experiencing data bottlenecks that stalled critical decisions
Insulet Corporation’s mission is to improve the lives of people with diabetes, enabling its customers to enjoy simplicity, freedom and healthier lives through innovative technology. With Omnipod, a tubeless, automated insulin delivery system designed for simplicity and control, Insulet is transforming diabetes management in a major way. Because Insulet relied on data to enhance every aspect of their operations — from product development to customer experience — real-time monitoring of the Omnipod’s performance was critical for troubleshooting issues, detecting product gaps and securing product reliability. Manufacturing teams needed visibility into machine efficiency across U.S. and Malaysian manufacturing facilities to optimize performance. Long term, they aimed to implement predictive maintenance to minimize downtime and reduce waste.
On the customer service side, AI-powered assistants were introduced to help agents find information faster and resolve cases more quickly. AI was also used to ensure regulatory compliance by identifying cases that were misclassified and should have been flagged as complaints. Manual approaches previously made it difficult to identify these overlooked cases, which increased the risk of non-compliance. Finally, the data engineering team wanted to use AI to automate error detection and debug more quickly.
However, the company’s data infrastructure struggled to keep pace with these initiatives, especially the ones around AI. Manufacturing teams lacked a unified view of production performance, forcing them to rely on fragmented data sources for important aspects of their business, such as product quality monitoring and customer experience. Also, SQL Server limitations made it difficult to analyze large datasets, restricting long-term performance tracking and delaying insights into product quality. Highly manual, time-consuming extract, transform and load (ETL) processes further slowed the flow of data, creating unnecessary bottlenecks in analytics and decision-making. This incomplete picture of data, hindered sales productivity, demand forecasting and supply chain optimization — while slowing product development and innovation.
On the external side, Insulet aimed to use IoMT (Internet of Medical Things) data streaming to improve the customer experience, particularly during major software rollouts like Omnipod 5’s iOS integration. The ability to track adoption rates and detect potential issues in real-time became increasingly important, yet existing systems struggled to process all the live data. As Bill Whiteley, VP of AI, Analytics and Advanced Algorithms at Insulet, explained, “Data pipelines took over 13 hours to update and caused our teams major delays in identifying product issues. Unfortunately, this impacted the customer experience since our dashboards couldn’t display more than two weeks' worth of historical information and limited long-term analysis of device performance.”
Not to mention, integrating operational data from Salesforce and SAP required costly third-party tools and further complicated an already inefficient practice. With all these disconnected systems, Insulet experienced challenges in creating a governed, scalable data environment capable of supporting its AI aspirations. With hopes of creating a “data flywheel” strategy that would feed directly into their goals for the future, Insulet turned to the Databricks Data Intelligence Platform.
Unifying data to drive innovation and operational efficiency
To modernize their data ecosystem for product innovation and operational efficiency, Insulet adopted Databricks' unified data platform, paving the way for their engineering teams to focus on extracting insights, rather than managing infrastructure. At the foundation of this transformation was Delta Lake, an open source storage layer that structured, cleaned and made data instantly available for analysis. By replacing fragmented systems with a centralized, scalable architecture, Insulet gave all teams seamless access to real-time data for monitoring manufacturing workflows, optimizing Omnipod performance and supporting all AI applications.
Layered on top of Delta Lake was the star of Insulet’s transformation: Lakeflow Connect, Databricks’ newest product that makes it easier to efficiently ingest data into its lakehouse architecture. Lakeflow Connect elevated Insulet’s data ingestion and processing by automating the integration of key enterprise applications, like Salesforce and Workday. Previously reliant on costly, third-party ETL tools, Insulet could now bring in near real-time, high-quality data with minimal effort required from its data engineers. “The Lakeflow Connect Salesforce Connector was especially important to our data transformation since it allowed us to directly ingest customer feedback, sales data and service interactions from the Salesforce CRM directly into Databricks,” Bill explained.
Manufacturing teams, once burdened with hours of manual reporting, leveraged a unified production dashboard to track software rollouts in real-time and prevent disruptions that impacted patient care. Plus, streaming and real-time analytics — powered by Spark, an open-source engine for large-scale data processing, and Delta Live Tables — compounded these improvements by replacing batch processing with continuous, low-latency data streaming. These changes helped an eight-person team to do the work of 100+ employees, completely changing how Insulet monitored Omnipod’s performance.
Thanks to Unity Catalog, a unified governance and data management solution that ensured secure access for Insulet’s various business units, data was better organized and easily discoverable. Databricks SQL, a fully-managed, serverless data warehouse, further unlocked this new level of data accessibility, which gave data analysts the power to run queries on demand without needing to manage clusters.
Further removing engineering bottlenecks, Insulet enhanced its AI and automation capabilities with the Databricks Assistant. As a context-aware AI assistant, it simplified SQL development by automating query writing through a conversational interface, detecting errors and debugging various issues for Insulet’s data teams, which freed up time for insights and innovation. Assistant also helped customer care and product support teams retrieve relevant information to support customers as needed.
With the Databricks Data Intelligence Platform, Insulet streamlined data ingestion, managed workflows, optimized resource allocation and enabled more efficient, data-driven decision making across its business.
Cutting costs and complexity to improve patient outcomes
Insulet has achieved substantial improvements, savings and overall efficiency with the Databricks Platform. In manufacturing alone, data-driven optimizations have resulted in tens of millions of dollars in savings. The transition from outdated ETL processes — that previously took over 13 hours — to live streaming has enabled real-time data access, while SQL queries were reduced from 40+ commands to just two. Not only did this drastically decrease data engineering workloads, it also cut query times by 83% with serverless processing.
Beyond cost savings, Databricks has helped Insulet’s ability to scale, innovate and optimize workflows across their manufacturing, engineering and customer service teams. “These enhancements have driven a 12x increase in data processing speed, a 25% boost in operational efficiency and a 97% reduction in TCO,” concluded Bill. More granularly, the Omnipod’s product monitoring has expanded from tracking just two weeks of historical data to two years, providing long-term visibility into performance, trends and potential issues. Such enhancements help ensure software updates are guided by live user data that inform UX advancements.
Down the road, Insulet is poised to deepen its AI and data capabilities to further enhance customer outcomes. The integration of SAP with Databricks, as well as an expansion of supported data sources from Lakeflow Connect, will help unify and continue to reduce fragmentation.
Next up, AI-driven predictive maintenance will harness IoMT data to deliver sharper, in-the-moment insights into device performance. And as the final cherry on top, Insulet plans to push AI-powered predictions directly into Salesforce to automate decision-making. With a fully modernized data ecosystem, Insulet is advancing its current operations while laying the foundation for continuous evolution in diabetes management.