WHOOP Improves AI/ML Financial Forecasting While Enhancing Members’ Experiences

Balancing physical activity with relaxation is vital for performing well at work, in the gym and throughout daily life. WHOOP helps people unlock better performance by using biometric data from its advanced wearable devices, which are designed to be worn around the clock. Members rely on the WHOOP app to monitor health metrics and access WHOOP Coach, an AI-powered performance coaching feature.

Data is central to the WHOOP member experience — and to the operational teams responsible for designing new features, forecasting revenue and retaining customers. Since moving to Snowflake from Amazon Redshift and Dremio, departments across WHOOP now enjoy expedited access to data and insights, allowing them to quickly understand members’ needs and make more informed business decisions.

At WHOOP, innovative feature launches, ongoing international expansion and a popular free trial has yielded an impressive number of new members — and exponentially more data. “As we started to experience hypergrowth, our data began doubling or tripling year-over-year,” says Matt Luizzi, Director of Business Analytics at WHOOP. “We needed a platform that could scale with us.”

But prior to Snowflake, data was spread across the data warehouse, data lake and production databases at WHOOP. Maintaining and scaling the company’s previous data warehouse, Amazon Redshift, was costly and operationally burdensome for the small data team. Cluster resizing jobs led to 15 minutes of downtime twice per day and long wait times during off-hours. According to Luizzi, “If it was quarter-end and we were working late, our data warehouse would just turn off and go super slow after that. It was a big pain point as we continued to scale.” Ensuring proper data governance for hundreds of terabytes of data — including large amounts of personally identifiable information (PII) — was also challenging.

Additionally, WHOOP also experienced cost challenges with Dremio, which the team used to query and manage Apache Iceberg tables in the company’s data lakehouse. This required EC2 instances to always be on, which demanded extra infrastructure costs and overhead to manage.

Recognizing the need for change, Luizzi turned to Snowflake to centralize the company’s data in one scalable, secure data platform. The AI Data Cloud’s elastic performance engine, Horizon Catalog data governance features like role-based access control (RBAC) and support for Apache Iceberg tables meet evolving needs at WHOOP — without jeopardizing budget. Cost monitoring dashboards built in Snowsight provide WHOOP with granular visibility into their Snowflake spend while eliminating any unexpected expenses. “We’ve never had a problem with our Snowflake cost. It’s been very controllable, manageable and there’s a lot of visibility built into the platform,” Luizzi says.

With Snowflake, the data team at WHOOP spends less time troubleshooting slow queries and infrastructure issues, which leaves more time for extending data access across the organization. Snowflake has become a catalyst for data-driven experimentation at WHOOP, leading to further consolidation of data. “We’ve done a lot of work to get to where 100% of our company’s data is queryable in Snowflake,” Luizzi says.

Powering the company’s data warehouse and data lake workloads with Snowflake means there’s less infrastructure to procure, manage and support. By consolidating onto Snowflake and using Apache Iceberg tables, workloads are now easier to manage and more performant, saving 20 hours of compute every day. Even better, data scientists, engineers, product developers, marketers and other business users at WHOOP know where to find reliable data. “We have a lot of focus on product analytics,” Luizzi says. “Being able to see which features are resonating — or not resonating — with members helps us direct our work and improve the overall experience.”

Visiting Snowflake’s Customer Experience Center in San Mateo, California, inspired Luizzi to collaborate with finance leadership at WHOOP and streamline revenue forecasting. With Snowpark and Hex, a Snowflake partner, the team rapidly built an automated AI/ML cohort model that automatically identifies patterns and trends within the company’s finance data. This new model has allowed WHOOP to reduce its reliance on CSV files while tripling model dimensionality for greater model accuracy, performance and scalability. These improvements translate into simpler financial forecasting for the WHOOP team and a more complete understanding of members.  

Moving forward, WHOOP plans to use Snowflake’s AI/ML features in tandem with Streamlit in Snowflake. The analytics team is currently working on a large language model (LLM) chatbot to help employees across the organization quickly access data, better understand metrics and easily surface answers to their questions. 

Snowflake Cortex AI has been valuable for testing proof of concepts, such as the support ticket classification solution at WHOOP. “Using Cortex LLM Functions, I could run three LLMs side by side and see which produced the best results,” Luizzi says. “Ultimately, I found that it was more accurate than humans churning through hundreds of tickets per day. That’s super powerful since it’s directly available in Snowflake.”

Polaris Catalog, an open source catalog for Apache Iceberg tables, will also give the analytics team new levels of choice, flexibility and control over their data. “The flexibility and vendor-neutral control of both Apache Iceberg and Apache Polaris open source projects are very appealing to ensure interoperability and no lock-in,” Luizzi says. “The team’s very excited to get going with Polaris.”