Where product analytics fails: Measuring invisible products

Traditional product analytics assumes users, sessions, and a UI, but what if your product is invisible? At companies like CircleCI and Mux, I worked on infrastructure and developer tools where no user interaction was the goal: if someone showed up, it meant something had failed. Today, AI platforms, SDKs, and embedded tools increasingly operate the same way; quietly integrated into production workflows, triggered by systems, not people.

In this talk, I’ll share how to build analytics functions for products without interfaces, where “engagement” means system activity, not clicks. This talk is designed for analytics engineers, data scientists, and product data teams supporting infrastructure, AI platforms, APIs, observability tools, internal developer products, and anyone trying to measure success when there’s no traditional user behavior to track. We’ll explore how to model adoption through API usage, job runs, asset flows, and organization-level patterns. You’ll leave with a practical framework for redefining engagement, building trustworthy metrics, and making invisible products measurable.