Mixpanel

Mixpanel was one of the earliest companies to build product analytics around events rather than pageviews — a distinction that sounds technical but fundamentally changed how product teams measure what matters. Founded in 2009, it pioneered the idea that tracking "user clicked checkout button" is more valuable than tracking "homepage was loaded," and that understanding user journeys through funnel analysis requires event-level granularity. This philosophy became the standard for modern product analytics, and Mixpanel still serves thousands of companies from early-stage startups to publicly traded companies who depend on its dashboards for critical product decisions.

Mixpanel ingests billions of events per day across its customer base, processes them through a pipeline that must maintain low latency for real-time dashboards while also handling complex retrospective queries for funnel and retention analysis. Its JQL (JavaScript Query Language) and the more recent Spark-based analysis backend must handle queries that often span months of historical data across billions of events — a computationally demanding operation that can degrade during periods of infrastructure stress or when particularly expensive queries run across the cluster simultaneously.

When Mixpanel is having issues, product teams experience: the Events dashboard showing a gap in event counts for recent hours (the clearest sign of an ingestion pipeline delay), funnel reports returning loading spinners that eventually time out, user profile lookups failing to return data even for known user IDs, saved reports that previously loaded quickly now taking minutes to resolve, SDKs successfully sending events (the network call returns 200) but those events not appearing in the dashboard, and the Mixpanel Notifications (messaging) feature failing to deliver scheduled in-app messages.

Live Mixpanel platform status is tracked on Outage.gg. An event ingestion delay is among the more subtle but consequential failures a product team can experience, as it can make a real drop in usage look like a data pipeline problem or vice versa.