PostHog

• CategoryProduct analytics
• PricingFreemium
• Starts atFree / $0
• LicenseMIT
• Created2020
• GitHub stars23k
• Script size~50KB
• CookielessNo
• Self-hostYes

PostHog is the open-source product analytics platform. Funnels, retention, paths, session replay, feature flags, A/B testing, surveys, error tracking, all in one product. Self-hostable on Kubernetes; managed cloud tier with a generous free plan (1M events/month). YC-backed, fast-moving, the closest thing to a one-vendor product-analytics stack in 2026.

• Product teams who need funnels + feature flags + session replay in one tool
• Startups who want PostHog Cloud free tier and the option to self-host later
• Teams escaping Mixpanel / Amplitude pricing

Skip PostHog if you only need web analytics, it is overkill. Skip if you need a polished mature product-analytics tool; Mixpanel and Amplitude are still ahead on dashboard polish.

PostHog is the right default for product-led companies in 2026. Self-host the EU instance for compliance, use feature flags + experiments out of the same tool, ship session replay when you need it. Pair with Plausible for marketing-site web analytics.

The argument for PostHog is not that any single module beats the category leader. It is that funnels, retention, paths, feature flags, experiments, session replay, surveys, and error tracking all read from the same event stream and the same person identity. That is the expensive part when you stitch separate vendors together. You flag a feature, experiment on it, watch a replay of the session where it broke, and never write an identity-mapping layer to join those three systems. The cost of that breadth is weight. The client bundle is far heavier than a privacy-first web tracker, so dropping PostHog onto a marketing landing page you have optimised for Core Web Vitals is a poor trade. Run it where the product is, run something light where the traffic is, and stop trying to make one tool cover both jobs.

PostHog is MIT-licensed and self-hostable, but the honest framing is that a production self-hosted deployment expects Kubernetes with ClickHouse, Kafka, and Redis behind it. That is a platform team's project, not an afternoon. The project itself has steered smaller users toward Cloud for exactly this reason, and there is an EU Cloud region that answers most of the data-residency questions people were self-hosting to solve. Unless a contract genuinely forbids a third-party processor, the EU region costs you less than standing up the cluster. On billing, pricing is usage-based per product, so the bill tracks event volume and replay minutes rather than seats. That scales kindly for a small team with a lot of users and much less kindly if you instrument every mouse movement. Configure sampling and event filtering on day one rather than after the first surprising invoice.

• Official homepage →
• GitHub repo →

The 30-min call is where your analytics choice becomes a real implementation, an event-tracking plan that fits your product, and a privacy story that does not need a 4-screen cookie banner.