Self-hosting your web analytics: what you gain, what it costs

The pitch for self-hosted analytics is one sentence: your visitors' behavior is your data, and right now it probably lives in someone else's cloud, on their terms, at their price, until their next deprecation. Here's the honest version of both columns.

What you gain

• Ownership: raw events in your own Postgres. Query them with SQL, join them with your product data, keep them forever. No API quotas, no sampling, no "this report is unavailable for your tier".
• Privacy posture: no third party receives your visitors' data, which simplifies GDPR conversations enormously, and lets you run cookieless if you choose.
• Ad-blocker resilience: blocklists target known analytics vendors' domains. A tracker served from your own domain family is far less likely to be filtered, so your numbers are closer to true.
• One tool instead of five: traffic, replay, heatmaps, vitals and error tracking usually means four subscriptions and four snippets slowing your site. A unified pipeline is one snippet and one bill (your infrastructure).

What it costs

You operate it: a database, a deployment, occasional migrations when you upgrade. Managed platforms (a Supabase project plus a Vercel app) shrink this to roughly the effort of running any side project, with infrastructure cost in the tens of dollars a month at millions of events, but it is not zero. You also own your uptime: if ingestion goes down, those events are gone. A well-built tracker fails silently and never harms the host site, but the gap in your charts is yours to own.

Who should (and shouldn't)

Good fit: developers and teams who already deploy web apps, run several sites, care about data ownership, and want behavioral depth (replay, frustration signals) without enterprise pricing. Poor fit: teams with nobody comfortable deploying software, or organizations that need certified, audited analytics contracts.