Plausible Analytics

Founder Uku Taht kept the first stats engine deliberately simple: write pageviews to PostgreSQL and aggregate at query time. Donald Knuth’s “premature optimization” quote was the permission slip.

That worked until it did not — but it got them paying customers and a product people could self-host later without a PhD in distributed systems.

A blog post about ditching Google Analytics hit the front page of Hacker News. April brought 272 signups — more than the previous nine months combined — and dashboard loads of 5–6 seconds on the demo graph.

Enterprise prospects asked for plans above their largest tier (~10M pageviews/month). Taht started evaluating column-oriented OLAP instead of pre-aggregating every possible chart upfront.

ClickHouse landed in production in May 2020; demo graph loads dropped from ~6s to ~500ms. MRR crossed $1k that same month. Docker support followed so the community could self-host Postgres + ClickHouse together.

Later infrastructure posts describe plain Docker on Hetzner, ingesting over a billion events per month from tens of thousands of sites — scaling vertically first, horizontal scale-out as a planned next step, not day-one Kubernetes.