Five steps, together
Provisioning, keys, schema, DNS, verification — one guided session, run by our engineers alongside your DevOps team.
Deployment walkthroughSelf-hosting
Observer is self-hosted product analytics: ClickHouse, the collector, the API and the dashboard run on YOUR infrastructure — deployed and operated jointly by our team and your DevOps. A 2 GB VPS class is enough, and there are no per-traffic, per-site or per-seat charges. Ever.
The deployment, end to end
Everything below happens on your hardware, in a joint session with your DevOps team. Source access is granted after NDA and licensing — the deployment itself never depends on you reading our code first.
Provisioning, keys, schema, DNS, verification — one guided session, run by our engineers alongside your DevOps team.
Deployment walkthrough2 GB RAM class. Four zero-dependency Node 22 processes and ClickHouse 24.8. Docker Compose or bare Node under systemd — no orchestrator.
What it runs onSites, collector, ClickHouse, API + dashboard — the collector is the only internet-facing component. Everything else stays inside.
The architectureDisk-spool replay survives ClickHouse outages, alerts evaluate every minute, upgrades are config swaps — not migration projects.
Operations modelAt ~60M events a month, the infrastructure bill is roughly €60–90 — your hardware, sized together. Your license does not scale with traffic.
The economicsThe 6-spec browser privacy suite runs against your deployment during onboarding — the report goes to your DevOps, not to a marketing PDF.
Verify privacyThe only public line
Everything server-side is handled during onboarding. What your sites carry is a single async, cookie-less script tag — served from your own subdomain, locked to your allowlisted domains.
<!-- generated per site during onboarding — this is the whole install -->
<script async src="https://sa.yourdomain.com/v.js" data-site="obs_XXXX"></script>
Everything else — ClickHouse, the collector, keys, schema, TLS — is provisioned jointly with your DevOps during onboarding. The snippet is domain-locked: pasted on any other site, the collector refuses it (403).
The operating model
Observer is privately licensed: access follows an NDA and a license agreement, and deployments happen together with your DevOps team. Our engineers drive the provisioning, keys, schema and privacy verification on your hardware — your team keeps root, keeps the network, and keeps the data. After go-live we stay in the loop: runbooks at onboarding, upgrades as config swaps, monitoring shared.
The stack runs on machines you control — no third-party endpoint, no subprocessor chain, no international transfers.
Deployment and operation are done with your DevOps, not handed over as a checklist. You are never alone with a broken ingest.
No per-traffic, per-site or per-seat charges. The marginal event costs zero — that is a property of your hardware, not our pricing tier.
Request access, sign the NDA, and we schedule the joint session with your DevOps team. Five steps later, the dashboard is alive on your own infrastructure.