The Best Open-Source Analytics Tools Compared (2026)
How to read this comparison
Every tool here is open source, self-hostable, and a credible Google Analytics replacement. None is best at everything, and a comparison claiming otherwise is marketing.
The useful question is not "which is best" but which job are you hiring it for. Three different jobs show up repeatedly:
1. "Replace GA4, keep the features." → Matomo
2. "Just tell me my traffic, simply and privately." → Plausible or Umami
3. "Understand behaviour in my product." → PostHog
InsightsTrack sits between 2 and 3: web analytics with behavioural depth, without product-analytics weight.
The table
| Matomo | Plausible | Umami | PostHog | InsightsTrack | |
|---|---|---|---|---|---|
| Licence | GPL-3.0 | AGPL-3.0 | MIT | MIT + enterprise | MIT |
| Storage engine | MySQL (row) | ClickHouse | Postgres/MySQL (row) | ClickHouse | Postgres + DuckDB |
| Analytics DB to operate | MySQL | ClickHouse | Same DB | ClickHouse + Kafka | None (embedded) |
| Cookieless default | Configurable | ✅ | ✅ | ❌ | ✅ |
| Heatmaps | Paid plugin | ❌ | ❌ | ✅ | ✅ |
| Session replay | Paid plugin | ❌ | ❌ | ✅ | ❌ |
| Funnels | ✅ | Basic | Basic | ✅ | ✅ |
| Web Vitals | ❌ | Partial | ❌ | Partial | ✅ |
| Error tracking | ❌ | ❌ | ❌ | ✅ | ✅ |
| Feature flags | ❌ | ❌ | ❌ | ✅ | ❌ |
| Ecommerce | ✅ | Basic goals | ❌ | ✅ | Goals + funnels |
| Tag manager | ✅ | ❌ | ❌ | ❌ | ❌ |
| SQL access | Limited | Via ClickHouse | ❌ | ✅ HogQL | ✅ Built-in editor |
| Minimum realistic RAM | 2–4 GB | 4 GB+ | 1 GB | 16 GB+ | 2 GB |
Storage engine is the decision nobody makes deliberately
It is worth understanding, because it determines both your dashboard speed and your weekend.
Row stores (Matomo, Umami) read entire rows to answer a question about two columns. Simple to operate — it is the database you already run — but aggregations get slower as data grows. Matomo compensates with a pre-aggregation cron; Umami largely does not, which is fine until it isn't.
ClickHouse (Plausible, PostHog) is a columnar database built for this workload and very good at it. The cost is that it is a distributed system you now operate: sizing, merges, replication, upgrades.
Embedded columnar (InsightsTrack) uses DuckDB inside the API process — columnar read performance with no service to run. The ceiling is lower than a ClickHouse cluster: it is one process, so it suits a handful of concurrent dashboard users rather than hundreds of analysts.
Columnar vs row storage covers the mechanics, and DuckDB vs ClickHouse vs Postgres covers when to pick which.
Picking by job
Replacing GA4 feature-for-feature, ecommerce included → Matomo. Nothing else has the catalogue. Budget for the operational weight and the archiving system.
A public site where you want traffic numbers and nothing else → Plausible if you want polish and a hosted option, Umami if you want the smallest stack and MIT.
Product analytics on a logged-in app → PostHog. Session replay, flags, and experiments are the product. Use their cloud unless you are prepared to run ClickHouse.
Web analytics with behavioural depth, minimal ops → InsightsTrack. Heatmaps, Web Vitals, error tracking, funnels, and raw SQL, from Postgres plus an embedded engine. See the full comparison against GA4.
What "open source" hides
Three things to check before committing, regardless of which you pick:
- Licence terms. AGPL (Plausible) has a network clause that matters if you build a service on it. MIT (Umami, InsightsTrack) does not.
- Self-hosted parity. PostHog's self-hosted edition is not feature-equivalent to cloud, and some Matomo capabilities are paid plugins. "Open source" does not always mean "everything, free."
- Who operates it. The honest cost of self-hosting is your team's time, not the server. A tool needing one container is a genuinely different commitment from one needing a cluster.
Frequently asked questions
Which open-source analytics tool is easiest to self-host?
Umami and InsightsTrack are the lightest — a database plus an app. Matomo needs PHP and MySQL, and PostHog's self-hosted stack is substantially heavier than the rest.
Does "open source" always mean free to use commercially?
No, and the licence matters. MIT and GPL differ from AGPL, which adds a network clause affecting hosted services, and several projects reserve some features under a separate commercial licence.
Which tool has the smallest tracking script?
Plausible and Umami, both around 1 KB. Heavier scripts are collecting more — heatmaps, Web Vitals, error tracking — so compare what a script does before comparing weight alone.
Why does the storage engine matter when choosing?
Because it determines your operational burden more than the feature list does. A ClickHouse-backed tool and a SQLite-backed one may look similar in screenshots and differ enormously in what you maintain.
Can I try these without committing?
Yes — every tool here has a live demo or a Docker image that runs in minutes. Running your real traffic through two in parallel for a fortnight beats any comparison table, including this one.
Try before committing
All of these install as a single script tag and can run in parallel. Two weeks of overlap tells you more than any comparison table, this one included — including how each tool's numbers differ, which they will (metrics explained).
Try InsightsTrack free
Open-source, self-hosted, privacy-first analytics — explore the live demo, no install required.
Open live dashboard