Self-Hosted Analytics Comparison

Why Self-Host Your Analytics?

Google Analytics is free because you are the product. Every pageview, scroll depth, and conversion event feeds Google’s ad targeting machine. Self-hosted analytics gives you full ownership of your traffic data, zero third-party tracking, and — in most cases — a cleaner, faster dashboard than GA4’s labyrinthine interface.

But “self-hosted analytics” is not one tool. There are 11+ serious options, and they solve different problems. This guide compares all of them so you can pick the right one without deploying five different stacks first.

If you just want to ditch Google Analytics entirely, see Self-Hosted Alternatives to Google Analytics.

Prerequisites

• Familiarity with Docker Compose (all tools below run in Docker)
• A Linux server with at least 1 GB RAM (some tools need more — see the comparison table)
• A domain name for accessing your dashboard remotely
• A reverse proxy for SSL termination

No Docker Compose configs in this guide — each tool’s dedicated setup guide has the full stack. This guide is about choosing the right tool.

The Landscape at a Glance

Self-hosted analytics tools fall into three categories:

1. Simple web analytics — page views, referrers, top pages. Lightweight, privacy-first, cookie-free. Replaces GA for content sites. Examples: Plausible, Umami, GoatCounter, Fathom Lite, Ackee, Shynet.

2. Full web analytics — everything above plus goals, funnels, segmentation, custom dimensions, heatmaps. Replaces GA4 feature-for-feature. Examples: Matomo, Open Web Analytics, Countly.

3. Product analytics — event tracking, user identification, session replays, feature flags, A/B testing. Replaces Mixpanel/Amplitude. Examples: PostHog.

GoAccess is the outlier — it parses server access logs directly, requires no JavaScript snippet, and runs as a terminal application or static HTML report.

Comprehensive Comparison Table

	Plausible	Umami	Matomo	PostHog	GoAccess	Ackee	GoatCounter	Fathom Lite	Shynet	Open Web Analytics	Countly
License	AGPL-3.0	MIT	GPL-3.0	MIT (self-host)	MIT	MIT	EUPL-1.2	MIT	Apache-2.0	GPL-2.0	AGPL-3.0
Language	Elixir	Next.js (Node)	PHP	Python/TS	C	Node.js	Go	Go	Python (Django)	PHP	Node.js
Database	ClickHouse + PostgreSQL	PostgreSQL or MySQL	MySQL or MariaDB	ClickHouse + PostgreSQL	None (reads log files)	MongoDB	PostgreSQL or SQLite	SQLite	PostgreSQL	MySQL	MongoDB
Tracking method	JS snippet	JS snippet	JS snippet	JS snippet	Server log parsing	JS snippet	JS snippet (or pixel)	JS snippet	JS pixel (no-JS capable)	JS snippet	JS SDK
Script size	~1 KB	~2 KB	~22 KB	~70 KB	N/A (no script)	~1.5 KB	~3.5 KB	~1.5 KB	N/A (pixel)	~75 KB	~50 KB
Cookie-free	Yes	Yes	Optional (yes by default in recent versions)	No (uses cookies)	N/A	Yes	Yes	Yes	Yes	No	No
GDPR-friendly	Yes — no consent banner needed	Yes — no consent banner needed	Yes if configured correctly	Requires consent	Yes — no JS, no cookies	Yes — no consent needed	Yes — no consent needed	Yes — no consent needed	Yes — no consent needed	Requires consent	Requires consent
Dashboard	Single page, real-time	Multi-page, real-time	Complex, GA-style	Complex, product-focused	Terminal or static HTML	Single page, minimal	Single page, simple	Single page, minimal	Multi-page, detailed	GA-style	Complex, multi-view
Custom events	Yes	Yes	Yes	Yes (core feature)	No	No	Limited	No	No	Yes	Yes (core feature)
Funnels	Yes (paid cloud; CE basic)	No	Yes	Yes (core feature)	No	No	No	No	No	Yes	Yes
Session replays	No	No	Yes (plugin)	Yes (core feature)	No	No	No	No	No	No	No
API	Yes (REST)	Yes (REST)	Yes (REST)	Yes (REST)	JSON output	Yes (GraphQL)	Yes (REST)	No	Yes (REST)	Yes (REST)	Yes (REST)
Mobile analytics	No	No	Yes (SDK)	Yes (SDK)	No	No	No	No	No	No	Yes (SDK)
Import from GA	Yes	No	Yes	No	No	No	Yes (CSV)	No	No	No	No
RAM usage (idle)	~500 MB (ClickHouse)	~150 MB	~256 MB	~1 GB+ (ClickHouse)	~10 MB	~100 MB	~30 MB	~20 MB	~150 MB	~200 MB	~400 MB
Best for	Content sites, blogs, SaaS marketing pages	Content sites, multi-site tracking	GA replacement with full feature parity	SaaS product analytics, startups	Server operators who avoid JS, bandwidth analysis	Minimal privacy-first tracking	Personal sites, blogs, small projects	Minimal single-site tracking	Pixel-based privacy tracking	Legacy PHP environments	Mobile + web product analytics
Actively maintained	Yes	Yes	Yes	Yes	Yes	Low activity	Yes	No (archived)	Low activity	Low activity	Yes

Decision Tree

Stop reading the table. Here is what you should actually deploy:

If you run a content site, blog, or marketing page

Use Plausible. It is the best balance of useful data, minimal resource usage, privacy compliance, and active development. The single-page dashboard shows everything you need — top pages, referrers, countries, devices — without clicking through five menus. No cookies, no consent banner, GDPR-compliant by default.

If Plausible’s ClickHouse dependency bothers you (it needs ~500 MB RAM), use Umami instead. Nearly identical feature set, runs on PostgreSQL alone, lower RAM footprint (~150 MB). The trade-off: Umami’s dashboard requires more clicks to get the same information, and it lacks GA import.

For a detailed head-to-head, see Plausible vs Umami.

If you need Google Analytics feature parity

Use Matomo. It is the only self-hosted tool with funnels, segmentation, custom dimensions, heatmaps (via plugin), session recording (via plugin), and a GA data importer. If your marketing team is used to GA and will revolt without similar depth, Matomo is the answer.

The cost: Matomo’s interface is complex, its PHP codebase is heavy, and the tracking script is 22 KB. See Plausible vs Matomo and Umami vs Matomo for trade-off details.

If you need product analytics (events, funnels, session replays, feature flags)

Use PostHog. It is not a web analytics tool — it is a product analytics platform. Event tracking, user identification, session replays, feature flags, A/B experiments, all self-hosted. PostHog replaces Mixpanel, Amplitude, and LaunchDarkly in one stack.

The cost: PostHog is resource-heavy (ClickHouse + PostgreSQL + Redis + multiple services). Plan for 2 GB+ RAM minimum. It also uses cookies and requires a consent banner under GDPR.

For the distinction between web and product analytics, see Matomo vs PostHog.

If you refuse to run JavaScript on your site

Use GoAccess. It parses Nginx, Apache, or Caddy access logs directly. Zero JavaScript, zero cookies, zero tracking pixels. It runs in the terminal or generates a static HTML report. Ideal for server operators who want traffic visibility without modifying their site.

The trade-off: no real-time dashboard (unless you use the WebSocket mode), no referrer detail beyond what the HTTP Referer header provides, no custom events.

If you want the absolute lightest option

Use GoatCounter. A single Go binary with SQLite, ~30 MB RAM, minimal JS snippet. No cookies, GDPR-compliant, actively maintained. Perfect for personal sites and small projects where Plausible’s ClickHouse dependency is overkill.

If you track mobile apps alongside web

Use Countly or PostHog. Both have native mobile SDKs (iOS and Android). Matomo also has mobile SDKs but they are less polished. None of the lightweight tools (Plausible, Umami, GoatCounter) support mobile app tracking.

Tools to avoid for new deployments

Fathom Lite — archived by its maintainers. The commercial Fathom product is cloud-only now. The open-source Lite version works but receives no updates.

Ackee — development has slowed significantly. It also requires MongoDB, which is unusual overhead for a minimal analytics tool.

Shynet — clever pixel-based approach but low commit activity. Fine if you specifically need no-JS tracking; otherwise Plausible or GoatCounter are better maintained.

Open Web Analytics — still PHP 7.x era. Functional but dated. Use Matomo if you need a full-featured PHP analytics stack.

Resource Requirements Summary

For capacity planning on your server:

Tool	Minimum RAM	Database overhead	Suitable for shared server?
GoatCounter	64 MB	SQLite (none)	Yes
Fathom Lite	64 MB	SQLite (none)	Yes
GoAccess	32 MB	None	Yes
Ackee	256 MB	MongoDB (~200 MB)	Marginal
Shynet	256 MB	PostgreSQL (~100 MB)	Marginal
Umami	256 MB	PostgreSQL (~100 MB)	Yes
Matomo	512 MB	MySQL/MariaDB (~200 MB)	Marginal
Open Web Analytics	256 MB	MySQL (~200 MB)	Marginal
Plausible	768 MB	ClickHouse + PostgreSQL (~500 MB)	No — give it its own resources
Countly	1 GB	MongoDB (~400 MB)	No
PostHog	2 GB+	ClickHouse + PostgreSQL + Redis	No — dedicated server recommended

If you are running analytics alongside 10 other services on a single mini PC, GoatCounter, Umami, or GoAccess are the practical choices.

Privacy and Compliance

Every analytics tool collects some data. Here is what matters for GDPR, CCPA, and ePrivacy compliance:

No consent banner needed (cookie-free, no personal data stored): Plausible, Umami, GoatCounter, Fathom Lite, Ackee, Shynet, GoAccess.

Consent banner recommended (uses cookies or stores personal identifiers): Matomo (configurable — can be cookie-free), PostHog, Open Web Analytics, Countly.

If eliminating consent banners entirely is a priority, your shortlist is Plausible, Umami, or GoatCounter.

Migration Path from Google Analytics

If you have historical GA data you want to preserve:

• Plausible — has a built-in Google Analytics importer (both UA and GA4)
• Matomo — has a Google Analytics importer plugin
• GoatCounter — accepts CSV imports (manual formatting required)
• All others — no import path; you start fresh

For most sites, starting fresh is fine. Historical GA data loses value quickly, and the metrics are not directly comparable between platforms anyway.

Common Mistakes

Deploying PostHog when you need page view counts. PostHog is a product analytics platform. If you just want to know which blog posts get traffic, you are deploying a semi-truck to carry a grocery bag. Use Plausible or Umami.

Choosing Matomo “just in case” you need advanced features later. Matomo’s operational overhead (PHP, MySQL, cron jobs for report archival) is significantly higher than Plausible or Umami. If you are not using funnels, segmentation, or custom dimensions today, start simple and migrate later if needed.

Ignoring ClickHouse’s memory requirements. Both Plausible and PostHog use ClickHouse, which is memory-hungry. On a 1 GB VPS, ClickHouse will OOM-kill your other services. Plan for 2 GB minimum if running Plausible alongside other apps, 4 GB if running PostHog.

Running multiple analytics tools simultaneously. Each JS snippet adds page weight and tracking overhead. Pick one. If you want log-based analysis alongside JS analytics, GoAccess is the only one that adds zero page weight since it reads server logs.

Next Steps

Pick your tool based on the decision tree above, then follow the dedicated setup guide:

• Self-Host Plausible — recommended for most sites
• Self-Host Umami — lightweight alternative to Plausible
• Self-Host Matomo — full GA replacement
• Self-Host PostHog — product analytics
• Self-Host GoAccess — server log analysis

Once deployed, add the tracking snippet to your site’s <head> tag (or configure log parsing for GoAccess) and verify data is flowing within minutes.

Related

• Best Self-Hosted Analytics
• Self-Hosted Alternatives to Google Analytics
• How to Self-Host Plausible
• How to Self-Host Umami
• How to Self-Host Matomo
• How to Self-Host PostHog
• How to Self-Host GoAccess
• How to Self-Host Ackee
• How to Self-Host GoatCounter
• How to Self-Host Fathom Lite
• How to Self-Host Shynet
• How to Self-Host Open Web Analytics
• How to Self-Host Countly
• Plausible vs Umami
• Plausible vs Matomo
• Umami vs Matomo
• Matomo vs PostHog
• Docker Compose Basics
• Reverse Proxy Setup