Analytics

Superset75.4k★

Superset is a full business intelligence platform you can self-host for free, with interactive dashboards, charts, and reports without writing code. Think Tableau or Looker, but open source. Connect it to Postgres, MySQL, ClickHouse, BigQuery, or dozens of other databases, then build interactive dashboards with drag-and-drop. Fully free under Apache 2.0. No feature gating, no user limits, no enterprise unlock. You get the complete BI platform: SQL editor, chart builder, dashboard designer, role-based access, and scheduling. At, this is one of the most popular open source projects period. The catch: Superset is resource-hungry and complex to operate. The Docker Compose setup works for testing, but production deployment requires Redis, a metadata database, Celery workers for async queries, and careful memory tuning. Expect 4-8GB RAM minimum. And the learning curve is real. Building useful dashboards takes time, and the UI can feel overwhelming. If you just need a few charts, Metabase is much simpler to get running.

Metabase50.2k★

Metabase lets people who don't write SQL build charts and dashboards by clicking, on top of your existing database or warehouse (20+ supported sources), with a SQL editor for when clicking runs out. The open source edition is AGPL and free with unlimited users, unlimited dashboards, and AI questions on your own model key. It runs as one Docker container or JAR, but the built-in H2 app database isn't for production, so point it at Postgres from day one. Then patch fast. 2026 brought a run of critical advisories, including an unauthenticated SQL injection (CVE-2026-72898) exploited in the wild for admin takeover, plus several code execution bugs. Row and column permissions, SAML and JWT SSO, caching controls, auditing, whitelabeling and multi-tenant embedding are Pro: $575/mo with 10 users, then $12 each, cloud or self-hosted. Cloud Starter is $100/mo with 5 users. Enterprise starts at $20,000/year. Solo and small teams: free self-hosted, or Starter to skip ops. Medium teams: Pro once SSO or row-level rules matter. Larger teams: price Superset before signing. The catch: customer-facing embeds on the free edition are view-only, carry a "Powered by Metabase" banner, and drill-through needs Pro.

PostHog40.7k★

PostHog puts product analytics, session replay, feature flags, A/B tests, surveys, error tracking, logs and LLM observability behind one SDK and one bill. The free cloud tier resets monthly: 1M analytics events, 5K recordings, 1M flag requests, 100K exceptions, 1,500 survey responses and 10 GB of logs. Self-hosting is a hobby option, not a product. PostHog's docs call it officially unsupported, suggest it only under 300K events a month, keep every paid feature Cloud-only, and publish no CVEs because security fixes ship straight to the latest image. Past the free tier, anonymous events cost $0.00005 each and identified events (tied to a user profile) $0.000198, so how you instrument moves the bill about 4x. Recordings run $0.005. Enforced SSO is the $250/mo Boost package, SAML is $750/mo Scale, RBAC is Enterprise only. Solo and small teams: free for a long time. Medium teams: usage pricing with billing limits. Larger teams: Scale or Enterprise. The catch: the MIT code is real but the escape hatch is decorative. Need self-hosting? Look at Umami, Plausible or Matomo, plus Unleash for flags. And in November 2025 the Shai-Hulud worm pushed malicious posthog-js and posthog-node versions to npm for about five hours. Pin your SDKs.

Umami39.9k★

Umami is the privacy-first swap for Google Analytics: pageviews, referrers, UTM campaigns, custom events, funnels, retention and revenue, with no cookies and no consent banner. It's MIT licensed and free to self-host. Since v3 it has grown well past a simple counter, adding session replays, heatmaps and an MCP endpoint for AI assistants. Self-hosting is docker compose up -d: one Node app plus PostgreSQL (v3 dropped MySQL). Change the default admin/umami login and the placeholder APP_SECRET on day one. Then keep patching: this year's disclosures include a critical authenticated SQL injection (fixed in 3.0.3), a flaw that exposed any site's stats from its public website ID (fixed in 3.2.0), and the October 2026 3.4.1 hotfix for a critical Next.js RCE. Umami Cloud bills monthly. Hobby is free for 100K events and 1 website. Pro is $20/month for 1 million events, 20 websites and 10 members. Business is $200/month for 10 million events plus replays and heatmaps. Solo: Hobby or self-host. Small teams: Pro. Larger teams: Business, or self-host. The catch: Cloud jumps $180 a month from Pro to Business, and email reports and the streaming API are not in the self-hosted build at all. Plausible and Fathom are the closest rivals.

Plausible29.8k★

Plausible replaces Google Analytics for people who only wanted the dashboard. Traffic, sources, top pages and goals sit on one page, with no cookies or personal data, so no consent banner. The self-hosted Community Edition (CE) is free under AGPLv3. CE runs on Docker Compose with Postgres and ClickHouse. Plan on 2GB of RAM and a CPU with SSE 4.2 or NEON, which ClickHouse requires. It ships as a long-term release twice a year, and its bot filtering is basic, without Cloud's data-center IP blocklist. Cloud is priced by pageviews. At 10k a month: Starter $9/month (one site), Growth $14/month (3 sites, 3 members), Business $19/month (funnels, revenue attribution). At 1M those become $69, $104 and $139. Yearly billing gives two months free. CE never gets funnels, user journeys, revenue goals, SSO or the Sites API. Solo: Starter. Small teams: Growth. Larger teams: Business or Enterprise, or CE if the basic dashboard is enough. The catch: on CE, patching is your job. In June Plausible disclosed a critical unauthenticated remote code execution bug (CVE-2026-8467) via a /storybook endpoint in CE 3.0 through 3.2.0. Umami and Matomo are the other self-hosted picks; PostHog covers the user-level analytics Plausible skips.

dataease24.7k★

DataEase turns your databases and spreadsheets into dashboards and big-screen displays by dragging fields onto charts, no SQL required. It comes from Fit2Cloud, the Chinese company behind JumpServer and 1Panel, and it is GPL-3.0. Version 3 landed in late August 2026 with a new permission model, a spreadsheet-style report builder, and support for Postgres, SQL Server, and Oracle as its own metadata database instead of MySQL only. Install is a one-line script run as root on a Linux server with 2 cores and 4 GB of RAM or more. The connector list is strong: MySQL, Postgres, ClickHouse, Doris, StarRocks, Redshift, Excel, CSV, and APIs. The paywall sits lower than it looks. Fit2Cloud's own edition comparison shows the free community edition has no user management, no permissions, no scheduled reports, and no SSO. It is one admin sharing dashboards by link. Teams that need separate logins buy Pro at ¥60,000 (about $8,400) per deployment, a year of maintenance included. metabase/metabase and apache/superset give you real multi-user access for free. The catch: the new spreadsheet module is Enterprise-only, V3 changes how resources and permissions work, so read the upgrade guide first, and support and docs are mostly in Chinese.

Matomo22.0k★

You self-host it, you own the data, your visitors' browsing habits stay on your servers. GDPR compliance without the headache of cookie consent banners (in most configurations). GPL v3, used by over a million websites including government agencies and universities. It tracks page views, sessions, referrers, goals, e-commerce conversions, basically everything Google Analytics does. Self-hosting is free with no feature limits. Their cloud-hosted version starts at $23/mo for 50K hits. On-premise premium plugins (like A/B testing, funnels, custom reports) cost $229/year per plugin or $629/year for the bundle. The catch: Matomo's self-hosted interface feels dated compared to modern analytics tools like Plausible or PostHog. It's powerful but overwhelming: hundreds of reports, dense UI, lots of configuration. And the PHP/MySQL stack means you need traditional web hosting, not just a static server. For teams that just want pageviews and referrers, Plausible or Umami are simpler.

niubigeo8.9k★

NiubiGEO checks whether AI models know your product exists. It runs the questions your customers would actually ask through OpenRouter, OpenAI, Claude, Gemini, Perplexity, and DeepSeek, records whether you get mentioned, notes which competitors show up instead, and traces the citations behind each answer. Search rankings used to be the whole game. Now a lot of buying research never touches a results page. Setup is Docker compose or Node 22 and up, a clone, and an env file with at least one provider key. It queries with web search and grounding turned on where the provider supports it, so you pay that provider's per-call rate. A broad prompt set across six providers adds up faster than people expect. Apache 2.0, self-hosted, no license fee. Marketing people and founders at small companies are the fit. Larger teams already pay for a commercial visibility tracker, and this gets most of the same signal for the price of the API calls. The catch is that model answers are non-deterministic and shift with every model update. One run is a snapshot, not a measurement. Trustworthy output means the same prompt set on a schedule over weeks. The tool gives you the harness for that, not the discipline.

jitsu5.1k★

Jitsu is an open-source event collection and data pipeline tool. It captures events from your web and mobile apps and streams them directly to your data warehouse or analytics tools. Segment-style routing, warehouse-first design. MIT-licensed. Self-hosted via Docker. Simpler to operate than RudderStack for teams with fewer destination needs. The warehouse connector is the strong suit: Snowflake, BigQuery, ClickHouse, Redshift. SDK support covers JavaScript, Python, and mobile. Teams that want to capture web events and land them cleanly in a warehouse without paying Segment's per-MTU pricing get immediate value here. Less useful if you need 200+ marketing-tool destinations. The catch: Jitsu is leaner than RudderStack or Segment, which means fewer pre-built connectors. If your use case is warehouse-only, that is fine. If you need Salesforce, HubSpot, or ad platform integrations, the connector library gets thin fast.

rudder-server4.5k★

RudderStack is a Customer Data Platform that collects events from your web, mobile, and server sources and routes them to 200+ destinations: data warehouses, analytics tools, CRMs, and advertising platforms. The open-source alternative to Segment. Apache-licensed, free to self-host. Setup requires Docker Compose and a Postgres database. The data plane handles event ingestion at scale. You manage infrastructure; RudderStack handles routing logic. A control plane UI is available separately for destination configuration. Engineering teams that track user events and pipe them to Snowflake, BigQuery, or Amplitude without the Segment bill get a real cost save here. The routing logic is comparable to Segment. Self-hosted means you control data residency. The catch: the self-hosted control plane is less polished than the cloud version. Destination support can lag Segment. And high event volumes require serious infrastructure to match Segment's managed reliability.

ossinsight2.5k★

OSSInsight crunches over 10 billion GitHub events to show you what's actually happening in open source. Star velocity, contributor growth, PR merge rates, issue response times, head-to-head repo comparisons. The hosted version at ossinsight.io is free and requires no account. GitHub's built-in Insights tab looks like a toy next to this. The platform is powered by TiDB (PingCAP's distributed database), which is the quiet part out loud: this project exists partly as a TiDB showcase. That said, the analytics are genuinely useful. Curated collections track trends across AI frameworks, databases, LLM tools, and more. The Data Explorer lets you write custom SQL against the full GitHub event archive. Developers evaluating which project to adopt get real data instead of vibes. Open source maintainers track their project's health metrics. Anyone doing market research on developer tools gets a free analytics platform that would cost thousands to build from BigQuery. The catch: public GitHub data only, no private repos. Self-hosting requires TiDB plus terabytes of ingested archive data, which is realistic only for organizations with dedicated infrastructure teams. And since PingCAP built this as marketing, long-term independent maintenance is an open question.

talivia2.3k★

Talivia ties your web analytics to your actual revenue. It's a self-hosted, open source "revenue-first" analytics platform for founders that combines the usual web analytics and session replay with revenue attribution pulled straight from payment processors like Stripe, LemonSqueezy, and Polar. Instead of two dashboards that never talk, you see which traffic actually turned into money. It's a Next.js and Postgres app. The pitch is the join most tools skip: connecting behavior to dollars in one place. It goes after PostHog, Plausible, and Matomo on the analytics side, and the "datafast" style revenue-analytics products directly. Solo founders and small teams self-host it for free. The paid path is Talivia Cloud, which adds the managed hosting plus richer integrations like Search Console, GitHub activity, and social mentions. The catch is freshness. This is very new, the public history is thin, so maturity and long-term maintenance are real questions before you build reporting on it. It's also open-core: the most useful integrations sit behind the Cloud tier, so "free self-hosted" gets you the base, not the full picture. Promising, but pilot it before you trust it.

new-term-radar2.1k★

new-term-radar catches the names of new AI models, products and repos the day they surface, before keyword tools list them. It reads X accounts you pick, Hugging Face trending, fast-rising GitHub repos and AI directory sitemaps, then a small model filters out common words. Built for SEO sites, newsletters and directories. The code is MIT. Setup is a clone, a config file, one API key and python -m radar daily. A bundled GitHub Actions workflow runs it on a schedule. The three agent skills and MCP server are wrappers around that CLI. The code is free; the data isn't. X posts, Google Trends, search volumes and the judging model all run through AIsa, a pay-per-call reseller, and the README estimates about $15 a month on the example config. Without the key you get Hugging Face, GitHub and sitemaps, unsorted. Solo and small teams doing new-term SEO: worth $15. Larger teams on Semrush or Exploding Topics: this finds terms earlier, not instead. The catch: the README's AIsa links carry the author's referral code (disclosed), and your pipeline depends on a reseller keeping X and Google feeds alive. One signal is already broken upstream. Hit rates are unproven; the author says so.