Umami vs PostHog: Which Self-Hosted Analytics in 2026?
You decided to stop handing your visitors to Google Analytics, went looking for a self-hosted replacement, and quickly ran into the two names everyone mentions: Umami and PostHog. On a features page they look like rivals. They are not, really. One is a lightweight web-traffic counter you can run on a tiny box. The other is a full product-analytics platform that wants a small cluster to itself. Pick the wrong one and you either stand up a 32 GB stack you never needed, or you end up with a pretty pageview chart when what you needed was a funnel.
Here is the honest breakdown so you land on the right one the first time.
What each one is actually for
Umami answers "how many people came, from where, and what did they look at?" It tracks pageviews, referrers, devices, and custom events, cookie-free and privacy-first, and it does that one job cleanly. It is MIT-licensed and famously easy to run.
PostHog answers "why do users behave the way they do?" It bundles product analytics, session replay, feature flags, A/B testing, surveys, and error tracking into one platform. It is built to follow a user through a funnel, replay their session, and let you flag-gate a feature for 10 percent of them. Its core is MIT-licensed too, but the ambition comes with weight.
Side by side
| Dimension | Umami | PostHog |
|---|---|---|
| Best for | Website traffic analytics | Deep product analytics |
| Core features | Pageviews, referrers, devices, custom events | Funnels, session replay, feature flags, A/B tests, surveys |
| Infrastructure | App plus one database (PostgreSQL or MySQL) | ClickHouse, Kafka, Postgres, Redis |
| Memory footprint | Runs on a small VM | Around 32 GB RAM for a real deployment |
| Privacy | Cookieless, minimal data by design | Captures rich user-level data |
| Self-host difficulty | Easy | A distributed system to operate |
The infrastructure reality nobody warns you about
This is where the choice really gets made. Umami is one app and one database. It happily runs on a small VM, sips resources, and you will rarely think about it again.
PostHog is a distributed system. A real deployment runs ClickHouse for event storage, Kafka for ingestion, Postgres, and Redis together, which is why the guidance lands around 32 GB of RAM to start. PostHog is refreshingly honest about this: for most teams, they will tell you Cloud is the right answer. Even session replay, which once lived in ClickHouse, was moved to S3-compatible blob storage precisely because the original approach got expensive at scale.
None of that makes PostHog bad. It makes it a different class of tool. You are not paying that operational tax for a pageview counter, you are paying it for funnels, replays, and flags. If you will use those, it is worth it. If you will not, it is a lot of machine to watch a line go up.
So which should you run?
Pick Umami if you mainly want to know your traffic numbers, you care about privacy and GDPR simplicity, and you want something you can stand up in minutes and forget. For blogs, marketing sites, docs, and most small products, this is the right call and the cheaper one.
Pick PostHog if you are building a product and need to understand behavior, not just counts. When "where do users drop off in onboarding?" and "did this feature move the metric?" are questions you ask weekly, PostHog's funnels, replay, and flags earn their keep, and running it yourself keeps that behavioral data in your own infrastructure.
A quiet third option: run both. Umami on your marketing site, PostHog on the app. They are not mutually exclusive, and plenty of teams end up there.
A couple of gotchas
Do not under-provision PostHog. The most common self-hosting mistake is starting it on a small box and watching ingestion fall over. Give it real memory from day one, or use Cloud.
Umami's simplicity is a ceiling, not just a floor. If you outgrow pageviews and start needing funnels and cohorts, do not bolt them onto Umami. That is your signal to move up to a product-analytics tool.
Both keep your data yours. The whole point of self-hosting either is that visitor and user data lives on infrastructure you control, not a third party's.
The bottom line
Once you decide whether you need web analytics or product analytics, the tool picks itself. Match it to the question you are actually asking, and there is barely a decision left to make.
Deploy either one in a click on Elestio: Umami if you want simple and private, PostHog if you want the full product-analytics suite without hand-assembling ClickHouse and Kafka yourself.
Thanks for reading ❤️ See you in the next one 👋