PostHog review: analytics, session replay, and the free tier reality

Filed under Review

By Gerald · 14 August 2026

A desk with a calculator, a pencil, and printed pie and bar charts, representing product analytics and the numbers behind a usage bill

PostHog is a genuinely good bundle of product analytics, session replay, and feature flags for a small team, and the free tier is real. The catch is that the three products do not consume that free tier at the same rate, and almost nobody reads that part of the pricing page before they turn replay on for everyone.

I run PostHog on Flow. This review covers what it actually does well, where the free allowance runs out faster than people expect, and what self-hosting really involves if you go that route instead of the cloud plan. I checked current numbers on PostHog's own pricing page while writing this, in August 2026.

PostHog earns its reputation for event tracking and feature flags. Just watch session replay specifically, because it is the product most likely to put you on a bill you did not plan for.

What you get in one tool

PostHog's pitch is that you do not need four separate vendors for analytics, replay, feature flags, and experiments. One SDK, one dashboard, and every product shares the same underlying event data, so a feature flag rollout and the conversion metric it affects sit in the same place instead of two disconnected tools you have to cross-reference by hand.

That shared-data idea is the actual differentiator, not any single feature. Google Analytics does not do session replay or flags at all. A dedicated flag tool like LaunchDarkly does not know anything about your funnel data. PostHog's advantage is that a flag change and the metric it moves are queryable against each other natively.

The setup itself is straightforward. Drop the JavaScript snippet in, or install the SDK for your framework, and events start flowing within minutes. There is no lengthy schema design step before you see your first dashboard, which matters when you are trying to decide in an afternoon whether a tool is worth adopting at all.

Product analytics in practice

A retro control room with rows of desks, monitors, and wall panels under a grid of ceiling lights, representing a dashboard tracking many data streams at once — One event stream feeds analytics, replay, and feature flags at once, which is the part worth understanding before you turn all three on.

The core analytics product covers events, funnels, retention, and dashboards, and it does this part well. Autocapture picks up clicks and page views without instrumenting every button by hand, though I still recommend naming your key conversion events explicitly rather than relying on autocapture alone, because autocaptured event names are messier to build a funnel from later.

The free tier includes the first 1 million events a month at no cost, according to PostHog's current pricing page. Past that, pricing is volume-based and declines per event as you use more, starting around $0.00005 per event in the first paid tier and dropping well below that at high volume. For a small product, that first million covers a real amount of usage before you see a bill.

Session replay, and what it really costs you

This is the section that actually matters if you are deciding whether to turn PostHog on for a small product. Session replay is billed separately from events, and the free allowance is much smaller: the first 5,000 recordings a month, per PostHog's pricing page as of this writing. Past that, replay is priced per recording, not per event, starting at roughly $0.005 per recording in the first paid tier, with mobile session replay priced separately and starting higher.

Five thousand recordings sounds like a lot until you realize a recording is created per session, not per user, and PostHog by default records a meaningful share of sessions rather than a tiny sample. A product with a few hundred daily active users can burn through the free replay allowance in a few weeks without anyone deciding to do that on purpose. If you want replay, turn on sampling from day one rather than recording everything, and check your usage dashboard weekly for the first month.

The lesson generalizes: treat event tracking and session replay as two different budgets inside PostHog, because they are, and the one that runs out first is not the one people expect.

Feature flags and experiments

Feature flags share the same free allowance structure: the first 1 million flag requests a month are free, then volume pricing kicks in starting around $0.0001 per request and dropping at scale, according to PostHog's pricing page. What makes this genuinely useful rather than a checkbox feature is that a flag's rollout percentage and the metric you care about live in the same analytics engine, so you can build an experiment and read its result without exporting anything to a spreadsheet.

For a small team, this is the part of PostHog that replaces a separate paid tool most convincingly. A standalone feature flag service is a real recurring cost for something PostHog includes in the same free event allowance.

Experiments build directly on flags: you define a flag with variants, PostHog assigns visitors, and the experiment results page reads straight from the same event stream you are already capturing. There is no separate export-and-import step to run a basic A/B test, which is the part that usually eats an afternoon with other tools.

Self-hosting: possible, not easy

PostHog's core is MIT licensed and there is an official open-source deployment path, which matters if data residency or avoiding a per-event bill is the priority. In practice, the self-hosted "hobby" deployment via Docker Compose is explicitly aimed at up to about 100,000 events a month on a single machine, per PostHog's own documentation, and PostHog has sunset support for self-hosted Kubernetes deployments. Past hobby-project scale, self-hosting means running and scaling the ClickHouse-backed pipeline yourself, with no commercial support from PostHog if something breaks.

That is a fair tradeoff to know going in. Self-hosting solves a genuine ownership or compliance need, but it is not a free version of the managed product at production scale. If you need meaningfully more than 100,000 events a month self-hosted, budget real ops time for it, not just a docker-compose file.

On privacy specifically, PostHog does not store data in cookies or local storage for a visitor until consent is given, according to PostHog's own privacy documentation, and it supports a fully cookieless tracking mode. That gets you most of the way to a defensible GDPR setup, but you are still the one responsible for showing a consent banner and configuring the mode correctly. PostHog provides the mechanism; it does not make the legal decision for you.

Who should use PostHog and who is better served elsewhere

PostHog is the right call if you actually want analytics, replay, and flags working off the same data, and you are comfortable being deliberate about session replay sampling from the start. It is a strong replacement for stitching together three separate vendors when you are a small team without the headcount to manage that integration work.

It is the wrong call if all you want is simple, privacy-respecting pageview counts, where a lighter tool without PostHog's surface area is a better fit, or if you need production-scale self-hosting without an ops person to run it. For the rest of what I run Flow on, the full stack I use covers where PostHog fits alongside Vercel, Convex, and Stripe, and developer tools without subscriptions is a useful companion if PostHog's usage-based model is not what you are looking for at all.

Frequently asked questions

Is PostHog free for small projects? Yes, up to real limits. The free tier includes the first 1 million events, the first 5,000 session recordings, and the first 1 million feature flag requests each month, per PostHog's current pricing page. A small product can genuinely stay free for a while, especially if replay is sampled rather than left on for everyone.

How much does session replay cost on PostHog? The first 5,000 recordings a month are free. Past that, pricing starts at roughly $0.005 per recording in the first paid tier, with mobile replay priced separately and higher. Replay is the product most likely to exceed the free tier first, since a recording is created per session rather than per event.

Can I self-host PostHog for free? Yes, the core product is MIT licensed and there is a free Docker Compose deployment path. It is documented as suited to hobby-scale usage, around 100,000 events a month, and PostHog no longer supports self-hosted Kubernetes deployments, so scaling past hobby volume is on you.

Is PostHog GDPR compliant without a cookie banner? PostHog does not store tracking data in cookies or local storage until a visitor consents, and it supports a cookieless mode, according to its own privacy documentation. That is a genuinely privacy-friendly default, but you still need to implement and configure consent handling correctly yourself; the tool provides the mechanism, not the compliance decision.

Does PostHog replace Google Analytics? For most product-focused teams, yes, and it does more: funnels, replay, and flags on top of basic pageviews and events that Google Analytics also covers. If all you need is simple traffic counts with no product data, a lighter analytics tool may be less to configure.

My verdict

PostHog earns the recommendation. Just treat session replay as its own budget from day one, sample it rather than recording everything, and self-host only if you are ready to actually run the pipeline once you outgrow the hobby scale.