GA4 vs PostHog: when to use Google Analytics, PostHog, or both
GA4 vs PostHog compared on data model, funnels, session replay, experiments, pricing, EU hosting and SQL. When to pick GA4, PostHog, or both.
“PostHog or Google Analytics?” sounds like a question about which tool is better. It is really a question about what you need to measure. GA4 is built around marketing: where people came from, what a campaign cost, what it brought in. PostHog is built around the product: what individual users do, where they get stuck and whether they come back.
A disclosure first: we build Kaen, who works with both tools, so we have no reason to push either one. Our own site, kaen.cz, runs on PostHog (EU cloud). Everything we say about the two tools is based on their official documentation and pricing pages as of September 2026.
GA4 vs PostHog at a glance
| GA4 | PostHog | |
|---|---|---|
| Best for | Marketing sites, e-commerce, lead gen | Products, SaaS, logged-in apps |
| Data model | Events with parameters; sessions derived from the session_start event | Events; identified events attach to a person profile |
| Funnels | Funnel exploration | Funnels with time to convert, historical trends and click-through to replays |
| Retention | Cohort exploration, user lifetime | Retention insights, cohorts |
| Session replay | Not included | Included (5,000 free recordings a month) |
| Feature flags and experiments | Not included (Google Optimize shut down on September 30, 2023) | Included; every experiment is backed by a feature flag |
| Google Ads | Native link: conversions from key events, audiences, Ads data inside GA4 | Experimental destination for sending conversions to Google Ads, not officially supported |
| Price | Free; Analytics 360 is paid (pricing via sales) | Monthly free tier, then usage-based |
| Data retention | Explorations: 2 or 14 months (360: up to 50 months) | Events: 1 year on free, 7 years on paid plans |
| EU hosting | EU traffic is collected on EU servers, then forwarded to Analytics servers for processing; IPs used only for geolocation | EU cloud in Frankfurt, or self-host |
| Sampling and limits | Explorations sampled above 10M events per query; 25 parameters per event | The main limit is cost per volume; billing limits available |
| Consent and cookieless | Consent mode; advanced mode sends cookieless pings and models the gaps | cookieless_mode (always or on_reject), users counted with a server-side hash |
| Self-host | No | Yes, hobby deploy via Docker Compose (MIT), no support |
| SQL | BigQuery export (standard: 1M events daily batch limit) | SQL inside the app (HogQL on ClickHouse), plus batch export to BigQuery |
One note on retention: GA4’s retention setting does not touch standard aggregated reports. The 2- or 14-month window applies to explorations and funnel reports — exactly where funnel analysis happens.
When GA4 is the right call
GA4 makes sense when your site is mainly a store or a set of landing pages and you buy traffic to it.
- Google Ads is your main channel. Once linked, you create Google Ads conversions from GA4 key events, see campaign performance inside Analytics and send audiences to remarketing. On the PostHog side, Google Ads is an experimental destination PostHog doesn’t officially support.
- You don’t need to know individual users. You want to know which channels and pages sell, not what one person did inside an account.
- You want zero tool cost. Standard GA4 is free. The BigQuery export itself is free too; you only pay BigQuery for storage and queries.
What to watch: explorations in the standard version are sampled above 10 million events per query, and data for explorations and funnels is kept for 2 or at most 14 months. There is no session replay and no A/B testing in GA4, so you need another tool for those (see Microsoft Clarity vs. Hotjar). How to get a usable funnel out of GA4 is covered in funnel analysis in GA4.
When PostHog is the right call
PostHog fits when your product is an app and what matters is what logged-in users do in it.
- You need a person view, not just a session view. Identified events attach to a person profile, which unlocks segmentation by user properties, cohorts and feature flag targeting.
- You want the funnel, the replay and the test in one place. From a funnel you click straight into recordings of the people who dropped off. Then you test the hypothesis with an experiment on a feature flag, using Bayesian or frequentist statistics.
- You want the data in the EU. PostHog Cloud EU runs on servers in Frankfurt. If your requirements are stricter, you can self-host, but PostHog offers no support for self-hosted instances and recommends Cloud above 1 million events a month.
- You want SQL without another tool. You query events, persons and connected data warehouse sources directly in PostHog.
What to watch: you pay for volume. The first million events a month are free; the first paid band starts at $0.00005 per event and the price drops as volume grows. PostHog says anonymous events can be up to 4x cheaper than identified ones, which is why the web SDK captures anonymous events until you call identify(). Set a billing limit on day one. Replays are kept for 30 days on the free plan and 90 days on pay-as-you-go.
When to run both
For companies with a marketing site and a logged-in product, running both is usually the sensible setup:
- GA4 for marketing: traffic sources, campaigns, Google Ads conversions, the path to sign-up.
- PostHog for the product: onboarding, activation, feature usage, retention, experiments.
To make it work, decide upfront which tool owns which metric. Sign-ups measured in GA4 and in PostHog will rarely match exactly, because each tool counts users differently and handles cookie consent differently. If you haven’t split ownership, meetings turn into debates about which number is right instead of what to do about it.
What to watch when you switch
- History doesn’t move by itself. PostHog’s migration path is a GA4 export to BigQuery, renaming events (for example
page_viewto$pageview) and importing through an SDK. Plan for a period where both tools run in parallel while you compare numbers. - Name your events first. A migration is the moment to clean up your taxonomy. Write down which events make up your funnel before you implement anything.
- Decide when to identify users. Calling
identify()changes both the bill and what you can see. A sensible default: identify at sign-up or login. - Handle consent. GA4 uses consent mode, PostHog uses
cookieless_mode. In cookieless mode you shouldn’t callidentify(), andaliasevents are dropped. - Don’t forget Google Ads. Switch GA4 off entirely and you lose the native conversion import. Either keep GA4 for ads, or plan a replacement and confirm conversions still reach Ads.
- Check retention. In GA4, make sure the setting is 14 months, not 2. In PostHog, check what your plan covers.
The tool is only half of it
Whichever you choose, a well-configured tool improves nothing on its own. Funnel leaks last for months not because analytics can’t show them, but because nobody looks at the funnel on a regular schedule. Replays nobody watches and funnels nobody opens are worth nothing.
This is where Kaen comes in, so the disclosure again: we build him. Kaen reads GA4 or PostHog through read-only OAuth. He proposes a funnel from your events, checks it daily and hunts for lost revenue weekly. He prices every finding in CZK and proposes a fix or an A/B test. If tracking is missing or broken, Foundation audits and sets up GTM, GA4 and events first. Nothing ships without your approval.
Frequently asked questions
Is PostHog a good alternative to GA4?
For a product or SaaS, often yes: PostHog puts funnels, retention, session replay, feature flags and experiments in one tool, and you can keep the data in its EU cloud. For a site that lives on Google Ads, GA4 stays more practical, because it links to Google Ads natively and turns key events into Ads conversions without extra plumbing.
Is PostHog free?
It has a monthly free allowance. According to PostHog’s pricing page in September 2026, that is 1 million product analytics events, 5,000 session recordings and 1 million feature flag requests. Above that you pay by volume, and you can set a billing limit.
Can I run GA4 and PostHog at the same time?
Yes, and it is common for companies with a marketing site and a logged-in product: GA4 covers acquisition and campaigns, PostHog covers product behavior. Decide upfront which tool is the source of truth for which metric, or the numbers will drift apart in every meeting.
Does GA4 sample data?
In standard GA4, explorations are sampled when a query exceeds 10 million events. Analytics 360 raises the quota to up to 1 billion events per query. When sampling applies, the data quality icon shows what percentage of the data the result is based on.
The takeaway
GA4 is the better fit for a marketing site or store that runs on Google Ads. PostHog is the better fit for a product where you need individual users, replays and experiments, ideally with data in the EU. A company with both usually runs both, with a clear split of what each one measures. If you’re also weighing session replay tools, continue with Hotjar alternatives.
The outcome depends less on the tool than on the routine. If nobody looks at your funnel every week, that is the job Kaen does: he reads GA4 or PostHog, finds where people drop off and proposes the fix or the test. Nothing ships without your approval.