PostHog vs Plausible vs GA4 for SaaS (2026)

Most founders install Google Analytics on launch day because it's free and everyone has it. Three months later it can say how many people visited the pricing page, but not whether anyone who signed up last week used the product — and the cookie banner it needs is hiding some of the visitors it was meant to count. Choosing analytics for a SaaS is really two decisions, and in the UK, a consent decision too.

Which analytics tool should a SaaS startup use?

Use PostHog for product analytics inside your app and a cookieless web analytics tool — Plausible, or Vercel Analytics if you host on Vercel — for your marketing site. Add GA4 only when you run Google Ads or a marketing team already reports from it.

The three aren't straight rivals, because they answer different questions. Web analytics counts anonymous visitors: how many, from where, landing on which pages, converting to sign-up at what rate. Product analytics follows signed-in users through your app: which features they use, where they drop out of a funnel, whether they come back in week four — with session replay and feature flags on top. Plausible is purely the first. PostHog is built for the second, with a web analytics view included. GA4 tries to do both, but is designed above all to feed Google's advertising products. Most SaaS teams need one of each, and the handover point is the sign-up button.

PostHog vs Plausible vs GA4 compared

Prices are as of October 2026 and vendors change them, so check PostHog's pricing and Plausible's plans before you commit.

For UK founders, the cookies and hosting rows matter most: Plausible wins both by design, and PostHog can match it with its EU cloud and cookieless mode.

How much do PostHog, Plausible and GA4 cost?

At MVP scale, PostHog and GA4 cost nothing and Plausible costs from $9 a month after a 30-day trial — but the cheapest invoice isn't the cheapest choice once you count consent and setup time.

PostHog gives each product a free allowance that resets monthly. The free plan includes one project and a year of data retention, and simply stops collecting at the limits, so you can't be billed by surprise. Add a card and you keep the allowance, gain six projects and seven-year retention, and pay from $0.00005 per event between 1M and 2M, falling with volume. Set a billing limit per product, and watch autocapture: every click it records is a billable event.

Plausible has no free plan. After the trial, Starter is $9 a month (one site, three years of retention), Growth $14 (up to three sites) and Business $19 (up to ten sites, five years of retention, funnels, custom properties and the Stats API) — all at 10k monthly pageviews, rising with traffic, with two months free on yearly billing. Want a landing-page-to-sign-up funnel on the marketing site? Budget for Business.

GA4 is free; Analytics 360 is Google's enterprise edition, with no public price. GA4's real costs are consent setup and the hours spent bending its reports to SaaS questions.

Do you need a cookie banner for analytics in the UK?

Not for genuinely aggregate, cookieless analytics like Plausible; yes for session replay, per-user tracking and anything that feeds advertising — which covers PostHog's full feature set and almost every real-world GA4 setup.

The rule is regulation 6 of PECR: you can't store or read information on a user's device — cookies, localStorage, fingerprinting — without consent, unless an exception applies. The Data (Use and Access) Act 2025 rewrote that rule and added a "statistical purposes" exception, in force since 5 February 2026, and the ICO published its final guidance on 29 April 2026. The exception covers analytics only when the sole purpose is aggregate statistics to improve your service, any provider acts purely as your processor, and users get clear information plus a simple, free way to object. The ICO is explicit that recordings of individual visitors, tracking or profiling individuals, sharing conversions with ad partners and any advertising purpose still need opt-in consent.

Applied to the three tools:

• Plausible sets no cookies and keeps no persistent identifiers. It counts visitors with a hash of IP address, user agent and domain plus a salt deleted every 24 hours, and never stores raw IPs. Mention it in your privacy notice and offer an opt-out.
• PostHog keeps an ID in a cookie and localStorage by default, so treat identified users and session replay as consent-based. Setting cookieless_mode: "always" stores nothing on the device and counts users with a daily-salted server-side hash, losing replay, surveys and identity beyond a day. "on_reject" runs fully for people who accept your banner and counts the rest cookielessly. Both need "Cookieless server hash mode" switched on in project settings.
• GA4 relies on cookies and is built around Google Ads, so it needs a banner connected to Consent Mode. That passes four signals — ad_storage, analytics_storage, ad_user_data, ad_personalization — in basic mode (tags blocked until consent) or advanced mode (cookieless pings while consent is denied, used to model conversions).

PECR is only half of it. Analytics about identifiable users is personal data under UK GDPR, so you need a DPA with the vendor and a covered transfer route. EU hosting sidesteps most of that, because UK adequacy regulations cover the EEA — and new PostHog projects in EU organisations discard IP addresses by default. Our UK GDPR guide for SaaS startups covers the rest.

So the most affordable compliant setup is also the simplest: cookieless web analytics on the marketing site, PostHog on EU hosting with consent in the app, and nothing that feeds ad networks until you actually buy ads. This is general guidance, not legal advice.

What should an MVP actually track?

Five or six events, not five hundred: sign-up, your activation event, two or three core actions and upgrade — each named consistently and tied to a user.

Pageviews mislead early-stage SaaS because they measure curiosity, not value. Say a launch post brings 3,000 visitors: the chart looks like traction, but what matters is whether the 40 who signed up did the thing your product exists for, and came back the week after. A starter event plan:

• user signed up, carrying the source and campaign from the marketing site, so you learn which channels produce users who stay — the input for getting your first 100 users.
• Your activation event: the one action that shows a user got value — first report generated, first invoice sent, first teammate invited. Define it before launch; it's the number your first months are judged on.
• Two or three core actions for the features you believe are the product, such as report exported or integration connected.
• subscription started and subscription cancelled, sent server-side from your Stripe webhook so ad blockers can't drop them:

import { PostHog } from 'posthog-node'

const posthog = new PostHog(process.env.POSTHOG_KEY!, {
  host: 'https://eu.i.posthog.com',
  flushAt: 1, // send immediately on serverless
  flushInterval: 0,
})

posthog.capture({
  distinctId: userId,
  event: 'subscription started',
  properties: { plan: 'pro', interval: 'month' },
})
await posthog.shutdown()

Then watch one funnel (sign-up to activation) and one retention chart, weekly. With dozens of users rather than thousands, analytics shows you where to look and conversations tell you why — the rhythm we set out in the first 90 days after an MVP launch.

When does GA4 still make sense?

When you're buying Google Ads, or someone else's reporting already depends on it. Three situations justify it:

• Paid search is a real channel. GA4 links natively to Google Ads for conversion import, audiences and attribution — with consent, since advertising is never exempt.
• A marketing team or agency already reports from it.
• You want raw data in a warehouse cheaply. The BigQuery export is free to set up and handles up to 1M events a day on a standard property.

What it's poor at is the product: no session replay or feature flags, funnels that reach back 14 months at most on the free tier, and people who decline cookies modelled or missing. GA4 beside PostHog is fine; GA4 as your product analytics is not.

Our default: PostHog in the app, Plausible on the site

Analytics follows the SaaS MVP tech stack we ship with — Next.js, Postgres, Clerk, Stripe and Resend on Vercel — with PostHog for product analytics and Plausible or Vercel Analytics for the marketing site.

In practice: PostHog on the EU cloud, loaded but opted out until the user consents, and identified with their auth user ID at sign-up so anonymous browsing joins the account. Billing events come server-side from Stripe. The marketing site runs Plausible, or Vercel Analytics when the client wants one fewer vendor, and passes the visitor's source through to sign-up. Choosing the activation event is a product decision, not a technical one, so we settle it with the founder during product strategy work.

We deviate in three cases. When the client runs Google Ads, GA4 joins the marketing site behind a consent banner. When the client wants one tool and no banner, we run PostHog alone in cookieless mode and give up session replay. When a contract demands analytics on the client's own servers, we self-host — otherwise never.

Frequently asked questions

Does Plausible need a cookie banner in the UK?

Plausible's own position is no: it sets no cookies, stores nothing on the visitor's device and is designed so that no consent banner is required. Aggregate analytics like this is what the statistical purposes exception, in force since February 2026, was written for — provided your privacy notice mentions it and people have a simple way to object. You still need a banner if anything else on the site, such as an ad pixel, sets non-essential cookies.

Can PostHog replace Google Analytics?

For most SaaS companies, yes — PostHog includes a web analytics dashboard billed as ordinary events, so one tool can cover the marketing site and the app. The case for a separate site tool is consent: full PostHog uses cookies, Plausible doesn't. What PostHog doesn't replace is GA4's Google Ads integration, so if paid search drives your pipeline, keep GA4 alongside it.

Is it legal to use GA4 in the UK?

Yes — with a consent banner, a privacy notice that discloses it and tags that don't fire before consent. Google says it doesn't log or store UK users' IP addresses, performing IP lookups on UK, EU or Swiss servers before forwarding data for processing. Non-compliant setups usually fail on implementation: tags firing early, a reject option harder to find than accept, or ads features nobody consciously switched on.

Should you self-host your analytics?

Rarely at MVP stage — both PostHog and Plausible are open source, but self-hosting moves patching, backups and scaling onto your team. PostHog's self-hosted deploy is officially unsupported and wants a server with 4 vCPUs and 16GB of RAM; Plausible's Community Edition is updated twice a year and lacks funnels. Either vendor's EU cloud solves data residency for far less effort.

Want analytics that answer real questions from day one?

The event plan is cheapest to get right before the first line of tracking code — activation event, consent banner and billing events included. Book a free scoping call and we'll map the analytics around your product, then quote the build fixed-price.