PostHog

Product analytics, session replay, feature flags, experiments — and a dozen more tools, honestly reviewed. No marketing fluff, just what you actually get when you self-host it.

TL;DR

• What it is: An open-source developer platform that bundles product analytics, web analytics, session replay, feature flags, A/B testing, error tracking, surveys, a data warehouse, and more under a single SDK [1][3].
• Who it’s for: Product engineers and technical founders who want tight feedback loops between shipping and learning, without stitching together five separate SaaS subscriptions [1][3].
• Cost savings: Mixpanel and Amplitude both start around $28–$61/mo and scale aggressively with monthly tracked users. PostHog Cloud offers 1 million events/mo free and usage-based pricing after that. Self-hosted runs on your own infrastructure at whatever a VPS costs you.
• Key strength: Breadth. One SDK captures analytics, session recordings, feature flag evaluations, and errors simultaneously — meaning one integration replaces Mixpanel + Hotjar + LaunchDarkly + Sentry for many teams [1][3].
• Key weakness: The self-hosted “hobby deploy” is explicitly not for production. Running PostHog properly at scale means operating ClickHouse + Kafka + PostgreSQL + Redis — a significant infrastructure commitment most non-technical founders shouldn’t attempt [GitHub README].

---

What is PostHog

PostHog started as a product analytics tool and has spent the last four years absorbing adjacent categories. The GitHub README now lists twelve distinct capabilities: product analytics, web analytics, session replays, feature flags, experiments, error tracking, surveys, a data warehouse, data pipelines, LLM analytics, workflows, and an AI assistant. The company describes it as “an all-in-one developer platform for building successful products” — which is ambitious to the point of straining credibility, but the feature list is real [GitHub README][1].

What makes the pitch coherent is the single SDK. You instrument once, and every tool in the stack draws from the same event stream. When a user hits an error, you can immediately jump to their session recording to see what they were doing. When a feature flag experiment finishes, the analytics to evaluate it are already there. That tight coupling between capture and analysis is the actual product thesis — not “another analytics tool,” but a closed loop between shipping and learning [1][3].

The project sits at 32,088 GitHub stars. It is backed by a real company (Y Combinator-backed, Series C) with a commercial cloud offering, but the core codebase is open source. The license situation is mixed — the platform core uses an open-source license, while certain enterprise features are proprietary. The GitHub repository shows “NOASSERTION” for license detection, which typically signals a multi-license structure [GitHub repository].

PostHog Cloud is available in US (Virginia) and EU (Frankfurt) regions, which matters for GDPR compliance [homepage].

---

Why people choose it

The 200 Product Hunt reviews aggregate to a 5.0 rating, which is unusually high. The most frequently cited reasons: detailed analytics, session recording, open source, developer-friendly design, feature flags, and easy integration [3]. That’s not a random cluster — it’s the exact feature set that would let a technical team cancel three or four separate SaaS subscriptions.

Versus Mixpanel and Amplitude. The traditional product analytics tools have event-based pricing that gets expensive fast, no session replay, no feature flags, and no experiments without add-ons or additional tools. A typical growth-stage startup running Mixpanel + Hotjar + LaunchDarkly is spending $200–$500/mo or more. PostHog’s free tier — 1 million events, 5,000 session recordings, 1 million feature flag requests per month — covers most startups entirely [homepage][3]. Tom Snyder’s review at marketertools.co puts it plainly: PostHog “consolidate[s] multiple point tools into one platform” [1].

Versus Google Analytics 4. GA4 is free and handles web traffic well, but it’s not a product analytics tool in any serious sense. It doesn’t do session replay, feature flags, or experiments. For teams that need to understand product behavior rather than just marketing traffic, GA4 is not the comparison. PostHog’s web analytics dashboard is actually the weakest module in the suite — it exists because users asked for it, not because PostHog beats GA4 at SEO dashboards [5].

The engineer-first design philosophy. Multiple reviewers call out that PostHog doesn’t use dark patterns, doesn’t hide pricing, doesn’t require a sales call, and treats developers as intelligent adults [1][3]. The Product Hunt community specifically mentions “developer-friendly” as a top differentiator. PostHog publishes its company handbook, sales manual, and strategy publicly — this is unusual for a SaaS company and signals genuine commitment to transparency [homepage].

The non-technical caveat. Product Hunt reviewers are honest about the downside: “the UI could be more polished” and it “can feel heavy for non-technical teams” [3]. Marketertools.co notes “the expanding feature set may overwhelm smaller teams” and that “non-technical users might find it less accessible than alternatives” [1]. PostHog is explicitly built for product engineers — if your primary analytics user is a non-technical marketer, this probably isn’t the right tool.

Self-use signal. PostHog uses PostHog to run their own marketing. Their marketing team published a detailed post showing how they track new org signups, organic SEO, content performance, circuit breaker metrics, and attribution — all in PostHog dashboards. That includes using feature flags and experiments to iterate their own pricing page [5]. A company that eats its own cooking at this level is a credibility signal.

---

Features

Core analytics:

• Event-based product analytics with autocapture (instruments clicks, form submissions, page views automatically without custom events) [GitHub README]
• Funnels, retention, user paths, cohorts, and SQL query editor for power users [GitHub README][1]
• Web analytics dashboard tracking traffic, sessions, conversion, web vitals, and revenue [GitHub README]
• Annotations to mark deploy dates, experiments, pricing changes — avoids the “why did this drop six months ago” problem [5]

Session intelligence:

• Session replays for web and mobile with heatmaps [GitHub README]
• Direct navigation from an error to the user session that triggered it [1]
• Rage clicks, dead clicks, scroll depth — standard replay toolkit [GitHub README]

Experimentation:

• Feature flags with percentage rollouts, cohort targeting, and local evaluation [GitHub README]
• A/B and multivariate experiments with statistical significance tracking [GitHub README]
• No-code experiment setup option [GitHub README]

Error tracking:

• Captures errors, surfaces alerts, tracks resolution [GitHub README]
• Cross-referenced with session replays and analytics [1]

Data infrastructure:

• Data warehouse that syncs from Stripe, HubSpot, external warehouses, and 120+ other sources [homepage]
• Data pipelines with 25+ destinations and custom webhook support [GitHub README]
• SQL editor + BI visualization on top of unified data [homepage]

AI:

• LLM analytics for apps using LLMs — traces, generations, latency, cost tracking [GitHub README]
• PostHog AI assistant that answers questions about product usage and will eventually suggest code fixes [homepage]
• Workflows for automated messaging and user actions [GitHub README]

Surveys:

• No-code survey builder with templates [GitHub README]
• In-product NPS, CSAT, feature feedback — tied to the user records in analytics [GitHub README]

---

Pricing: SaaS vs self-hosted math

PostHog Cloud:

• Free tier: 1 million events/mo, 5,000 session recordings/mo, 1 million feature flag requests/mo, 1 million data warehouse rows/mo [homepage]
• After free tier: $0.00005/event, $0.005/recording, $0.0001/flag request, $0.000015/warehouse row (pricing decreases with volume) [homepage]
• No seat-based pricing, no “call sales” tier for standard usage

PostHog claims 98% of customers use it entirely within the free tier. That’s plausible — 1 million events per month is more than most early-stage startups consume.

Self-hosted:

• Software: free (open source core)
• Infrastructure: see Deployment section — this is not a $5 VPS situation

Mixpanel for comparison:

• Free: up to 20M events/mo (generous, but no session replay or flags)
• Growth: starts at $28/mo for 1,000 monthly tracked users, scales with MTU count
• Enterprise: custom pricing, typically $500–$2,000+/mo for larger teams

Amplitude for comparison:

• Free: up to 50K monthly active users (but limited features)
• Plus: $61/mo
• Growth/Enterprise: custom, typically $1,000+/mo

The full-stack replacement math: Say you’re running Mixpanel ($100/mo) + Hotjar ($99/mo) + LaunchDarkly ($75/mo) + a basic error tracker ($29/mo) = $303/mo. PostHog Cloud covers all four categories and the first year you’re almost certainly in the free tier. Even at modest growth, you’re looking at $20–$80/mo in PostHog usage versus $300+/mo across four tools [1][3].

That math is why PostHog’s growth has been rapid. The individual tools it replaces are each defensible products — the bet is that the integration value of having them all on one event stream outweighs best-in-class depth in any single category.

---

Deployment reality check

This is where the honest review diverges sharply from the marketing. PostHog’s README explicitly says:

“The fastest and most reliable way to get started with PostHog is signing up for free to PostHog Cloud”

The self-hosted “hobby deploy” is labeled as not for production. PostHog’s full infrastructure stack includes ClickHouse (columnar database for analytics at scale), Kafka (event queue), PostgreSQL (application database), Redis (cache), and a Django application server. This is not a docker-compose up that you run on a $5 VPS and forget about [GitHub README].

What the hobby deploy actually is: A Docker Compose configuration for local development or small-scale personal use. It works. But the moment you start collecting meaningful event volume, ClickHouse’s memory requirements grow fast. The README warns this is not supported for production workloads.

What production self-hosting actually requires:

• A Kubernetes cluster or multiple VMs to run each component
• Operational knowledge of ClickHouse, which is not a beginner database
• PostHog’s own documentation says to use their managed cloud for teams without dedicated infrastructure engineers
• Minimum viable production setup: probably $100–$300/mo in cloud infrastructure, depending on event volume

Who can realistically self-host PostHog:

• A DevOps/infrastructure engineer who has run ClickHouse before
• A company with existing Kubernetes infrastructure and an ops team
• A team with strict data residency requirements that justifies the operational overhead

Who should not attempt to self-host PostHog:

• Non-technical founders (the intended audience of this site)
• Small teams without dedicated ops bandwidth
• Anyone who calls a VPS “a server” without knowing what ClickHouse is

The gap between “open source and self-hostable” (true) and “easily self-hostable by non-technical founders” (false) is significant. PostHog Cloud’s free tier is generous enough that for most unsubbed.co readers, it’s the correct choice — self-hosting the analytics infrastructure is a distraction from the product you’re actually building.

---

Pros and cons

Pros

• Single SDK replaces 4–5 tools. One instrumentation, unified data across analytics, replays, flags, experiments, error tracking, surveys [1][3]. This is the core value proposition and it’s real.
• Genuinely generous free tier. 1M events, 5K recordings, 1M flag requests per month for free — this covers most early-stage startups entirely [homepage].
• Usage-based pricing with no seat tax. You pay for data volume, not the number of people on your team viewing dashboards [homepage].
• EU Cloud option. Hosted in Frankfurt for GDPR compliance without self-hosting overhead [homepage].
• Transparent company. Public handbook, public pricing, public roadmap, no “contact sales” gating for standard features [homepage][1].
• Developer-friendly. Autocapture means less instrumentation work upfront. SQL access means power users aren’t blocked by a limited query builder [1][3][GitHub README].
• 32,000+ GitHub stars and active development — not a dormant project [GitHub].
• LLM analytics built in — relevant for the growing cohort of founders building AI-powered products [GitHub README].

Cons

• Self-hosting is not beginner-friendly. The “hobby deploy” is explicitly not for production. Real self-hosting requires ClickHouse, Kafka, and Kubernetes proficiency [GitHub README]. This is a significant caveat for a tool marketed under the self-hosted umbrella.
• Feature breadth creates UI complexity. Product Hunt reviewers note “UI could be more polished” and the platform can overwhelm non-technical users [3]. Twelve product categories in one interface is a lot of surface area.
• Not the best-in-class at any single thing. Hotjar beats it on UX research features, Amplitude beats it on analytics depth, LaunchDarkly beats it on enterprise feature flag management. The bet is integration value over depth — a real trade-off [1].
• Autocapture creates noise. The automatic event capture is convenient but generates large volumes of events that require cleanup and filtering before the data is useful [1].
• Open-source license is mixed. The NOASSERTION license detection on GitHub signals a multi-license structure — enterprise features are proprietary, not open source [GitHub repository].
• Heavy for what it is. Running ClickHouse and Kafka for a 10-person startup is overengineered infrastructure. You’re renting a data warehouse when you might need a filing cabinet.

---

Who should use this / who shouldn’t

Use PostHog if:

• You’re a technical founder or product engineer who instruments your own analytics.
• You’re currently paying for multiple tools (Mixpanel + Hotjar + LaunchDarkly + something else) and want to consolidate.
• You want to run experiments and tie results directly to the same analytics where you track user behavior.
• You’re building an LLM-powered product and want trace-level visibility into model usage and cost.
• GDPR is a concern and you want EU-hosted infrastructure without operating it yourself.

Use PostHog Cloud specifically (not self-hosted) if:

• You’re a non-technical founder who just needs analytics to work.
• You don’t have infrastructure engineering bandwidth.
• The free tier covers your volume — it probably does.

Consider self-hosting only if:

• You have a dedicated DevOps engineer or infrastructure team.
• Data residency requirements prohibit third-party cloud, even EU-hosted.
• Your event volume is high enough that cloud costs become significant (millions of events/day).

Skip it (use Plausible or Umami) if:

• You only need web traffic analytics — page views, referrers, countries, devices.
• You don’t track product events at all.
• You want a 5-minute Docker Compose install that a non-technical person can maintain.

Skip it (use Mixpanel or Amplitude) if:

• Your primary analytics user is a non-technical marketer or product manager who needs a polished, purpose-built UI.
• You don’t need session replay, feature flags, or experiments.
• You want best-in-class funnel analysis without learning a 12-product platform.

Skip it (stay on separate best-in-class tools) if:

• Your engineering team has strong opinions about LaunchDarkly, Sentry, and Amplitude separately, and integration overhead is not your bottleneck.
• You need enterprise-grade feature flag governance that PostHog’s proprietary tier doesn’t fully cover.

---

Alternatives worth considering

• Mixpanel — purpose-built product analytics, polished UI, better for non-technical PMs, no session replay, usage-based pricing by MTU.
• Amplitude — more enterprise-grade analytics depth, better data governance, no replay or flags, expensive at scale.
• Plausible — lightweight, privacy-first web analytics only. Self-hostable on a single VPS. No product analytics, no flags. For traffic dashboards, dramatically simpler.
• Matomo — Google Analytics replacement for web traffic. Self-hostable, GDPR-friendly, no product analytics depth or feature flags.
• Hotjar — session replay and heatmaps done better than PostHog. No analytics or flags. If replay is your only need, Hotjar is more mature.
• LaunchDarkly — dedicated feature flag management with enterprise controls. More mature flag governance than PostHog. Expensive.
• Flagsmith — open-source feature flags only, self-hostable on a single server, much simpler deployment than PostHog.
• GrowthBook — open-source A/B testing and feature flags, connects to your existing data warehouse. If you only need experiments, not the full stack.
• Sentry — dedicated error tracking, more mature than PostHog’s error module, better source maps and alerting. If errors are your primary concern.

For a non-technical founder trying to understand how people use their product: the realistic shortlist is PostHog Cloud free tier vs. Mixpanel free tier. Both are free at small scale. PostHog wins if you want session replay and will eventually need feature flags. Mixpanel wins if you want a cleaner analytics UI without learning twelve product categories.

---

Bottom line

PostHog is the right tool for a specific person: a technical founder or product engineer who instruments their own code and wants a single platform instead of four separate SaaS bills. The free tier is genuinely generous, the EU Cloud option handles GDPR without self-hosting complexity, and the integration between analytics, session replay, and feature flags is the real product value — not any individual module.

The honest caveat: the self-hosting story is significantly less accessible than tools like Plausible or Activepieces. “Open source and self-hostable” is true, but self-hosting PostHog in production requires ClickHouse and Kafka expertise that most non-technical founders don’t have and shouldn’t need for analytics. For the unsubbed.co target reader, PostHog Cloud’s free tier is almost certainly the right call — it’s not a SaaS bill you’re escaping, it’s a bill you probably won’t hit.

If you do eventually hit the free tier ceiling and want to self-host to control costs, that’s when you need an infrastructure engineer or a deployment service to handle the ClickHouse cluster. That’s exactly the kind of one-time setup work that upready.dev handles for clients.

---

Sources

1. Tom Snyder, marketertools.co — “PostHog: The Open-Source Product Analytics App That Wants to Steal Google Analytics’ Lunch” (January 11, 2026). https://www.marketertools.co/blog/posthog-review
2. Product Hunt — “PostHog Reviews (200 reviews, 5.0)”. https://www.producthunt.com/products/posthog/reviews
3. Product Hunt — “PostHog — Literally every piece of SaaS that a product engineer needs”. https://www.producthunt.com/products/posthog
4. Andy Vandervell, PostHog Blog — “How (and why) our marketing team uses PostHog” (September 28, 2022). https://posthog.com/blog/posthog-marketing

Primary sources:

• GitHub repository and README: https://github.com/posthog/posthog (32,088 stars)
• Official website: https://posthog.com
• Pricing page: https://posthog.com/pricing
• PostHog Cloud signup: https://us.posthog.com/signup