PostHog review for product analytics
PostHog review for product analytics
PostHog in 2026 is the right product analytics tool for engineering-led teams that want one bill instead of four (analytics, session replay, feature flags, surveys) and the option to self-host. It's the wrong pick for analyst-heavy orgs running >100M events per month, where Mixpanel and Amplitude still beat it on dashboard polish and PM-friendly exploration.
That's the verdict. Below is the honest version with real 2026 pricing, where the bundle pays for itself, where it stops paying, and what it actually costs to deploy correctly.
What PostHog actually is in 2026
PostHog is an open-source product analytics platform that has, over the last three years, quietly turned into nine products in one shell: product analytics, web analytics, session replay, feature flags, A/B experiments, surveys, heatmaps, error tracking, and a data warehouse layer with reverse-ETL.
The company's pitch is straightforward: you don't need Mixpanel for analytics, LaunchDarkly for flags, Hotjar for replay, Typeform for surveys, and a half-built error tracker on top. You need one tool that your engineering team owns, with HogQL (a SQL-like query language) underneath when you outgrow the dashboards.
You can run it two ways. PostHog Cloud is hosted in either US (AWS Virginia) or EU (Frankfurt) regions. PostHog Self-Hosted is the open-source build you run on your own Kubernetes cluster, which matters more in 2026 than it did pre-Schrems II.
PostHog isn't a marketing analytics tool. It doesn't compete with GA4 for SEO attribution or with Segment for CDP routing. It competes with Mixpanel, Amplitude, Heap, June, and Statsig for the question "what are users actually doing in our product."
PostHog 2026 pricing, in one table
PostHog's pricing is usage-based per product, and the page is more honest than most. Here's what you'll actually pay in May 2026, after the free tier:
| Product | Free tier (per month) | Starting paid rate | Volume floor |
|---|---|---|---|
| Product analytics | 1M events | $0.00005 / event | $0.0000090 / event at 250M+ |
| Session replay | 5,000 recordings | $0.0050 / recording | $0.0015 / recording at 500K+ |
| Feature flags | 1M requests | $0.0001 / evaluation | $0.000010 / request at 50M+ |
| Surveys | 1,500 responses | $0.10 / response | $0.010 / response at 20K+ |
| Error tracking | 100K exceptions | $0.00037 / exception | scales down |
| Data warehouse | 1M rows + free historical | $0.000015 / row | $0.000001 / row at 1B+ |
Three things to notice. First, the free tier is generous enough that PostHog claims more than 90% of accounts never pay. For a pre-PMF SaaS doing 200K monthly events, you're free for the foreseeable future. Second, the per-unit rate drops sharply as you scale, but the absolute bill still climbs (a 50M-event SaaS pays roughly $1,200 a month on analytics alone). Third, session replay is the line item that surprises people; mobile recordings are billed at roughly double web rates, so an unsampled mobile app can run $4-6K/month on replay alone.
Enterprise pricing is custom and adds SSO, SAML, dedicated support, and HIPAA BAAs. If you're asking, you already know.
Where PostHog wins
The all-in-one bundle math
This is the thesis. Run the math at 5M events per month with a typical SaaS replay sample rate (10% of sessions, ~15K recordings) and active feature flagging:
- Mixpanel Growth: ~$500/mo
- LaunchDarkly Pro: ~$650/mo (per-MAU pricing bites)
- Hotjar Business: ~$200/mo
- Sentry Team for error tracking: ~$100/mo
- Total: ~$1,450/mo across four contracts, four onboarding tickets, four SDK installs
PostHog at the same volume runs ~$300/mo all-in. The savings aren't the headline, the consolidation is. Your engineers maintain one SDK. Your data lives in one place. Your retention cohort can be filtered by feature flag exposure in a single query, which is genuinely hard when those tools are split.
The free tier is genuinely free
A real free tier with 1M events, 5K replays, 1M flag requests, 100K exceptions, 1,500 survey responses, and unlimited team members is unusual. Mixpanel's free tier is 1M events but caps users; LaunchDarkly's free tier caps you at 1,000 MAUs and one project. PostHog's free tier carries most pre-PMF SaaS for 12-18 months without surprise upgrades.
Self-host is the lock-in escape hatch
You can take the open-source build and run it on your own infrastructure. At >50M events per month, the math starts to favor self-host (figure $2-4K/mo of EKS plus Postgres plus ClickHouse plus an engineer who babysits it) over PostHog Cloud's per-event bill.
More importantly, self-host means your event data physically does not leave your VPC. Mixpanel can't offer that. Amplitude can't offer that without a six-figure private-cloud contract. For European fintech, healthcare, or anyone with a customer DPA that bans US data transfer, this is the answer.
EU data residency is real, not marketing
PostHog Cloud EU is in Frankfurt, runs as a separate tenant, and survives Schrems II audits. If your customers are German enterprises or EU-regulated, this isn't a nice-to-have. It's the reason you pick PostHog.
Where PostHog breaks
Dashboards still trail Mixpanel for non-engineers
This is the honest part. PostHog's product analytics interface in 2026 is good, but it's clearly designed by engineers for engineers. Building a multi-step funnel that a PM can self-serve, with cohort comparisons and a saved view your CEO can read on Monday, is faster in Mixpanel. The interface is calmer. The defaults are better.
If your product team has zero SQL fluency and zero patience for "where do I click to add a breakdown?", Mixpanel is still the easier sell. Same with our take on the best analytics tools for SaaS, where PostHog tops the engineering-led list and Mixpanel tops the PM-led list.
Scale gets expensive past 100M events/month
The per-event rate looks tiny ($0.00005), but multiply it across 100M events and you're at $5,000/mo on analytics alone, before replay or flags. By 500M events you're at $15-25K/mo. At that volume, you're either negotiating an Enterprise contract or self-hosting. Both are real choices, but neither is the cheap path the marketing implies.
Session replay mobile is double-rate
PostHog charges roughly 2x for mobile session recordings versus web. An unsampled iOS app with 30K weekly sessions burns through the free 5K recordings in a day. Always set a sample rate (PostHog's SDK supports per-platform rates) and exclude session replay from your highest-traffic auth flows.
HogQL needs SQL fluency
HogQL is PostHog's SQL-like query layer, and it's how you get past the dashboards into real analysis. It's powerful. It also requires that someone on your team can write a window function. If nobody can, you're stuck with the prebuilt insights, which are good but not unlimited.
PostHog vs Mixpanel vs Amplitude vs Statsig
Here's the honest landscape. The right tool depends on who runs analytics at your company, not on which is "best."
| Tool | Free tier | Paid starts | Best at | Weakest at |
|---|---|---|---|---|
| PostHog | 1M events + 5K replays + 1M flags | $0.00005 / event | All-in-one for engineer-led teams, self-host | Non-engineer dashboards |
| Mixpanel | 1M events | $24/mo + tiers | Polished funnels, PM self-serve | No flags, no native replay |
| Amplitude | 50K MTU | Custom enterprise | Analyst-led orgs, ML-driven insights | Cost, no flags |
| Statsig | 1M events | $0.00002 / event | Experimentation depth | Lighter replay/surveys |
If you're picking flags specifically, our roundup of the best feature flag platforms in 2026 walks through PostHog vs LaunchDarkly vs Statsig in more depth, and our take on the best A/B testing tools for SaaS covers PostHog Experiments vs Statsig vs Optimizely. If error tracking is your priority and analytics is secondary, see our review of the best error tracking tools for startups, where PostHog's exception tracking ranks well but doesn't beat Sentry for stack-trace depth.
Who should buy PostHog (and who shouldn't)
Buy PostHog if:
- You're an engineering-led SaaS at 0-50M events per month and you want one tool, one bill, one SDK.
- You ship feature flags weekly and want flag exposure data joinable to retention cohorts in one place.
- You're EU-regulated or you have a customer DPA that bans US data transfer.
- You'd rather give your engineers HogQL than write a six-figure check to Amplitude.
Skip PostHog if:
- Your analytics consumer is a non-technical PM or marketing lead who needs Mixpanel-grade dashboards out of the box.
- You're an enterprise with strict procurement, an existing analyst stack on Snowflake + Looker, and no appetite for a 9-product platform you'll only use 2 products of.
- You're past 200M events per month and not willing to either self-host or sign Enterprise.
What it actually costs to deploy PostHog correctly
The pricing page tells you the tool cost. It doesn't tell you the implementation cost, which is where teams underestimate.
A correct PostHog deployment includes: an event taxonomy and naming spec, autocapture audit (it captures everything by default and that's a privacy footgun), session replay sampling rates per platform, feature flag governance (who can ship a 100% rollout, who can't), reverse-ETL into your warehouse for finance and growth queries, and a runbook for when ClickHouse query performance degrades.
That's 20-60 hours of senior engineering time, depending on whether you're a 5-person SaaS or a 50-person scale-up. Skip those steps and you'll either re-instrument in six months or eat a privacy incident.
If you don't have someone in-house who's deployed PostHog before, you have two options: hire a fractional analytics engineer for a month, or book one of the engineers on Cadence who already has PostHog deployments under their belt. Cadence's pool of 12,800 engineers includes hundreds with shipped PostHog work, and our median time to first commit is 27 hours, so a typical PostHog-from-scratch project ships inside a week. Every engineer on Cadence is AI-native by default, vetted on Cursor, Claude Code, and Copilot fluency before they unlock bookings, which matters here because the event taxonomy work is exactly the kind of thing AI tooling accelerates from days to hours.
For most early-stage teams, that means a Mid engineer at $1,000/week ships a clean PostHog deployment, autocapture audit, and reverse-ETL into Snowflake or BigQuery. For a scale-up doing flag governance and self-host evaluation, a Senior at $1,500/week is the right tier. Either way, it's a one-week scope, not a quarter.
If you're picking your analytics stack right now and want a second opinion before you commit, run your tooling through Ship or Skip for an honest grade. If you've already chosen PostHog and need someone to actually deploy it well, the founders booking flow gets you a vetted engineer on the work in 48 hours.
FAQ
Is PostHog actually free?
Yes. The free tier covers 1M events, 5,000 session recordings, 1M feature flag requests, 100K exceptions, and 1,500 survey responses per month with no credit card required. PostHog says more than 90% of its accounts never pay. Most pre-PMF SaaS stays free for 12-18 months.
PostHog vs Mixpanel: which should I pick?
Pick PostHog if your team is engineer-led and you want one tool for analytics, feature flags, and session replay. Pick Mixpanel if your PMs and growth team need polished funnels and cohorts without learning SQL or HogQL. Mixpanel still wins on dashboard ergonomics for non-engineers; PostHog wins on consolidation and self-host optionality.
Should I self-host PostHog?
Self-host once you cross roughly 50M events per month or you have a hard EU/Schrems-II data residency requirement that PostHog Cloud EU doesn't cover. Below that volume, PostHog Cloud is cheaper than the infrastructure plus on-call burden of running ClickHouse and Postgres yourself.
How long does it take to implement PostHog?
Plan 20-60 senior engineering hours for a real deployment: event taxonomy, autocapture audit, session replay sampling, feature flag governance, and reverse-ETL into your warehouse. The 5-minute install gets you autocapture; everything else is the actual work.
Does PostHog work for mobile apps?
Yes, with iOS, Android, React Native, and Flutter SDKs. Watch the session replay rate carefully: mobile recordings are billed at roughly 2x web rates, so sample aggressively (5-10% is usually enough) and exclude high-traffic flows like login.