Mixpanel vs Amplitude vs PostHog vs Heap: 2026 Test
PostHog wins if you want pricing you can actually forecast and the option to self-host. Amplitude has the best free tier of the four for MTU-priced teams, because it bundles session replay, feature flags, and AI analytics at zero cost. Mixpanel is still the sensible default for event-volume shops that can stomach per-event overages. Heap only makes sense if you're already inside Contentsquare's orbit and don't mind a sales call to learn what it costs. That's the short version of this mixpanel vs amplitude vs posthog comparison. The rest is how I got there.
The eval that started this
I ran all four tools against the same B2B SaaS event stream. One product, one real account, my own credit card on the free tiers. The test volumes were fixed: roughly 1M events per month at the low end, which for our sample worked out to about 40K monthly tracked users, scaling toward 10M events over a projected four quarters. Same events, same funnels, same retention questions asked of each dashboard.
Here's the thing that makes this comparison genuinely annoying to run. The four tools don't price on the same unit. Mixpanel and PostHog bill on raw events. Amplitude bills on monthly tracked users. Heap bills on sessions. So "which is cheaper" has no clean answer. It depends entirely on how many events your average user fires. A charting-heavy analytics product and a read-mostly content app can have identical MTU counts and 10x different event volumes. Keep that mismatch in your head for the rest of this piece, because it's the whole ballgame.
How I scored these (the 8-axis rubric)
I graded each tool on eight axes, all from hands-on use rather than vendor benchmark decks. Vendor benchmarks are marketing. I don't score off them. The axes:
- Event/MTU cost: what you actually pay at 1M events and at scale.
- SQL & warehouse access: can you query raw data or sync to your warehouse, and at which tier.
- Session replay: is it included, gated, or an add-on.
- AI copilot quality: how useful the natural-language analytics actually are.
- Self-host option: real open-source deployment or not.
- Data governance: event approval, access controls, PII handling, plus the tier they live behind.
- Mobile SDK: iOS/Android coverage and setup friction.
- Activation setup speed: time from signup to a working funnel.
That rubric is the framework the rest of the article hangs on, and it's what the matrix below encodes. If an axis matters more to you than the others, weight it and re-read accordingly. Nothing here is scored off a vendor slide; I checked each claim against the current pricing pages, which is more work than it sounds because several of them contradict the third-party roundups.
The pricing gotcha nobody quotes back to you
Before any feature list, the trap. All four tools sell you on a generous-looking free or entry tier, and all four are priced on a meter that grows faster than your headcount. You add one onboarding flow, one analytics dashboard your customers use, one mobile screen with heavy instrumentation, and your event or session count doubles while your team stays the same size. "Basically free" becomes a real budget line within two quarters. I've watched it happen on my own test accounts.
Amplitude has the specific clause worth naming. Per Userpilot's June 2026 pricing breakdown, Amplitude's standard contracts commonly carry overage fees charged at 20 to 50% above your contracted per-MTU rate when you exceed your tier. That's not a penalty you negotiated. It's the default, and it's the kind of thing you find out about after the fiscal quarter it hits.
One more warning. Several third-party trackers publish free-tier numbers that don't match the vendors. You'll see Mixpanel's free tier cited as 20M events and Amplitude's as 50K MTUs in some roundups. Both are wrong against the current pages: Mixpanel's own pricing page states 1M free monthly events, while Amplitude's pricing page puts Starter at 10K MTUs and 2M events. When the numbers conflict, anchor to the vendor page and re-check the date on whatever roundup you're reading.
The canonical 4-way matrix
Real numbers only, all traceable to the sources named in the sections below.
| Axis | Mixpanel | Amplitude | PostHog | Heap |
|---|---|---|---|---|
| Pricing unit | Events | MTUs (or events) | Events | Sessions |
| Free tier | 1M events/mo | 10K MTUs / 2M events | 1M events/mo | 10K sessions/mo |
| Cost at ~1M events | Free | Free | Free | Free (if under 10K sessions) |
| Cost at ~10M events | ~$2,520/mo (Growth) | Plus $49/mo covers 25M events / 300K MTUs | ~$50/M, steps down at scale | Custom quote |
| Session replay | 20K/mo free (Growth) | Included free (Starter) | 5K/mo free | Not in public tier |
| SQL / warehouse | Higher tiers | Higher tiers | Yes, incl. free | Premier tier only |
| AI copilot | Yes | Yes (free tier) | Yes | Via Contentsquare |
| Self-host / OSS | No | No | Yes (MIT, ~100K events/mo cap) | No |
| Data governance | Enterprise-gated | Growth/Enterprise-gated | Available | Custom tiers |
| Contract style | Annual + overages | Annual + overages | No minimum, no contract | Sales-led custom |
Two things jump out. Amplitude's Plus plan at $49/month technically covers 25M events, which makes it wildly cheaper than Mixpanel's ~$2,520 at 10M, but only if your MTU count stays under 300K. And PostHog is the only row without "custom" or "Enterprise-gated" hiding somewhere.
Mixpanel, the event-volume default
The gotcha first: Mixpanel bills per event above the free allowance, and events are the one number you can't reliably cap without changing your instrumentation. Per Mixpanel's pricing docs, the first 1M events per month are free, and above that you pay $0.00028 per event on the entry Growth plan. That's $0.28 per 1,000, as the public pricing page confirms, with 20K free monthly session replays included. Run the math to 10M events and you're at roughly $2,520/month before any volume discount kicks in. That's not a rip-off. It's just linear, and linear meters punish spiky products. If your event volume triples during a launch week, your bill triples with it.
On the strength side, Mixpanel is fast to set up and the reporting is genuinely good. Funnels and retention are its home turf. I had a working funnel inside an hour on the free tier. Session replay being bundled at 20K/month is a real perk at this price point. Where it drops points on my rubric is governance and experimentation, which sit behind higher tiers. If you need event approval workflows or serious access controls, you're not on the plan you signed up for.
Who should skip it? Teams whose event volume is spiky and can't forecast overages. The per-event model turns every unpredictable traffic spike into an unpredictable invoice, and finance will not enjoy the conversation.
Amplitude, the MTU model
Overage clause first, because it's the surprise. As Userpilot's 2026 breakdown documents, exceed your contracted MTU tier and Amplitude bills the overflow at 20 to 50% above your per-MTU rate. On the public ladder, per UserCall's June 2026 writeup and Amplitude's own page, Starter is free with up to 10K MTUs and 2M events, Plus is $49/month on annual billing covering up to 300K MTUs or 25M events, and Growth and Enterprise are both custom-quoted. The jump from Plus to "call us" is where the pricing stops being knowable.
The free Starter tier is the strongest of the four for the money. Amplitude's pricing page states it bundles 2M events per month with session replay, A/B tests, feature flags, AI analytics, and unlimited seats. No card, no time limit. For a small team that wants replay and flags without adding two more vendors, that's a lot of surface area at zero cost. It scored highest of the four on my AI-copilot and free-tier-value axes.
Who should skip it? Teams that can't predict MTU growth, or that get itchy about custom quotes at scale. The MTU model rewards products with a stable, well-understood active-user base and quietly penalizes ones where a viral moment balloons your tracked-user count overnight.
PostHog, the open-source, transparent-pricing pick
This is the one that respects your ability to plan. PostHog's pricing page gives 1M analytics events free every month, then charges per captured event on a step-down usage model. Roughly $50 per million to start, falling steeply as volume climbs. No minimum spend, no annual contract, and you can set hard billing limits so a spike can't run away from you. The replay allowance is 5K sessions/month free. Of the four, it's the only tool where I could answer "what will this cost at 10M events" without emailing a salesperson.
The self-host story needs a caveat, though. PostHog's core is MIT-licensed and you can stand it up with a one-line Docker deployment. Tempting. But read the vendor's own words before you commit: per PostHog's GitHub repo, open-source deployments should scale to roughly 100K events per month, after which they recommend migrating to Cloud, and they explicitly provide no customer support or guarantees for self-hosted instances. That's an honest limit, stated plainly, and I respect them for putting it in writing. It also means self-host is a hobbyist or privacy-mandate move, not a cost-savings play at real volume.
Who should skip the self-host path? Anyone already past 100K events/month without infra staff to babysit it. Use Cloud instead. And if your analytics users are non-technical, know that PostHog rewards people comfortable near SQL and configuration. It's less hand-holdy than the others.
Heap by Contentsquare, the sales-led wildcard
Heap is the hardest to evaluate because most of it is hidden. Per Heap's pricing page, the only public tier is a free plan of up to 10K monthly sessions with 6 months of data history and SSO. Growth, Pro, and Premier are all custom-priced. This is the one to flag: data warehouse integration is reserved for the top Premier tier. So the feature most serious data teams care about sits behind the most expensive, least transparent door.
Heap's autocapture heritage is real, and inside the Contentsquare experience stack it has a story. But on my rubric it's docked hard for opacity. I couldn't score its per-session cost, its warehouse pricing, or its governance tier without booking a demo, and I'm not going to invent numbers I couldn't verify. Sessions as a pricing unit also make cross-tool comparison a headache, since one engaged user can generate many sessions.
Who should skip it? Anyone who wants self-serve pricing they can read on a page, or who needs warehouse access without committing to top-tier spend. It's built for buyers who are already Contentsquare customers and treat the sales cycle as normal.
Who else to consider
Worth knowing that the four above aren't the whole field. One newer entrant worth a look is Kixo, which takes a chat-first angle: you ask questions in plain language and it generates the charts and dashboards for you, with a visible reasoning trail so you can see how it got there. Useful if you don't want to teach a non-technical team the query builder. On paper it covers a lot of the same rubric ground: product analytics such as events, funnels, retention, cohorts, and user flows; session replay across web and native iOS/Android with heatmaps and privacy masking; mobile attribution and deep links; and audience segmentation with email and push tooling. Pricing is per-project on MAU-bracketed Free, Growth, and Enterprise plans.
The honest caveat: Kixo is a newer platform, and I don't have published pricing snapshots I can slot into the matrix with the same confidence as the four incumbents. So it sits outside the head-to-head scoring rather than inside it. If the natural-language approach interests you, treat it as a demo candidate alongside the others, not as a scored winner. For background on how this category is shaking out, see the AI-native product analytics explainer.
Worked example: the same 1M-event startup, four bills
Let me make the unit mismatch concrete with a tiny dataset. Take a hypothetical B2B SaaS product: 1M events/month today, which for its usage pattern maps to about 40K MTUs and stays comfortably under 10K sessions in the free window. Now project growth to 10M events over four quarters as adoption climbs.
Today (1M events / ~40K MTUs): Mixpanel is free at the 1M line. PostHog is free on the 1M allowance. Amplitude is free on Starter until it crosses 10K MTUs, and at 40K MTUs it's already over, so it's on Plus at $49/month. Heap is free only if sessions stay under 10K, otherwise custom.
Four quarters out (10M events): Mixpanel is now roughly $2,520/month per its per-event rate. PostHog is in the low hundreds per its step-down usage model, since the per-million rate drops as you scale. Amplitude's Plus plan still covers 25M events at $49/month, so on paper it's the cheapest, but only if MTUs stay under 300K. Blow past that and you're in custom-quote-plus-overage territory. Heap remains a phone call.
Here's the synthesis. The "cheapest" tool flips entirely on your events-per-user ratio. A product with few, heavy users (high events-per-MTU) crushes Mixpanel's bill and loves Amplitude's MTU cap. A product with many light users does the opposite. Cheap on events, brutal on MTUs. There is no universal winner on price. There's only the winner for your ratio, and you can't know it until you've measured your own events-per-user on a free tier for a month. Which is exactly why I ran all four.
Verdict: pick by constraint, not by brand
Choose by the constraint that actually binds you, not by whichever logo you've seen most.
When transparent, forecastable pricing is the priority, meaning you want to know your bill at 10M events without a sales call, PostHog is the pick, and it's the only one that also offers self-host. When you want free bundled AI analytics plus session replay and feature flags and your MTU count is stable, Amplitude's Starter tier is the most for-the-money. For a pure event-analytics shop at scale with predictable volume and budget to match, Mixpanel remains the sane default. And when you're already a Contentsquare customer, Heap slots into that stack. Otherwise its opacity is a real cost.
One more note on the overall pick. PostHog isn't for non-technical teams, and the open-source route stops making sense once you're past the ~100K events/month wall without infra staff. If you need SLA-backed support on a self-hosted instance, PostHog says plainly it won't give you one. So use Cloud, or pick a tool built around a support contract from day one.
Whatever you choose, do what I did. Put your own card on the free tier, run one real month of your own traffic, and read the bill before you sign anything annual.
FAQ
Which is cheapest: Mixpanel, Amplitude, or PostHog? It depends on your events-per-user ratio. At 10M events, Mixpanel runs about $2,520/month, PostHog's step-down usage model lands far lower, and Amplitude's $49 Plus plan covers up to 25M events, but only if you stay under 300K MTUs. Measure your own ratio on a free tier before deciding.
Does Amplitude's free plan really include session replay? Yes. Per Amplitude's pricing page, the free Starter plan includes 2M events per month, session replay, A/B tests, feature flags, AI analytics, and unlimited seats, with no card required.
Can I self-host PostHog for free at production scale? Not really. PostHog's own GitHub states open-source deployments scale to roughly 100K events/month before they recommend Cloud, and they offer no support or guarantees for self-hosted setups. Past that volume, Cloud is the intended path.
Why is Heap so hard to price? Because it's gone sales-led. Heap's public page only lists a free 10K-session tier. Growth, Pro, and Premier are custom-quoted, and warehouse integration is locked to the top Premier tier, so you can't evaluate real costs without a demo.