PostHog vs Mixpanel vs Amplitude: Pick One for Your Startup

Here is where I land, so you can stop reading early if you want. If you are a seed-stage product team and you want analytics, session replay and feature flags without stitching three vendors together, pick PostHog. If your team mostly needs clean event funnels and retention charts that a product manager can build without help, pick Mixpanel. If you have a dedicated analyst, a lot of stakeholders and a procurement process, pick Amplitude.

That is the whole decision for about eighty percent of the founders I talk to. The rest of this article is the reasoning, the pricing mechanics, and the traps I see people walk into after they choose.

One warning before we compare anything. The tool matters less than you think. A team with a sloppy event plan will get bad answers out of all three, and a team with a tidy plan will get good answers out of any of them. So we start with the plan.

Every analytics tool is a database of events with properties attached. Garbage in, garbage out. The single most valuable hour you will spend on analytics is writing down what you track and what you call it, before anyone touches an SDK.

A tracking plan is a spreadsheet. Four columns: event name, when it fires, properties, and who owns it. That is it. Keep the first version to 15 to 25 events. Teams that start with 150 events end up trusting none of them.

Naming rules that save you later

Pick one convention and enforce it. I like object then past-tense verb, written in snake case: project_created, invite_sent, subscription_started. Do not mix in "Clicked Button" next to "signup-complete" next to "UserUpgraded". Three weeks in, nobody will know which one is the real one, and your funnel will quietly split across duplicates.

Track outcomes, not clicks

Skip "button_clicked" with a label property. Track the thing that happened: the project got created, the invoice got paid, the teammate accepted the invite. Clicks are noise. Outcomes are what you will build funnels from.

Identify users properly

Decide on day one how anonymous visitors become known users, and call identify exactly once at login or signup with your own database user id. If you also work with accounts or workspaces, set up group analytics early. Retrofitting that after six months of data is painful in every one of these tools.

Put the plan in the repo next to the code, and make adding an event a pull request review item. That one habit does more for data quality than any feature any vendor sells you.

All three can answer the core questions: where do people drop off, which features get used, who comes back. The differences show up in what surrounds that core.

PostHog

PostHog is an all-in-one product platform. Event analytics, session replay, feature flags, A/B experiments, surveys and a data warehouse layer live in one app on one set of events. For a small team that is the point. You watch a replay of a user who dropped out of your funnel, then ship a flag that tests a fix, and you never left the tool or paid a second vendor. The interface is denser and more engineer-flavored than the other two. Product managers pick it up, but it takes a week.

Mixpanel

Mixpanel is the cleanest pure event analytics product of the three. Funnels, retention, flows and cohorts are fast to build and easy to share. If you hand a non-technical PM a Mixpanel login on Monday, they will have a useful funnel on Tuesday. It has added session replay and a warehouse connector in recent years, but analytics is where it is strongest, and I would not pick it for the extras.

Amplitude

Amplitude is the deepest on analysis. Behavioral cohorts, predictive audiences, strong governance features, a taxonomy manager and a mature experimentation product. It is built for organizations where several teams share one data set and someone owns data quality full time. If that is not you yet, you will pay for depth you do not use and spend more time configuring than learning.

My rule of thumb: PostHog is for teams that want one tool, Mixpanel is for teams that want one great chart builder, Amplitude is for teams that want one analyst to go very deep.

I am going to give you ranges and shapes, not exact numbers. All three vendors change their plans regularly, and any specific figure I print will go stale. Check each current pricing page before you decide, and run your own event volume through their calculators.

All three are usage-based and all three have a free tier. The meter is what differs.

• PostHog meters each product separately. Events, replay recordings, flag requests and surveys each have their own free allowance and their own per-unit price that drops as you grow.
• Mixpanel bills mainly on events, with add-ons for replay and other extras. Recent plans have leaned on monthly tracked events as the primary meter.
• Amplitude has historically leaned on monthly tracked users or event volume, with paid tiers that open up the heavier analysis and governance features and an enterprise tier you reach through sales.

In practical terms, here is what I see. A pre-launch or early product with a few thousand monthly users will usually pay nothing on any of the three. Somewhere between 10,000 and 100,000 monthly active users, bills commonly land in the low hundreds to low thousands of dollars a month, depending on how chatty your events are and which add-ons you switch on. Past that, you are in negotiation territory with all three, and the enterprise plans can run well into five figures a month.

What actually drives your bill

Event volume is the lever, and you control it. Auto-capture is convenient and expensive. Firing an event on every scroll or every keystroke is how a free tier disappears in a week. Track the 20 events in your plan, sample the noisy ones, and the bill stays small on any vendor.

The other surprise is the add-on stack. If you buy analytics from one vendor, replay from another and flags from a third, you are paying three minimums and three overage meters. That is the quiet financial argument for PostHog at the seed stage. Each product is not always cheaper in isolation, but you pay one vendor and only for what you use of each part.

Replay and flags are where the three tools diverge most, and where the all-in-one argument gets concrete.

Session replay

PostHog includes replay as a first-class product, tied directly to your events. You can click a funnel step, jump to the people who dropped there, and watch them. That loop is the fastest way I know to find out why a number is bad. Mixpanel and Amplitude both offer replay now, and both link it to their analytics, but it arrived later and feels more like an add-on. If replay is a daily tool for you, you can also buy it from a specialist like Hotjar or FullStory, but that is another bill and another place to keep privacy settings aligned.

Whichever tool you use, mask inputs by default and turn on masking for anything resembling a password, card number or health detail. Replay can capture more than you intended. Decide what is off limits before you start recording.

Feature flags and experiments

PostHog ships flags and experiments in the same product, using the same events, so you can roll out a change to 10 percent of users and read the result without exporting anything. Amplitude has a capable experimentation product that is strongest when you already live in Amplitude. Mixpanel has been the lightest here, and many Mixpanel teams pair it with a dedicated flag tool.

If feature flags are a serious part of your release process, read our comparison of LaunchDarkly vs Statsig vs Unleash before you assume your analytics vendor is the right home for them. A dedicated flag tool earns its place once you have several teams, audit needs or complicated targeting. Before that, the flags inside your analytics tool are plenty.

And if your goal is running real tests rather than simple rollouts, our walkthrough of building an AI A/B testing and experimentation engine covers the statistics and the plumbing that no vendor dashboard fixes for you.

If you have European users, or you sell to companies that run security questionnaires, this section can decide the whole thing.

PostHog is open source and can be self-hosted, and it also runs a cloud with an EU region. Self-hosting means your event data never leaves your infrastructure, which makes some data processing conversations much shorter. It also means you operate a data-heavy system, with a database, queues and upgrades to look after. For a team under ten engineers, I usually recommend the managed cloud with the EU region rather than self-hosting. Run it yourself only when a customer contract or a regulator forces your hand.

Mixpanel and Amplitude are managed cloud products. Both offer EU data residency options, and both publish data processing agreements and security documentation. Check the current plan details, because residency options and which tier includes them have changed over time.

What to do whatever tool you pick

• Do not send personal data you do not need. An internal user id is better than an email address in event properties.
• Gate the SDK behind a cookie consent choice where the law requires it, and test that nothing fires before consent.
• Mask replay inputs and exclude sensitive pages from recording.
• Write a deletion process. When a user asks to be erased, you need to be able to remove them from your analytics tool, and you should test that path once before you need it.

I am not a lawyer, and your obligations depend on your market and your data. Treat this as a checklist of questions to take to counsel, not legal advice.

Most startups I work with ship a mobile app, often on React Native with Expo, so the mobile story matters more than the vendor docs admit.

All three have React Native SDKs. The practical differences show up in three places.

Expo compatibility

Check whether the SDK works in Expo managed workflow or needs a development build with native modules. Analytics alone usually runs fine. Session replay on mobile is a different matter, because it hooks into native views, and that typically means a custom development build instead of Expo Go. Test it on a real device early. I have seen teams pick a tool in a planning meeting and discover the replay constraint in their first week of integration.

Event batching and offline behavior

Mobile users go through tunnels and airplane mode. Confirm the SDK queues events offline and flushes them later, and check how it handles timestamps so a delayed event lands at the right time. Also check how much it adds to your app size and startup time. Mobile replay in particular can add weight.

One identity across web and app

If you have a marketing site, a web app and a mobile app, use the same user id and the same event names on all three. Your funnel from landing page to first purchase only works if the pieces join up. This is another reason the tracking plan comes first. It is the document that keeps three codebases honest.

Plan for a small wrapper in your own code, one function that every screen calls to send an event. Your tracking plan lives behind that function, and swapping vendors later becomes a one-file change instead of a hundred-file change.

You will probably change tools at least once. Companies grow, needs shift and pricing changes. The cost of switching depends mostly on choices you make now.

First, keep the wrapper I described above. If your app code calls track("project_created", props) through your own function, the vendor underneath is a detail.

Second, get your raw events somewhere you own. All three can export event data to a warehouse such as BigQuery, Snowflake or Postgres, either through a built-in connector or through a pipeline tool like Segment or RudderStack. Warehouse export on the lower tiers is sometimes limited or priced as an add-on, so check this specifically on each pricing page. A copy of your raw events in your own warehouse is the best insurance against vendor lock-in, and it lets you join product events with billing and CRM data, which no analytics dashboard does well.

Third, accept that history does not migrate cleanly. Importing old events into a new tool is possible but messy, and charts built on imported data rarely match the originals. Most teams run both tools in parallel for 30 to 60 days, compare key numbers, then cut over and keep the old tool read-only for a while.

If you are building product-led growth infrastructure around this, with usage tracking, billing events and a warehouse, our breakdown of what product-led growth infrastructure costs shows where analytics sits in the larger budget.

Here is the rule I give founders. Find your row and stop agonizing.

Pre-launch to roughly 1,000 users

Pick PostHog on the free tier. You get analytics, replay and flags for free, and at this stage watching ten real people struggle with your onboarding teaches you more than any chart. Write the tracking plan, keep it small, and move on.

Seed stage with a product team of two to ten

Stay on PostHog if you want one tool and your engineers are involved in analytics. Choose Mixpanel if a product manager or growth lead owns analytics and wants to build charts alone, and you already have a flag tool or do not need one. You will pay low hundreds a month at most for a while on either.

Series A with a growth team

This is where Amplitude starts to earn its price, but only if you have a person whose job is analysis and several teams asking questions from the same data. If that person does not exist yet, hire them before you buy the tool. Otherwise PostHog or Mixpanel will carry you further than you expect, with a warehouse copy of your events as your safety net.

Enterprise buyers and regulated data

If your customers demand self-hosting, strict residency or audit trails, PostHog self-hosted or an Amplitude enterprise plan are the realistic paths. Let your security questionnaires make the call.

Notice what is missing from that list. I did not say "pick the tool with the most features". Nobody I know has failed because their analytics tool lacked a feature. Plenty have failed to act on what the data told them.

If you want a second opinion on your tracking plan, your stack or your growth instrumentation before you commit budget, we do this work every week. Book a free strategy call