Mixpanel vs Amplitude vs PostHog: Data Models, Real Pricing, and Who Each Is For

Three product-analytics tools dominate the shortlist whenever a team decides it is finally time to measure what users actually do: Mixpanel, Amplitude, and PostHog. They sound interchangeable in a demo. They are not. The differences live in the data model, the billing meter, and the question of who on your team is expected to answer the next analytics question without filing a ticket.

I have wired event tracking into all three, and the wrong choice is expensive in a way that does not show up until month four. You instrument the SDK, you ship, traffic grows, and then either the bill jumps by an order of magnitude or your product managers quietly stop opening the dashboard because every question requires a data engineer. This comparison is built around those two failure modes. Pricing figures below are the vendors' own published rates as of June 2026; analytics pricing changes often, so confirm against each pricing page before you commit a budget.

How Each Tool Models Your Data

Every product-analytics tool stores the same primitive: an event, attached to a user, with a timestamp and a bag of properties. The differences are in how that primitive is stored and queried, and that single architectural choice cascades into pricing, speed, and who can use the product.

The canonical event payload looks roughly the same across all three:

track("checkout_completed", {
  distinct_id: "u_8842",
  timestamp:   "2026-06-07T14:22:09Z",
  properties: {
    plan:        "growth",
    cart_value:  148.00,
    items:       3,
    ab_variant:  "checkout_v2"
  }
})

What diverges is the engine underneath. PostHog runs on ClickHouse, an open-source columnar database, which is why it can offer raw SQL access and a self-hosted build. Mixpanel uses a proprietary store purpose-built for behavioral queries, tuned so that funnels and retention return fast without you writing SQL. Amplitude sits closest to your warehouse, offering warehouse-native analysis that can query Snowflake, BigQuery, or Databricks more or less in place.

Events Versus Users as the Unit of Measure

The billing unit is the most consequential decision you will make, and the three tools split cleanly. Mixpanel and PostHog meter on events: every tracked action counts. Amplitude meters primarily on monthly tracked users (MTUs), with an events ceiling layered on top. This is not an accounting footnote. It changes which products get cheap and which get punished as you scale.

A rough way to predict your bill before you sign anything:

events_per_month ≈ MAU × actions_per_user_per_day × 30

# A consumer app: 50,000 MAU, ~8 tracked actions/day
≈ 50,000 × 8 × 30 = 12,000,000 events/month

An event-metered tool charges that twelve-million-event figure directly. An MTU-metered tool charges on the 50,000 users and largely ignores how chatty each one is. If your app fires many events per session — a media player, a trading screen, an editor — MTU pricing protects you. If you have a huge user base that each does very little, event pricing can be the cheaper meter. Model both before you decide.

The meter also changes how you instrument. Under event pricing, every extra track() call has a marginal cost, so teams get disciplined about what they log and sometimes under-instrument to save money, which quietly starves the analysis later. Under MTU pricing the marginal event is free once a user is counted, so you instrument generously and worry about the user count instead. Neither incentive is wrong, but they pull instrumentation habits in opposite directions, and that habit is hard to reverse once a codebase has hundreds of tracking calls in it.

Real Pricing as of June 2026

Marketing pages love the word "free," and all three back it up at the bottom of the range. The question is what happens at the first real scale step, which is where teams get surprised. Here is the published structure for each.

Tool	Billing unit	Free tier	First paid tier	Self-host
PostHog	Per event (per product)	1M events + 5K session recordings / mo	Usage-based; ~$0.00005/event after 1M	Yes (MIT, open source)
Mixpanel	Per event	1M events / mo, unlimited seats	Growth: ~$0.28 per 1K events above 1M	No
Amplitude	Per MTU (events ceiling)	10K MTU / 2M events, unlimited seats	Plus: $49/mo, up to 300K MTU or 25M events	No

Two things in that table deserve emphasis. First, every one of these tools now ships a genuinely usable free tier with a 1M-event or 10K-MTU floor and unlimited seats, which was not true a few years ago. You can run a real early-stage product on any of them at zero cost. Second, the curves diverge hard above the free line, and they diverge in different directions because the meters differ.

Where the Bill Actually Jumps

Take the 12M-event consumer app from the formula above and read it through each meter. On Mixpanel's Growth plan, the first 1M events are included and the rest bill at roughly $0.28 per thousand, which puts a twelve-million-event month in the low four figures. PostHog's per-event rate starts near $0.00005 and steps down steeply with volume, with published discounts reaching the 80 percent range at the hundreds-of-millions tier, so high-volume event streams get progressively cheaper per unit. Amplitude, metering on the 50,000 users rather than the 12M events, lands in a different place entirely and is often cheaper for chatty apps — until your user count, not your event count, is what grows.

The trap is benchmarking on today's volume. Instrument with growth in mind: pick three points on your roadmap — current, 6 months out, 18 months out — and price all three tools at each. The cheapest tool today is frequently not the cheapest tool at the scale you are actually building toward.

Two pricing details that hide in the footnotes are worth dragging into the light. Amplitude's $49 Plus rate is billed annually, so the monthly-equivalent commitment is the figure that lands on your card, and the headline price assumes the annual term. And on every one of these tools, the meter is not just the analytics events; session replay, feature-flag requests, and data-warehouse rows each carry their own line. PostHog is explicit that each product has its own metric, which means a single "analytics bill" can quietly become four meters running at once if you switch the other products on. Read the per-product pricing, not just the per-event headline.

Who Can Answer the Next Question

Pricing gets the spreadsheet attention. Accessibility decides whether the tool gets used at all. The relevant question is not "can this answer my question" — all three can answer almost anything. It is "who has to be in the room when a new question comes up."

Mixpanel is built so a product manager with no SQL can build a funnel, segment it, and break it by a property in a few clicks. That is the entire design intent, and it shows. There is no SQL editor in the core analysis flow because the premise is that you should not need one.

PostHog gives you the visual builder and a full SQL editor over the raw ClickHouse tables. When the UI cannot express your question, you drop into SQL and write it yourself. That is power, and it is also a fork in the road: teams without a SQL-fluent member tend to live in the visual layer and never touch the depth they are paying to have available.

Amplitude sits in between, with strong guided analysis plus SQL access for power users and warehouse-native querying for teams whose source of truth already lives in Snowflake or BigQuery. If your analytics center of gravity is the data warehouse, Amplitude's model fits the org chart you already have.

There is a second-order effect here that rarely makes the comparison tables. The accessibility tier you choose decides where your analytics knowledge accumulates. A no-SQL tool keeps that knowledge inside the product team, where the people closest to the feature ask and answer their own questions. A SQL-and-warehouse tool concentrates it in a data team, which produces more rigorous analysis but adds a queue between a question and its answer. Picking the wrong side for your org structure does not break anything; it just means the dashboard slowly goes stale because the people who have questions are not the people who can answer them.

The Self-Host Question

Only PostHog offers a self-hosted, open-source deployment, and this is the cleanest dividing line in the whole comparison. If you are in a regulated industry, cannot send user-level event data to a third party, or simply want to own the stack, PostHog is effectively the only option of the three.

The catch is that "free and open source" is not free to operate. Self-hosting PostHog means running ClickHouse, Kafka, Postgres, and Redis, and keeping them healthy under load. That is a standing infrastructure commitment, not a one-time install. The honest framing: self-host PostHog when data residency is a hard requirement, not to save money on the SaaS bill. At most volumes the managed cloud is cheaper than an engineer's time spent babysitting ClickHouse.

Beyond the Big Three: Where BI and CDPs Fit

A common mistake is treating product analytics as the whole data stack. It is one layer. Two adjacent categories regularly get confused with it, and choosing the wrong category is a more expensive error than choosing the wrong tool within a category.

Business intelligence tools answer different questions than product analytics. Metabase (open-source, SQL-first dashboards), Tableau (enterprise visualization), Hex (notebooks plus BI), and ThoughtSpot (search-driven BI) sit on top of your warehouse and answer "what does the business look like." Product-analytics tools answer "what did this user do, in sequence." You will often run one of each: a product-analytics tool for behavioral funnels and a BI tool for revenue and operational dashboards.

Customer data platforms are the plumbing, not the destination. Segment collects events once and fans them out to every downstream tool, so you instrument a single SDK and route the same stream to your analytics, your warehouse, and your marketing tools. Segment's Team plan starts at $120/mo for 10,000 MTUs. A CDP does not analyze anything; it pipes. The reason it appears in analytics comparisons is that it decides how clean your data is before any analytics tool ever sees it.

Category	Question it answers	Tools	Lives where
Product analytics	What did this user do, in sequence?	Mixpanel, Amplitude, PostHog	Behavioral event store
Business intelligence	What does the business look like?	Metabase, Tableau, Hex, ThoughtSpot	On top of the warehouse
Customer data platform	How do I collect once, route everywhere?	Segment	Between SDK and everything

The Honest Trade-Offs of Each

No tool here is bad. Each is sharp in one direction and dull in another, and the failure mode is buying a tool for a strength you will not use while ignoring the weakness that will hit you.

Where Each Tool Quietly Costs You

• Mixpanel: fastest path to a funnel for a non-technical PM, and a free tier that genuinely covers an early product. The cost is depth: no in-product SQL, and event-based billing that climbs as your tracking gets richer. Group Analytics for B2B is a paid add-on, not table stakes.
• Amplitude: deepest behavioral analysis and experimentation, with warehouse-native querying that fits warehouse-first orgs. The cost is that MTU pricing plus enterprise governance pushes teams off the free tier faster than expected, and the depth carries a learning curve a one-person team may never amortize.
• PostHog: the all-in-one, bundling analytics, session replay, feature flags, experiments, surveys, and a data warehouse under one usage-based bill, plus the only self-host option. The cost is that breadth is also surface area. There is more to learn, and self-hosting is a real ops commitment, not a checkbox.

One cross-cutting signal worth noting beyond features and price: independent review panels rate all three highly but not identically. On our own AI review panel, PostHog scores 8.4, Amplitude 7.8, and Mixpanel 7.7, a spread that tracks the breadth-versus-focus trade-off rather than any quality gap. Treat that as one input among several, not a tiebreaker. The right tool is the one whose meter and accessibility match your team, not the one with the highest aggregate score.

Which One Should You Actually Pick

Skip the feature-matrix paralysis. The decision collapses to three questions about your team and your data, answered in order. Stop at the first one that gives a hard answer.

A Three-Question Decision Framework

1. Must user-level data stay in your infrastructure? If yes, whether from a regulated industry, a data-residency rule, or a hard ownership requirement, choose PostHog self-hosted and budget for the ops. This requirement overrides everything below.
2. Who answers the next analytics question? If it is a non-technical PM working alone, choose Mixpanel for its no-SQL speed. If it is a data team that already lives in Snowflake or BigQuery and wants deep experimentation, choose Amplitude for warehouse-native depth.
3. Do you want one tool or a best-of-breed stack? If you would rather consolidate analytics, session replay, and feature flags under one usage-based bill, PostHog Cloud is the consolidation play. If you prefer specialized tools wired together through a CDP, run Mixpanel or Amplitude for analytics and let Segment handle the piping.

If none of those resolves cleanly, default to the cheapest free tier that fits your stack and instrument behind a CDP so the events are portable. The single most expensive analytics mistake is not picking the wrong tool. It is hard-wiring your tracking to one vendor's SDK, then discovering at scale that switching means re-instrumenting the entire app. Route through Segment or PostHog's own ingestion layer from day one, keep the event schema vendor-neutral, and the choice you make this quarter stops being a choice you are stuck with next year.