Mixpanel vs Amplitude vs PostHog vs Heap: What Product Analytics Actually Costs

Who each one is for

Mixpanel is the only one of the four that shows you a price before you talk to anyone. Its Growth plan is billed on event volume with a published rate card: $100/month at 1.5M events/month, $438/month at 5M, $1,635/month at 20M, all on yearly billing, and $0/month on a free plan that includes unlimited seats. Seats are not the meter, so a 40-person team and a 12-person team pay the same at the same event volume.

Amplitude publishes the entry and the shape of the bill, then stops. Every plan includes unlimited seats on every plan, the free tier carries 2M events/month forever, and Plus starts at $0 with published headroom to 70M events. Above that, Growth is custom, event based, which is stated honestly on the page rather than hidden.

PostHog meters every product separately and bills usage: 1M events/month of analytics free every month, then pay-as-you-go. Its own calculator opens at $0 across seven products, and the two prices it does publish outright are oddments: $0.05 per GB per month for log retention beyond the default window, $80 for 30 minutes for an onboarding call. That is a developer-shaped pricing page, not a procurement one.

Heap has a plan ladder and no price on it. The page lists Free, Growth, Pro and Premier, then puts custom session pricing and Contact Us where the number should be. If you need to budget Heap before a sales call, this page cannot help you and neither can Heap's own site.

Last verified 2026-10-21. Plans move and prices follow usage bands; confirm current rates before you quote this page.

Four tools, four meters

Product analytics looks like one category on a shortlist and is not one category in a budget line. Each vendor meters a different thing, so the same team at the same scale gets a different bill from each.

The shortlist at a glance

Each icon is the vendor's own declared site symbol, fetched once on 21 October 2026 from the same pages the prices come from. The line under each name is what that vendor puts on your invoice.

Mixpanel

Events tracked per month, and the only published rate card

Amplitude

Events per month: entry published, the Growth band quoted

PostHog

Usage per product: analytics, replay, flags, logs each metered

Heap

Sessions and events, priced per deal, nothing published

Seats do not appear in any of the four meters. Headcount is not what you are buying here.

Logos 4 of 4 tiles: each icon is the vendor's own declared site symbol, fetched once from the page the prices come from, normalised to PNG and hashed in the arm's logo-manifest.json. Trademarks belong to the vendors and are used here to identify them. No third-party logo source, and no logo where the vendor was not collected.

What each vendor lets you know without talking to sales

Captured 21 October 2026 from the vendor's own pricing page, one fetch each, robots.txt checked against the exact path first. "Published" means the figure is in the document the server sent, not behind a click, a form, or a calculator you have to drive.

Mixpanel's price per million events, band by band

Yearly billing, as published on the vendor's page on 21 October 2026. The right-hand column is ours, computed from the published monthly figure divided by the event volume in that band. It is the arithmetic a buyer has to do anyway, and it is where the trap sits.

What a million events costs at each band

Our arithmetic: the vendor's published monthly price for the band, divided by the band's event volume. The point is the shape, not any one bar.

Every input figure is quoted in the table above this chart. The line that goes the wrong way is the one that matters: doubling volume from 1.5M to 2M events raises the cost per million.

Four things the numbers say

Each vendor, properly covered

When to choose one over another

Which tool for which situation

When to buy none of them

If you cannot yet say what you will do with the answer, none of these is the purchase. Product analytics pays for itself when a specific decision changes: which onboarding step to fix, which cohort to revive, which feature to cut. Name the decision, and the tool choice gets much easier. Name nothing, and you will buy a dashboard nobody opens, which is the most common failure in this category.

If your event volume is genuinely tiny, a warehouse plus SQL answers the same questions at a fraction of the meter. The tools here earn their keep on behavioural depth, cohort work and replay, not on counting.

And if the real problem is that your analytics tool cannot see the funnel because the tracking code was never written, no price list fixes that. Instrument first, then price. This is the same gap our extractability work measures from the other direction: what a machine can actually read out of your pages. The free extractability audit checks it for AI crawlers; the instrumentation gap is the identical problem for analytics vendors.

The implementation traps

Questions people actually ask

Four readers, four things to take

Founders at seed and Series A. Two of these four vendors give you a free year or a free tier with real limits on it, and the thresholds are published: under $8M raised for Mixpanel, under $10M and under 20 employees for Amplitude. Apply before you sign anything, because the scholarship is a plan, not a discount you negotiate later.

Buyers at three to fifty million dollars of annual revenue. Put the volume question first: how many events per month will you actually send once mobile, backend and product events are all in? That number decides which band you are in, and only one vendor on this page lets you look up its price.

Practitioners running the evaluation. Run the PostHog calculator at your own volume and screenshot the result. It is a real quote in every sense except that it is not on the page, and you should log the date you ran it.

Researchers and journalists. The census table is the citable object: one of four vendors publishes a rate card, none prices by seat, and two of the four put their number behind a form. Every figure traces to a captured document with a SHA-256 printed below, so the claim can be checked rather than argued.

What this does not show

Where this census sits in the rest of our work

Four studies that bear on the same question: what a vendor's own page lets you find out before you talk to them.

How we checked

Four vendor pricing pages were captured with a single compliant HTTP fetch each on 21 October 2026. Before any fetch, each host's robots.txt was read and the exact target path tested against the User-agent: * block; every path was allowed. The full check, including the two hosts whose robots files ban automated access in other arms of this research, is recorded in the collection protocol at docs/plans/vendor-comparisons-protocol.md in the DataDab repository.

A figure is published on this page only if it appears in the server-rendered text of the captured document. Where a vendor's prices exist but only inside a client-side calculator, we say so instead of quoting the number. The hashes below let anyone re-verify that a claim on this page matches the document we fetched, and we correct the page if a vendor's page changes.

No vendor paid for placement on this page and no vendor was given advance sight of it. The census deliberately includes a "no" because omitting opaque vendors would remove the finding.