Traffic analytics tells you who visited your site. Product analytics tells you what they did after they signed up, and that second half decides whether your SaaS lives. Product analytics platforms are event-based tools that track individual user actions (signups, key actions, upgrades) inside your product, then stitch those events into funnels, retention curves, and cohort reports. Web analytics counts pageviews; product analytics counts people reaching their first win.
Where I’d start, by job: for a seed-stage startup that wants analytics, session replay, and feature flags under one roof, pick PostHog. For polished funnel and cohort analysis with transparent pricing, Mixpanel; for the deepest behavioral analysis as you scale, Amplitude. For retroactive “I wish we’d tracked that” questions, Heap. When you need in-app guides and analytics together, Pendo. If experimentation is the main event, Statsig. For preset B2B SaaS reports without learning a query builder, June. And Google Analytics 4? Run it for acquisition, but don’t pretend it answers product questions.
The concepts that decide your insights and your bill
Before you compare vendors, get these straight, they determine both what you learn and what you pay:
- Activation event, the single action that says a user “got it”: sent their first campaign, created their first project, invited a teammate. Define one, make it falsifiable (“did X within Y days of signup”), and build your first funnel around it. Most complaints that “our analytics are useless” trace back to never defining activation.
- MTU/MAU billing, most vendors bill by monthly tracked users (MTUs): every unique person who fires one event in a month, whether or not they were active. Dormant accounts, lurkers, and one-time triers all count. This is the number that blows up your bill, not raw events.
- Retention curve, the share of a signup cohort still active after N days. I judge retention by the curve’s shape: a curve that flattens means a stable base of returning users; a curve that decays to zero is a leaky bucket no acquisition channel can outrun.
- Auto-capture vs explicit tracking, Heap-style auto-capture records everything by default so questions can be asked retroactively; explicit tracking (Amplitude, Mixpanel) records only the named events you define. Auto-capture lowers effort and raises noise; explicit tracking is the opposite.
Quick job-to-tool mapping I use with founders:
- Everything under one roof, one bill → PostHog.
- First serious contract, self-serve pricing → Mixpanel.
- Data science ambitions at volume → Amplitude.
- Retroactive questions, no tracking plan → Heap.
- Onboarding guides and nudges driven by the same data → Pendo.
- A/B testing is the main event → Statsig.
- Monday-morning reports with near-zero setup → June.
The shortlist: 8 product analytics platforms
Amplitude, best for deep behavioral analysis as you scale
Amplitude is the heavyweight of the category: funnels, cohorts, and behavioral segmentation with the most polished charting I’ve used. It earns its keep once you have volume and a team asking nuanced questions, “which actions in week one separate retained users from churned ones” is an Amplitude question. It has added session replay and experiments, though analytics remains the core.
Pricing: a free Starter plan covering roughly 50,000 monthly tracked users; Growth and Enterprise tiers are quote-based, and Amplitude is firm about not publishing numbers.
Quote-based pricing means you’re budgeting blind, and Amplitude is the bill most likely to surprise you at scale, MTU growth plus seat creep compounds fast.
Mixpanel, best all-rounder for self-serve startups
Mixpanel has been the pragmatic choice for a decade: fast funnels, retention, and cohort reports with a UI your whole team learns in a week. Its free tier is the most generous of the big three, and unlike Amplitude it publishes its pricing. For a self-serve SaaS with one product, it’s the safest first contract.
Pricing: a free plan with about 20 million events and 1 million MTUs per month; the Growth plan is usage-based at roughly $0.28 per MTU with volume discounts, their calculator gives you the exact number before you talk to anyone.
MTU billing counts every unique person who fires a single event, including dormant lurkers, so a viral signup wave can triple your bill without adding one active user.
PostHog, best all-in-one platform for a lean startup stack
PostHog is the “bundle everything” option: product analytics, session replay, feature flags, A/B testing, surveys, and a data warehouse in one platform. For a two-to-ten-person startup, replacing three vendors with one integration and one bill is worth real money and real engineering time.
Pricing: generous free allowances per product, about 1 million analytics events and 5,000 session recordings monthly; beyond that, usage-based at roughly $0.0002 per analytics event, with enterprise features quote-based. Their pricing page includes a live calculator.
PostHog covers several jobs from one dashboard, and auto-capture across eight products creates event soup if nobody owns a tracking plan.
Heap, best for retroactive “I wish we’d tracked that” analysis
Heap’s differentiator is auto-capture: it records every click, tap, and form submission by default, so you can define events and build reports after the fact. When you discover in week twelve that you needed “used the CSV import” as an event in week one, Heap is the only tool where history is still intact.
Pricing: a free plan up to about 10,000 sessions/month; paid tiers have been listed from around $360/month billed annually, and Heap is only partially transparent above that, confirm before committing.
Pricing turns enterprise-ish quickly beyond the free tier, and capturing everything means heavy datasets and an event catalog that needs active curation to stay usable.
Pendo, best when you also need in-app guides
Pendo pairs product analytics with in-app guides, onboarding checklists, and a feedback module, aimed at product teams at mid-market SaaS companies. If your roadmap includes tooltips, announcements, and adoption campaigns driven from the same data as your analytics, Pendo’s bundle is coherent.
Pricing: Pendo Core is free up to about 5,000 MTUs; paid tiers are quote-based.
It’s heavy and priced for mid-market, and its analytics are a tier below Amplitude or Mixpanel, you’re buying the guides as much as the charts.
Statsig, best for experimentation-first teams
Statsig treats feature flags and A/B testing as the core product, with product analytics and session replay attached. If your team ships behind flags and measures everything with controlled experiments, its experiment engine and statistics layer are the strongest in this list.
Pricing: a free tier sized generously for early-stage products; beyond it, usage-based pricing with per-event rates on the pricing page and the top tier quote-based. Confirm the current free cap before you standardize on it.
Analytics is the supporting act, not the headline, dashboarding and reporting depth trail Amplitude and Mixpanel.
June, best preset reports for B2B SaaS
June is opinionated on purpose: instead of a blank canvas, it generates the standard B2B SaaS reports (activation, retention, engagement, feature usage) from a handful of tracked events. For a founder who wants Monday-morning answers without learning a query builder, it’s the fastest path to a working dashboard.
Pricing: a free plan for small products; paid plans from roughly $60–70/month for the first 10,000 tracked users, scaling with volume. June’s tiers have changed before, confirm current rates.
The presets that make it fast also cap it. The moment you ask a nonstandard question, you feel the flexibility gap.
Google Analytics 4, best free acquisition layer, not a product tool
GA4 belongs in this list because founders keep asking whether it can replace the rest. It’s excellent (and free) for acquisition, channels, and landing-page behavior, and it ties into Google Ads. But it measures sessions and pages, and its data model makes user-level funnels and cohort retention by plan or role painful.
Pricing: free; the GA360 enterprise contract runs about $50,000/year, negotiated through Google sales.
Use it to measure acquisition. For product questions, use a tool with deeper event and funnel analysis.
Product analytics comparison
| Tool | Best for | Starting price | Standout capability | Main limitation |
|---|---|---|---|---|
| Amplitude | Deep behavioral analysis at scale | Free (about 50k MTUs); Growth quote-based | Cleanest cohort and behavioral reporting | Quote-based pricing past the free tier |
| Mixpanel | Self-serve startups, all-round | Free (about 20M events); Growth about $0.28/MTU | Fast, flexible funnels and retention | MTU billing punishes lurkers |
| PostHog | All-in-one analytics + replay + flags | Free (about 1M events/mo); then about $0.0002/event | Analytics, replay, and flags in one bundle | Breadth over depth; event soup risk |
| Heap | Retroactive auto-capture analysis | Free (about 10k sessions); from about $360/mo | Define events after the fact | Opaque, enterprise-leaning pricing |
| Pendo | Analytics plus in-app guides | Free Core (about 5k MTUs); paid quote-based | Guides and analytics in one platform | Heavy; analytics not its strongest side |
| Statsig | Experimentation-first teams | Free tier; usage-based beyond | Feature flags and A/B tests as first-class objects | Reporting layer thinner than the big two |
| June | Preset B2B SaaS reports | Free (small products); from about $60–70/mo | Standard reports you don’t have to build | Presets limit flexibility |
| GA4 | Free acquisition/traffic layer | Free; GA360 about $50k/yr | Free, universal, ties into Google Ads | Not built for product questions |
The activation analytics setup I run
- Define activation in one sentence. With the whole team: “a user is activated when they [do X] within [Y days] of signup.” For my last product it was “sent a first campaign within 7 days.” If the team can’t agree on the sentence, that’s the real problem, not the tooling.
- Instrument the minimum viable taxonomy. Six to ten events: sign_up, the activation event, the two or three core actions around it, and upgrade. One naming convention, one document, one owner. Every event beyond that needs a written justification.
- Build one funnel and check it weekly. Visit → signup → activation → first core action → paid. Five steps, reviewed weekly. Most activation problems are one bad step, not a bad product.
- Read retention by cohort shape. Weekly signup cohorts, eight-plus weeks out. I don’t care about the absolute number early; I care whether the curve flattens (stable returning users) or decays toward zero (leaky bucket).
- Segment activation by source, plan, and platform. Two channels with similar traffic can have wildly different activation rates. Kill or fix the channels that never activate, that’s marketing-attribution territory, and analytics is what makes it measurable.
- Wire the recovery touch. When someone signs up but doesn’t activate within Y days, trigger a nudge (email or in-app) and measure whether it moves the number. Your analytics stack tells you where the leak is; SaaS onboarding software is how you fix it.
Product analytics FAQ
How much does a product analytics platform cost?
Free is real at the start: PostHog’s roughly 1 million monthly events, Mixpanel’s 20 million events, and Amplitude’s approximately 50,000 MTUs cover most seed-stage products. Past free, expect MTU-based pricing, Mixpanel’s Growth plan at about $0.28 per MTU is the most public number in the category, while Amplitude, Pendo, and Heap go quote-based. Budget rule of thumb: at 10,000 MTUs expect roughly $100–300/month; at 100,000 MTUs, four figures.
Amplitude vs Mixpanel, which should a startup pick?
Either can do the job. My tiebreakers: Mixpanel’s free tier and transparent MTU pricing make it the lower-risk first contract, while Amplitude’s behavioral tooling is stronger once you have volume and data-science ambitions. Teams rarely regret either choice, they regret picking from a demo instead of pricing out their projected MTU count.
Do I need product analytics if I already run GA4?
Yes, if your product is the app behind the login. GA4 measures sessions and pages; product analytics measures people and actions. GA4’s data model, event limits, and page-centric reports make funnel-by-role analysis and cohort retention painful. Run both: GA4 for acquisition, a dedicated tool for the product.
What’s a good activation rate?
Benchmarks vary wildly by product, so I’ll give you practitioner thresholds instead of a fake industry number: for self-serve SaaS I treat 20–25% of signups hitting the activation event as a floor and investigate anything below it. More important than the level is the trend, if activation is flat or rising while traffic grows, your funnel is healthy; if new channels push volume while activation falls, you’re buying the wrong users.
How many events should I track?
Six to ten to start. sign_up, your activation event, the two or three core actions around it, and upgrade will answer 80% of your questions. Every additional event needs a written justification, because event taxonomies rot: the teams I see with 200 events usually can’t answer basic questions, and the teams with eight usually can.
Sources
- Amplitude pricing (https://amplitude.com/pricing)
- Mixpanel pricing (https://mixpanel.com/pricing/)
- PostHog pricing (https://posthog.com/pricing)
- Heap pricing (https://heap.io/pricing)
- Pendo pricing (https://www.pendo.io/pricing/)
- Statsig pricing (https://statsig.com/pricing)
- June pricing (https://june.so/pricing)
- Google Analytics (https://marketingplatform.google.com/about/analytics/)
- G2 reviews of Amplitude (https://www.g2.com/products/amplitude/reviews)
- G2 reviews of Mixpanel (https://www.g2.com/products/mixpanel/reviews)
