Measure the lifecycle, then choose the tool
The order most teams use is backwards. They pick PostHog or Mixpanel first, instrument whatever is easy, then try to answer growth questions from the events they happened to collect. The result is a tool full of data that cannot answer the four questions that matter.
A B2B SaaS user-behavior analytics stack exists to answer four account-level questions, in this sequence:
- Did this account activate? Reach its first real value moment, measured per account or workspace, not per login.
- Is it adopting? Using the features that correlate with staying, at the depth that predicts renewal.
- Is it expanding? Adding seats, spreading to new teams, hitting limits. The shape of a product-qualified lead.
- Is it about to churn? The champion goes quiet, weekly use decays to monthly, feature breadth narrows.
Notice that all four are about the account. In consumer products the user is the unit. In B2B the buyer is the account and the seat is just one signal inside it. A measurement plan built on individual-user metrics will mislead a B2B team at every stage. That distinction drives everything else on this page, and it is the first thing we check when we run an analytics and insights engagement.
What to measure at each lifecycle stage
Takeaway: each of the four stages has one primary metric and a set of leading indicators that move 60 to 90 days before the renewal conversation. Track both, and define every metric at the account level.
| Stage | Primary metric | Leading indicators in product data |
|---|---|---|
| Activation | Activation rate (accounts reaching the value moment within a window) | Time to first value, setup completion, first invite sent, first core workflow run |
| Feature adoption | Breadth and depth of core-feature use per active account | Features touched per account, depth within the two or three retention-correlated features, second-week return |
| Expansion | Net revenue retention precursors (the product-qualified lead) | Rising weekly active users in an account, usage in new modules, invites from a new domain, hitting plan limits |
| Churn precursors | Account health decay | Champion stops logging in, weekly use decaying to monthly, feature breadth narrowing, billing or export-page spikes |
Activation
Activation is whether a new account reached its first real value moment inside a window, measured per workspace rather than per login. For reference, the OpenView 2023 Product Benchmarks put the median activation rate at 30 percent across the SaaS products surveyed, so half of products convert fewer than three in ten sign-ups to a first value moment. Useful as a sanity check, useless as a target to copy.
That copying instinct is the trap. Facebook's seven friends in ten days is the number everyone reaches for. It was a real internal finding from the growth team that put Facebook on the path to a billion users, and Chamath Palihapitiya described it openly. But copying it is the classic mistake. As Mixpanel's own writing on the subject puts it, magic numbers are an illusion: useful inside one product, not a law you can borrow. The Geckoboard team made the sharper point, that the seven-friends number confused a generation of growth people about correlation and causation. The friends did not cause retention. They were a sign of it.
Find your own value moment. Take a broad cross-section of your accounts that retained, work backward to the behavior they share that your churned accounts did not, and confirm it actually predicts retention before you optimise for it. That is a week of analysis, not a number lifted from a deck.
Feature adoption
Adoption is breadth and depth of use across the two or three features that correlate with renewal, not a raw count of features touched. The honest version measures whether an account returns to the core workflow in week two, because a single setup session followed by silence reads as adoption in a vanity dashboard and as churn in the data.
Expansion
Expansion is the product-qualified-lead signal: rising weekly active users in an account, usage in new modules, invites from a new domain, and plan limits getting hit. It is worth instrumenting well, because tracking product-qualified leads or accounts increased the likelihood of fast growth by 61 percent in the OpenView 2023 Product Benchmarks. The signal only pays off if it crosses into the CRM where sales can act on it.
Churn precursors
Churn shows up in product data before it shows up on the renewal line: a champion who stops logging in, weekly use decaying to monthly, narrowing feature breadth, and spikes on the billing or export pages. The stakes are real, since median private B2B SaaS gross revenue retention sat near 88 percent in 2024, meaning roughly one revenue dollar in eight leaks before any expansion (ChartMogul SaaS Retention Report, 2024).
For the reporting layer that turns these account signals into a board-grade cadence rather than a wall of charts, the mechanics live in our GA4 reporting guide, built on GA4's own reporting surfaces. Product analytics and web analytics answer different questions; you usually run both.
Event taxonomy is the part that decides everything
The taxonomy is the boring decision that determines whether the tool you bought is useful in eighteen months. Get it wrong and every funnel, every cohort, every retention grid is built on event names nobody trusts.
The convention worth adopting is object-action naming: a noun plus a past-tense verb, with context in properties rather than in the event name itself.
- Good:
Report Createdwith a propertyreport_type: pdf. One event, queryable by type. - Bad:
Created PDF Report,Created CSV Report,Created Dashboard. Three events that should have been one event with a property.
Keep a tracking plan as the single source of truth for which events exist and what properties they carry. Segment's tracking-plan templates and Amplitude's data-taxonomy guidance are the canonical references here, and both predate the current tool you will choose. The plan is the durable asset. The tool is replaceable.
The two failure modes sit at opposite ends. One is auto-capturing every click until the event list is a swamp nobody can read. The other is trying to design a perfect taxonomy in a spreadsheet before you ship a single instrumented event, which at a seed-stage company means you ship nothing for a month. Neither is discipline. Discipline is shipping the events that map to your four lifecycle stages first, then governing the taxonomy as the product grows. Start narrow, expand on evidence.
The four tools, by what they are actually for
PostHog, Mixpanel, Amplitude, and Heap are not interchangeable. They sit in different places and price on different units. Pricing models below are stated as categories with the vendor page linked so you can confirm the current free-tier numbers yourself, because every one of them moves at least once a year.
| Tool | What it is for | Pricing model | Free tier |
|---|---|---|---|
| PostHog | All-in-one, engineering-led. Analytics plus session replay plus feature flags plus surveys in one product. | Usage / event-based. Unlimited team seats. | Yes. 1M events and 5K session recordings per month, resets monthly. |
| Mixpanel | Best-in-breed pure-play analytics. The friendliest interface for a non-technical founder building funnels and cohorts. | Event-based. | Yes. Capped at 1M monthly events. |
| Amplitude | Enterprise-grade behavioral analytics with deeper data science. | MTU-based, with an event-volume option. | Yes. Starter plan to 10K monthly tracked users or 2M events. |
| Heap (Contentsquare) | Auto-capture analytics: records every interaction without manual instrumentation. Acquired by Contentsquare in December 2023. | Session-based, increasingly bundled into Contentsquare enterprise contracts. | Limited, and standalone pricing transparency has fallen since the acquisition. |
The honest read: PostHog is the engineering-led default because it removes two line items, session replay and feature flags, that you would otherwise buy separately. Mixpanel wins on interface for a product-led team. Amplitude is where you land when behavioral data science becomes a job, not a side task. Heap's auto-capture is genuinely useful early but the post-acquisition pricing story now points toward enterprise, so weigh that before you build on it.
One word on the product-qualified lead, because expansion measurement is where most teams underbuild. A PQL is not a single user crossing an event threshold. OpenView Partners, who codified the term inside the product-led growth movement around 2016, define it as an account that has reached real value and also fits your ideal customer profile. That means blending usage signals with firmographic fit, then pushing the score into Salesforce or HubSpot so sales acts on it. Usage alone sends your team free-tier tourists.
The most affordable user-behavior analytics for B2B SaaS startups
The most affordable user-behavior analytics for a B2B SaaS startup in 2026 is PostHog on its free plan. Not because it is the cheapest analytics tool in isolation, but because it is the cheapest stack. It bundles product analytics, session replay, and feature flags under one free meter (1 million events per month, per the PostHog pricing page in June 2026), with unlimited team seats. Choose a pure-play tool instead and you add a session-replay subscription and a feature-flag subscription to see the same picture.
Three realistic tiers, by stage:
| Stage | Typical stack | Why |
|---|---|---|
| Pre-seed / bootstrapped | PostHog free, or Mixpanel free, wired directly via SDK | Zero spend, value in a day, no data warehouse needed |
| Seed to Series A, under 50 staff | A CDP such as Segment or RudderStack feeding one analytics tool, plus a raw dump to a basic warehouse | Decouple tracking from the destination so you can switch tools without re-instrumenting |
| Series B and beyond | RudderStack to Snowflake or BigQuery, transformed by dbt, pushed to CRM via reverse ETL | Now you need to join product, billing, and CRM data, and you have a data team to run it |
Read the free-tier numbers off the vendor page on the day you decide. The allowances above were accurate in June 2026 and they will drift. Mixpanel runs an aggressive startup program; Amplitude's free Starter is real but the MTU cap bites faster than an event cap if your product has many light users. The runner-up to PostHog is Mixpanel, mostly on interface: a non-technical founder gets to a useful funnel faster.
The stack you choose at seed stage should survive to Series A. That is the whole reason the middle tier puts a CDP in front of the analytics tool: it means switching tools later costs a config change, not a re-instrumentation project. If the same site also needs the tracking baked in correctly from the first line of code, that is where leapbuzz website development and analytics work meet, before the bad-taxonomy debt accrues.
Build versus buy, and the consent constraint
Buy first. A warehouse-native stack, Snowflake or BigQuery plus dbt plus a BI tool plus reverse ETL, only earns its keep once you have a data team and need to join product events with billing and CRM data at a level a packaged tool cannot reach. Before that, a SaaS product-analytics tool gives you retention grids, conversion funnels, and behavioral cohorts in days that an engineer would spend weeks rebuilding, badly.
There is no event-volume number that triggers the switch. The trigger is organisational: a data team exists, and the analysis you need crosses product, revenue, and support data. Anyone who tells you to migrate at a specific million-events-per-month line is selling a heuristic, not advising you. The composable data warehouse is a destination for data-mature teams, not a starting point for a five-person product org.
Then there is consent, which is where confident advice gets companies into trouble. The claim that first-party product analytics is automatically exempt from consent is wrong. Under the ePrivacy rules, storing or reading non-essential information on a user's device needs consent, and a persistent analytics identifier in a cookie or local storage usually counts as non-essential. Running PostHog behind a reverse proxy on your own domain fixes ad-blocker loss and third-party data leakage, both real wins, but it does not by itself remove the consent requirement. France's CNIL offers a narrow audience-measurement exemption that is hard to qualify for; it does not generalise across the EU.
That tension, between the measurement you want and the consent you can lawfully obtain, is a strategy decision before it is an engineering one. It sits at the centre of how we approach analytics and insights and AI visibility work for B2B SaaS clients in technology, alongside the broader marketing technology stack the product analytics has to slot into. Measure the lifecycle, keep the taxonomy clean, and let the budget tier follow the stage. The tool is the easy part.