There is a specific moment when a SaaS founder realizes their analytics is the wrong kind. It usually happens in a board meeting. Someone asks "which feature keeps customers around?" and the founder opens GA4, which can report bounce rates and traffic sources in loving detail, and cannot answer the question at all. The tool is not broken. It was built for a different question.
This hub is about the other kind of analytics: what people do inside the product, on the web and in apps, and how to measure it without drowning in tooling.
What is the actual difference between web and product analytics?
Web analytics is acquisition-shaped: sessions, sources, campaigns, conversions. It answers "is the marketing working?" Product analytics is behavior-shaped: users, features, funnels, cohorts, retention curves. It answers "is the product working?"
The two look similar because both involve events and charts. They diverge on the unit of analysis. Web analytics thinks in visits. Product analytics thinks in people over time, which is why it needs a stable user ID and why retention, not conversion, is its native metric. If your revenue is subscriptions, people-over-time is the shape of your business, and visit-shaped data will keep giving you polite non-answers. The full comparison, including which one you need first, is in web analytics vs product analytics.
Why your product data is worse than your marketing data
Marketing tracking gets attention because ad money makes its gaps visible and painful. Product tracking decays quietly: events added by whoever built the feature, no naming convention, user IDs that reset between the marketing site and the app, and an activation event nobody has questioned since it was invented during onboarding week.
The symptom is always the same. The dashboards disagree with the database, the funnel has a step nobody believes, and every strategic discussion starts with twenty minutes of arguing about whose number is right. That argument is the cost. Measured in senior salaries, it is a large one.
The questions worth instrumenting for
Strip product analytics to the decisions it feeds and you get a short list:
- Activation: what early action separates users who stay from users who vanish? This defines onboarding's job. How to find it is covered in how to track a SaaS funnel that predicts revenue.
- Retention: who comes back, at what rhythm, and which behaviors precede leaving?
- Feature value: which features correlate with retention and expansion, and which are decoration?
- Conversion: where do trials and freemium users cross into paying, and what did they do differently beforehand?
Ten events chosen against these questions beat two hundred events chosen by autocomplete.
The mobile complication
Apps add two twists. First, the tooling shifts: Firebase's SDK feeding a GA4 property is the default Google path, and where that stops being enough is its own decision, covered in Firebase Analytics vs GA4. Second, ad measurement on iOS changed permanently with App Tracking Transparency: user-level attribution requires an opt-in most users decline, and the working replacements are aggregate by design. What still works is in mobile attribution after ATT.
The honest headline: on mobile, precise user-level ad attribution is not coming back. Businesses that accept this and build on aggregates outperform businesses that keep buying dashboards promising the old world.
Do you need PostHog, Mixpanel, Amplitude?
Sometimes. Not as often as their pricing pages suggest. The decision is not "which product analytics tool is best" but "which decision am I unable to make today?" A seed-stage product with a hundred users needs conversations, not cohort charts. A product with real usage and a retention question needs an event pipeline with a stable user ID, and any of the major tools will do that job; I compared the philosophies in GA4 vs PostHog.
What I push clients toward is sequencing: define the funnel and the activation hypothesis first, instrument the handoff between site and product properly, and only then choose tooling. Tools amplify instrumentation; they do not create it.
Where to start
If you run a SaaS or an app and suspect your product data cannot be trusted, the fastest diagnostic is unglamorous: pick one number the business repeats (trial conversion, activation rate, whatever is on the board deck) and trace it from the dashboard back to raw events. Either the trace holds, and you have a foundation, or it breaks somewhere specific, and you have found the real project. That trace is the first thing I do in every product analytics audit, and it has never once come back boring.