Quick answer
Product analytics is the practice of recording what people do inside a digital product, as events tied to people over time, and analysing it to see where they finish a task, where they stop and who comes back. It answers questions about behaviour, not only how much traffic arrived.
What is product analytics?
Product analytics is the practice of recording what people do inside a digital product (a web app, a mobile app, a site with sign-in) and analysing those actions to see where they succeed, where they stop and who keeps coming back. Amplitude defines it as the process used to understand how customers engage with digital products. The raw material is an event (a person did something at a time, with details attached) and a person (the same human across visits, devices and sessions). The output is answers to questions about behaviour, not counts of traffic.
Two things are easy to confuse with it. A dashboard of page views is not product analytics just because it sits inside a product. And product analytics is not product management: it supplies evidence for decisions, it does not tell you how to run discovery or prioritise a roadmap.
Product analytics vs web analytics
The difference is not that one uses events and the other does not. Google Analytics 4, a web analytics tool, logs sessions and page views as events too. The difference is the question each is built to answer, and therefore the unit it reports in.
| Web analytics | Product analytics | |
|---|---|---|
| Main question | How many people came, from where, and what did they view? | What did people do with the product, and did they come back? |
| Usual unit | Visit, page view, visitor | A person (or an account) and their actions over time |
| Typical reports | Traffic sources, top pages, bounce rate, goals | Funnels, retention, paths, segments, feature adoption |
| Time horizon | One visit at a time | Many sessions: day 1, week 2, month 3 |
| Who asks | Marketing, SEO, content | Product, growth, support, engineering |
| Weak spot | Cannot say whether the same people returned or finished a task | Says little about where visitors came from before they signed up |
Most teams need both, and a tool that keeps them in one project saves a lot of reconciling. A sign-up page that gets 3,000 visitors a week is a web analytics fact. Whether those 3,000 people ended up creating a project, and whether they were still around in week 2, is a product analytics fact.
The data model: events, properties and people
Every product analytics question is answered from the same four building blocks. Mixpanel's documentation describes the first three: events record a user's interaction with the product, event properties add detail such as an item's type and price, and user profiles hold attributes for each identified person.
- Event. A named action:
signup_started,project_created,report_exported. Names should be few, consistent and written down in a tracking plan. - Event properties. Details of that one action: plan, platform, error code, order amount.
- Person and user properties. Who did it, and what is stable about them: signup date, plan, country. A person only exists when you can recognise them across visits, which usually means a login id.
- Account (optional, for B2B). The company or team above the person, so you can ask whether the customer is still using the product, not just one user.
Collection can be automatic (autocapture of clicks and form submits) or deliberate (events you send from code). Automatic capture gets you started; named events carry the business meaning.
Five questions and the analysis that answers each
| Question | Analysis | What the result looks like |
|---|---|---|
| How many people finish this multi-step task, and where do they stop? | Funnel | Count and rate at each step, biggest drop marked |
| Do people come back after they start? | Retention | Share of a starting group active on day or week N |
| What do people do after (or before) a given point? | Path analysis | Most common next pages or events |
| Who are the people doing X, and how do they differ? | Segments and behavioral cohorts | A named group you can reuse in other analyses |
| Is a feature actually used, and by how many? | Event analysis, distribution | Unique people and times per person |
Eight worked cases are in product analytics examples, and the choice of headline metrics is covered in product analytics framework. The point here is the shape: each question is about people and steps, which a page-view table cannot answer.
Worked example: the same week, two readings
Sample data: a project-management app with a free sign-up. In one week, 3,000 visitors reach the sign-up page.
| Step | People | Step rate | Overall rate |
|---|---|---|---|
| Visited the sign-up page | 3,000 | 100.0% | |
| Started the form | 1,200 | 40.0% | 40.0% |
| Verified email | 840 | 70.0% | 28.0% |
| Created a project | 504 | 60.0% | 16.8% |
| Invited a teammate | 151 | 30.0% | 5.0% |
How to recompute: step rate = people at this step ÷ people at the previous step (504 ÷ 840 = 60.0%); overall rate = people ÷ 3,000 (504 ÷ 3,000 = 16.8%).
Three things this table tells you that a page-view report would not:
- Page views overstate reach. The "create project" page had 1,140 page views, but from only 640 distinct people. Of the 840 who verified their email, 640 opened it (76.2%) and 504 of those finished (78.8%).
- The drop is concentrated. Split the step from "verified email" to "created a project" by device: desktop 330 of 480 (68.8%), mobile 174 of 360 (48.3%). That 20-point gap is where to look first, for example in a session replay of mobile users.
- Return behaviour separates people. Of the 504 who created a project, 214 were active again in week 2 (42.5%). Among the 151 who invited a teammate, 98 returned (64.9%); among the 353 who did not, 116 did (32.9%).
The last point is an association, not proof. People who invite colleagues may simply be more committed. Product analytics shows where to look and which hypothesis to test; it does not replace testing the change itself.
Common mistakes
- Tracking everything. Hundreds of unnamed events make every analysis a search. Start from the five questions above and track what they need.
- Counting events instead of people. "5,000 exports" could be 5,000 people or 50 people exporting 100 times. Look at unique people and times per person.
- Treating anonymous traffic as users. Visitors who never sign in can be counted in totals but cannot be followed across devices or put in a person-level segment.
- Inconsistent names.
Signup,sign_upandsignup_doneare three events to a database. Use one name per action. - Reading association as cause. Retention differences between groups are hypotheses. Test the change before you scale it.
- Averages without distribution. A mean of 4 uses per person can hide 90 people at zero and 10 at forty.
Product analytics in TapCub
TapCub treats web, app and mini-program events as one project, so the sign-up example above runs on one data set. Here is how the same analysis goes.
- Collect events. The website script, the iOS / Android / Flutter / React Native / HarmonyOS SDKs, mini-program SDKs and the server HTTP interface all send events to the same project. Autocapture needs the Pro plan or higher. Record what you intend to collect in All-platform analytics › Data › Tracking plan.
- Identify people. Call
tapcub.login()when someone signs in. People then appear in All-platform analytics › Users › People with a day-by-day timeline. Whether a login binds to a profile depends on the identified-mode setting in the compliance center. Visitors who never sign in stay in the aggregates and have no profile. See identify users. - Build the funnel. In All-platform analytics › Behavior › Behavior analysis the Funnels tab is the default: 2 to 20 steps, a window of 7 days by default (up to 180), and people as the default unit. The largest drop is marked, and you can click a step to see who dropped. See build a funnel.
- Check return. The Retention tab shows whether a starting group is active on later days, weeks or months. Save the "created a project" group as a segment in All-platform analytics › Users › Segments & tags and apply it as a cohort to other analyses. See create user segments.
- Know the limits. TapCub measures behaviour; it does not run A/B tests or split traffic, and it does not score leads or send messages. Treat a retention gap as a hypothesis, then check the change in your own release process.
See also: Product Analytics Platform: How to Choose One (Event Model and Identity First).
See also: Product Analytics Tools: What to Compare Before You Pick One.
See also: Product Analytics: What It Is, Key Metrics and How to Start.
Sources
- Amplitude, The Amplitude Guide to Product Analytics.
- Mixpanel Docs, What is Mixpanel? (events, event properties and user profiles).
- Google Analytics Help, Automatically collected events (sessions and page views are events in GA4).
- Contentsquare, What is product analytics?.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Product Analyst
Focuses on event analysis, conversion funnels, and retention, using clear metric definitions and structured analysis to explain product performance.