The PM's Analytics Problem
Most product teams collect far more data than they use. They have event tracking everywhere, dashboards for everything, and yet the PM still cannot answer the basic question: "Is this feature working?"
Amplitude is one of the most popular product analytics tools, and for good reason. It is built specifically for product teams, not marketing teams or data engineers. But having Amplitude does not mean you are getting value from it. The difference between PMs who get actionable insights and PMs who just have another dashboard comes down to setup, discipline, and knowing which analyses actually drive decisions.
Setting Up Your Event Taxonomy
Event taxonomy is the foundation of everything you do in Amplitude. Get it right and every analysis is straightforward. Get it wrong and you spend more time debugging data than analyzing it.
Events vs properties
An event is something the user does: Sign Up, Create Project, Invite Team Member, Export Report. A property is a detail about the event or the user: plan_type: pro, project_count: 3, invite_method: email.
The most common mistake is creating too many events. If you have Click_Button_Save_Project_Settings and Click_Button_Save_Profile_Settings, you have two events where you should have one event (Save Settings) with a property (settings_type: project | profile).
Naming conventions
Pick a convention and enforce it across your entire team:
Events: Verb + Object in Title Case: Create Project, Submit Form, View Dashboard
Event properties: snake_case: project_type, form_name, dashboard_id
User properties: snake_case: plan_tier, company_size, signup_source
Document your taxonomy in a shared spreadsheet or wiki page. Every new event should go through a review before implementation. Undisciplined event creation leads to duplicate events, inconsistent naming, and unreliable data within months.
Essential events for SaaS products
At minimum, track these events for any SaaS product:
| Event | Why it matters |
|---|
Sign Up | Top of funnel, conversion tracking |
Complete Onboarding | Activation measurement |
[Core Action] | Your product's "aha moment" (e.g., Create First Dashboard) |
Invite Team Member | Virality and expansion signal |
Upgrade Plan | Revenue event |
Export/Share | Value realization |
Return Visit (automatic) | Engagement and retention |
Define what your product's core action is. For Figma, it is creating a design. For Slack, it is sending a message. For Amplitude itself, it is running a chart. This core action is the event you will build most of your analyses around.
Funnel Analysis
Funnel analysis is the analysis type PMs use most in Amplitude. It answers: "Where are users dropping off in a specific flow?"
Building a useful funnel
Define the conversion goal. Start with the end: what is the desired outcome? "User completes onboarding" or "User makes first purchase."
Map the steps. List every step the user takes to reach that goal. For onboarding: Sign Up → Verify Email → Create Workspace → Invite Team Member → Complete First Task.
Set the conversion window. How long should users have to complete the funnel? For onboarding, 7 days is typical. For a checkout flow, 1 hour. The window affects your conversion rate significantly, so choose it based on realistic user behavior.
Reading funnel results
Amplitude shows the conversion rate between each step. Look for:
The biggest drop-off. If 80% of users complete step 1 but only 30% reach step 2, that transition is where you should focus.
Time between steps. Amplitude shows median time between steps. If users take 3 days between "Create Workspace" and "Invite Team Member," that gap is an opportunity for a nudge email or in-app prompt.
Segment differences. Break the funnel by user property (plan tier, signup source, company size) to see if different user groups convert at different rates. A funnel that looks healthy in aggregate might have a severe problem for a specific segment.
Common funnel mistakes
Too many steps. A 10-step funnel is not useful because every step introduces noise. Keep funnels to 3 to 6 steps focused on the key conversion moments.
Mixing optional and required steps. If "Invite Team Member" is optional in your onboarding flow, do not include it in the main funnel. Create a separate funnel for the invitation flow.
Ignoring time windows. Default to "any time" and your funnel looks great because it includes users who converted 90 days later. Set a realistic window.
Retention Analysis
Retention is the metric that tells you whether your product delivers sustained value. Amplitude's retention charts are some of the most important analyses a PM can run.
Setting up retention
A basic retention analysis in Amplitude requires two things:
Start event: What counts as the beginning of the period? Usually Sign Up or First Visit.
Return event: What action indicates the user is still engaged? Usually your core action or simply Any Active Event.
Reading retention curves
The retention curve shows what percentage of users who started in a given period performed the return event in subsequent periods (days, weeks, or months).
Key things to look for:
The initial drop. Most products see 50% to 70% drop-off between day 0 and day 1. This is normal. The question is how steep the curve is.
The flattening point. A healthy retention curve flattens somewhere, meaning a stable group of users continues to return. If your curve never flattens and keeps declining toward zero, you have a retention problem.
Where the curve flattens. If it flattens at 20%, you retain 20% of new users long-term. Whether that is good or bad depends on your product category. B2B SaaS tools typically need 40%+ monthly retention to build a sustainable business.
Retention by cohort
Break retention by cohort to see whether your product is improving over time:
Time-based cohorts: Users who signed up in January vs February vs March. If later cohorts have better retention, your product improvements are working.
Behavioral cohorts: Users who completed onboarding vs those who skipped it. This tells you whether your onboarding flow actually predicts long-term retention.
Segment cohorts: Enterprise users vs SMB users, or users from paid acquisition vs organic. This informs where to invest in growth.
Cohort Comparison
Cohort analysis in Amplitude goes beyond retention. It lets you compare any behavior across user groups to find what drives success.
Building behavioral cohorts
Create cohorts based on actions users have or have not taken:
Power users: Users who performed the core action 10+ times in the last 30 days
At-risk users: Users who were active last month but have not logged in this month
Activated users: Users who completed onboarding within 7 days of signup
Feature adopters: Users who used a specific feature at least once
Comparing cohorts
Once you have cohorts defined, compare them across any metric:
Do power users use feature X more than average users? If yes, feature X might be driving engagement. If no, power users are getting value from something else.
Do activated users retain better than non-activated users? If the difference is large (e.g., 50% vs 15% at 30 days), your onboarding flow is one of the most important things to optimize.
Are at-risk users concentrated in a specific segment? If all your at-risk users are from a particular signup source or company size, the problem might be targeting, not product.
Turning cohort insights into action
The point of cohort analysis is not to produce charts. It is to produce decisions:
If activated users retain 3x better, invest in improving onboarding completion rates
If power users all use the same 3 features, make those features more discoverable for new users
If a specific segment churns at 2x the rate of others, either fix the product for that segment or stop acquiring them
Amplitude vs Mixpanel for PM Use Cases
Both tools serve product analytics, but they differ in ways that matter for PM workflows.
Where Amplitude is stronger
Retention analysis. Amplitude's retention charts are more flexible and easier to configure than Mixpanel's. If retention is your primary concern, Amplitude has the edge.
Behavioral cohorts. Defining and comparing cohorts is more intuitive in Amplitude. The cohort builder is a standout feature.
Notebooks. Amplitude's notebook feature lets you combine charts, text, and analysis into a single document. This is useful for weekly product reviews or sharing analysis with stakeholders.
Collaboration. Amplitude makes it easier to share analyses, create team dashboards, and maintain a shared analytics workspace.
Where Mixpanel is stronger
Event-level exploration. Mixpanel's event stream and individual user timelines are easier to navigate. When you need to debug a specific user's experience, Mixpanel is faster.
JQL (custom queries). Mixpanel's query language gives power users more flexibility for custom analyses that do not fit standard chart types.
Simplicity for small teams. Mixpanel's interface is slightly simpler for teams new to product analytics.
The honest answer
For most PM use cases (funnels, retention, cohorts, feature adoption tracking), both tools work well. The choice often comes down to which one your team already uses, which one your data team prefers to support, and pricing. Switching analytics tools mid-flight is expensive and disruptive. Pick one and invest in learning it deeply rather than chasing feature comparisons.
Common Setup Mistakes
These mistakes waste months of PM time and undermine trust in your data.
Tracking everything
"Let's track everything and figure out what we need later" sounds practical but creates a mess. You end up with 500 events, most of which no one looks at, and the events you actually need are buried in noise. Start with 20 to 30 core events and add more only when you have a specific question you cannot answer.
No event documentation
If only the engineer who implemented the tracking knows what user_action_v2 means, your analytics system has a bus factor of one. Document every event: what it tracks, when it fires, what properties it includes, and when it was last updated.
Inconsistent identity management
Users who sign up, log out, and return on a different device can appear as two different users if identity management is not configured correctly. Work with your engineering team to implement proper identity resolution (Amplitude calls this "User Identification" and has specific APIs for it). Bad identity data makes every downstream analysis unreliable.
Ignoring data quality
Set up a monthly data quality check:
Are any events logging significantly more or fewer times than expected?
Are event properties populating correctly, or are many values null?
Do the numbers in Amplitude match your backend database within an acceptable margin?
Treat data quality like product quality. If you would not ship a feature with a 15% error rate, do not trust analytics with a 15% data gap.
Making Amplitude Work for PMs
The PMs who get the most from Amplitude share these habits:
Define your key metrics first. Before opening Amplitude, write down the 3 to 5 metrics that matter most for your product area. Then build dashboards around those metrics. Do not explore aimlessly.
Check retention weekly. Make retention your default view. It is the single best indicator of product health, and it changes slowly enough that weekly checks catch trends without creating noise.
Share analyses, not screenshots. Amplitude lets you share live chart links. Use these instead of screenshots so stakeholders can explore the data and apply their own filters.
Combine quantitative and qualitative. Amplitude tells you what users do. It does not tell you why. Pair every quantitative finding with qualitative research (customer interviews, session recordings, support tickets) before making product decisions.
Clean up quarterly. Archive dashboards and charts that no one has viewed in 90 days. Remove deprecated events. A clean analytics workspace, like a clean backlog, is more useful than a large one.Free Resource
Enjoyed This Article?
Subscribe to get the latest product management insights, templates, and strategies delivered to your inbox.
No spam. Unsubscribe anytime.