Product Usage Analysis SOP

A product usage analysis shows which parts of your product people actually use, how frequently they come back and which early action predicts that they will stay. It takes two to three weeks once tracking is in place. Founders tend to believe their favourite feature is the reason people stay; usage data regularly points at something smaller and less glamorous. The goal is to find your product's real centre and build everything around it.

What is product usage analysis and when should you use it?

Product usage analysis studies behaviour inside the product: feature by feature, user by user. Where customer retention analysis answers "do people stay?", usage analysis answers "what do the people who stay do differently?"

It rests on three ideas from well-known growth work:

• Breadth, depth and frequency: how many features a user touches, how heavily, and how many days.
• The power user curve: Andrew Chen of Andreessen Horowitz described a histogram of how many days in a month each user is active. A curve with a bump on the right, a group active 20+ days a month, shows a product with a committed inner circle.
• The early predictive action (the "aha moment"): an action in the first days that strongly predicts retention. Facebook's growth team, as Chamath Palihapitiya described in talks, focused on getting new users to 7 friends in 10 days.

Use this SOP:

• When retention flattens and you want to know why the retained group stays.
• Before cutting or adding features, to see what people rely on.
• When your first-week flow loses people and you need to know what to steer them toward.
• When different customer types use the product differently and you need to choose who to serve.

What do you need before you start?

• Event tracking for every main feature, with user IDs. The digital behavior tracking SOP covers setup.
• At least 200 active users and 4+ weeks of history.
• A retention definition from your retention analysis, so you can label each user as stayed or left.
• Session recordings (Hotjar or PostHog) to watch what numbers can't explain.

How do you run it, step by step?

1. List features and name their events

Write every feature in a sheet with the event that proves someone used it: "Exported a report", "Shared a board". Skip passive events like page views. Ten to twenty feature events is enough for most early products.

2. Build the weekly feature use table

For each feature, calculate the share of weekly active users who used it at least once that week. Sort from highest to lowest. Features used by under 5% of active users are candidates for removal or hiding; the top two or three are your real product.

3. Plot the power user curve

Count, for each user active in the last 30 days, the number of distinct days they were active. Plot a histogram from 1 to 30 days. A curve that drops and stays low means casual use; a bump at 15 to 30 days means a group of power users worth studying. For weekly products, use days active per week instead.

4. Compare first-week behaviour of stayers and leavers

Split users who joined at least 8 weeks ago into those who are still active and those who left. Compare what each group did in their first 7 days. Look for an action, with a count and a time window, that most stayers did and few leavers did: "invited 2 teammates in 3 days", "created 3 projects in the first week".

5. Test the candidate action

Check the candidate against newer cohorts: do users who reach it retain at twice the rate of those who don't? Correlation is not cause, so confirm with five to eight calls with power users and watch ten session recordings of users who reached the action.

6. Turn findings into decisions

Write three lists: features to put at the centre, features to hide or remove, and changes to the first-week flow that steer new users toward the predictive action. Pass the lists into feature prioritization.

Template you can copy

PRODUCT USAGE ANALYSIS SHEET

Tab 1: Feature use (last 4 weeks)
Feature | Event name | % of weekly active users using it |
Trend vs last month | Keep / centre / hide

Tab 2: Power user curve (last 30 days)
Days active (1..30) | Number of users | % of monthly active users

Tab 3: Stayers vs leavers (users who joined 8+ weeks ago)
Action in first 7 days | % of stayers who did it | % of leavers who did it |
Ratio (stayers / leavers)

Tab 4: Candidate early action
Definition: "[action] [count] times within [days] days of signup"
Retention of users who reached it:      ___%
Retention of users who didn't reach it: ___%
Confirmed in calls/recordings (Y/N):    ___

Tab 5: Decisions
Centre: ___ | Hide/remove: ___ | First-week flow change: ___

Worked example: a project tool for wedding planners

This example is illustrative. A project tool for independent wedding planners has 640 monthly active users and 14 features. The feature use table shows that the client timeline is used weekly by 71% of active users, the budget tracker by 38%, and the built-in mood board by just 4%, even though it took two months to build.

The power user curve has a clear bump: 18% of monthly active users are active 20 or more days a month. In the stayers vs leavers tab, one action stands out: sharing a timeline with a client in the first 5 days. 64% of stayers did it, against 9% of leavers. Users who reached it retained at 3.1 times the rate of those who didn't.

The founder moves the "share with client" button into the first screen after signup, hides the mood board behind a menu and adds the action to her AARRR dashboard as the new activation event.

What should you measure, and when do you stop?

Stop when you have a candidate early action that holds across two cohorts and is confirmed by calls with power users. If no action shows a clear lift, the product has no strong centre yet: go back to the Sean Ellis test to find which segment loves it, then rerun this analysis for that segment alone.

Which mistakes should you avoid?

• Measuring clicks instead of value actions. Opening a menu shows curiosity, finishing a task shows value.
• Treating correlation as proof. Power users do many things because they are power users. Confirm the early action in conversations.
• Studying averages across all customer types. Agencies and freelancers use the same tool differently. Segment first.
• Keeping features because they were expensive to build. Usage decides, sunk cost doesn't.
• Forgetting the users who left. Stayers alone can't tell you what separates them.

How can AI speed this up?

Export the stayers vs leavers table and ask an AI assistant to rank actions by the ratio between groups and propose candidate early actions with count and time windows. AI tools inside PostHog and Mixpanel can also write the queries for you in plain language. Review the logic yourself: check that "stayers" and "leavers" come from the same signup period, and that the AI hasn't picked an action that only happens because someone was already retained.

To practise reading usage data and making product decisions under deadlines, the Fe/male Switch startup game puts first-time founders through these calls in a safe setting.

Where does this SOP fit in your validation plan?

Usage analysis follows customer retention analysis in the Product-Market Fit stage: first confirm people stay, then learn why. Next steps:

1. Rank the features to build or cut with feature prioritization.
2. Update your activation event in the AARRR metrics framework.
3. Link usage to customer goals with customer success metrics.

See all Product-Market Fit SOPs. The F/MS startup playbook module puts usage data into a full plan for an early-stage company.

FAQ

What is an aha moment in product analytics?

It is the early action that best predicts whether a new user stays, such as Facebook's 7 friends in 10 days. You find it by comparing the first-week behaviour of users who stayed with those who left.

What is a good DAU/MAU ratio?

It depends on how frequently the problem occurs. A messaging app needs a high ratio; a monthly invoicing tool does fine with a low one. Track your own trend, and use it only for products meant for daily use.

How many users do I need for product usage analysis?

About 200 active users and 4+ weeks of data give readable patterns. Below that, watch session recordings and talk to users instead.

Which tools are best for product usage analysis?

PostHog, Mixpanel and Amplitude cover events, cohorts and funnels, and all three have free tiers. Pendo adds in-app guides, and Hotjar adds recordings and heatmaps.