How To Keep Your Users | Startup School

Understanding user retention is crucial for startup success, yet most founders struggle to interpret the data properly. David Lee, a Y Combinator group partner who helped build Google Photos to over one billion users, shares the definitive framework for measuring whether you've truly made something people want through cohort retention analysis.

Key Takeaways

• Flat curves matter more than absolute numbers - The only metric that truly matters is whether your retention curves flatten over time, not the height at which they plateau
• Track cohorts, not total users - Group new users by acquisition period and follow their individual journey rather than analyzing your entire user base together
• Choose meaningful actions and timeframes - Select user actions that correlate with real value delivery and time periods that match your product's intended usage frequency
• Avoid common measurement mistakes - Don't pick overly broad time periods or superficial actions that inflate your numbers without reflecting genuine engagement
• Focus on accumulating users over time - Successful products retain even small percentages of users indefinitely, creating compound growth through user accumulation

Understanding Cohort Retention Fundamentals

Cohort retention tracks individual groups of new users over time instead of analyzing your entire user base together. This approach reveals how users continue engaging with your product after their initial experience.

The Three Essential Components

To implement cohort retention properly, you must define three critical elements. First, determine how to isolate cohorts - typically by grouping users based on when they first used your product. Week-by-week or month-by-month cohorts work well for most startups.

Second, select an action that qualifies someone as an "active user." Simply opening your app isn't sufficient. Choose actions that correlate with users receiving genuine value from your product.

Third, pick the appropriate time period for measuring subsequent usage. This should align with your product's intended usage frequency - daily for social apps, weekly for utilities, or quarterly for travel services.

Real-World Action Examples

Instagram might track users who "viewed three or more posts" rather than simple app opens. Uber could focus on "completed rides" instead of app launches. Google Photos measured users who "tapped to view photos full-screen" because this indicated genuine engagement with stored memories.

The best action filters out users who touch your product without receiving real value while identifying those who experience your core benefit.

Interpreting Retention Curves Correctly

The shape of your retention curve matters infinitely more than absolute retention percentages. A product retaining 20% of users indefinitely outperforms one that starts at 80% but continuously declines to zero.

Why Flat Curves Indicate Success

Flat retention curves enable user accumulation over time. Without this pattern, you're trapped on a treadmill - constantly acquiring new users while losing existing ones. Even retaining a small fraction of each cohort creates compound growth potential.

The only thing that matters is whether your cohort curves get flat. The shape of the curve is what matters, not the absolute number.

Google Photos demonstrated this principle perfectly. Despite 80% of users leaving quickly, the remaining 20% stayed active weekly indefinitely. This pattern gave Lee confidence that Google Photos would eventually reach 20% of all humans - which happened within four years.

Visualizing Retention Data

Triangle charts provide the foundation for cohort analysis. Each row represents a cohort (users acquired in January, February, etc.), while columns show retention in subsequent time periods. Converting these numbers to percentages and line graphs reveals retention curve shapes clearly.

The diagonal line in triangle charts represents single calendar periods, showing which cohorts contributed to total usage during specific months. This perspective helps identify whether growth comes from new acquisition or retained users.

Common Measurement Mistakes That Mislead Founders

Most founders unknowingly sabotage their retention analysis through predictable errors. Recognizing these pitfalls prevents self-deception about product-market fit.

Picking Overly Broad Time Periods

The biggest mistake involves choosing time periods that are too large for measuring retention. Quarterly or semi-annual periods make retention appear artificially high because users have more opportunities to return and be counted as active.

At Bump, Lee's team progressively widened their measurement periods from weekly to monthly to quarterly as investor meetings approached. Each expansion made retention look better while obscuring the reality that users weren't finding lasting value in the product.

Selecting Superficial Actions

Choosing actions that are too easy to achieve inflates retention metrics without reflecting genuine engagement. App opens or website visits can be gamed through notifications and alerts, creating false positives in your data.

Google+ exemplified this mistake by counting users as "active" simply for seeing notification bells in Gmail's corner. These weren't genuine social network users - just people checking what the red notification meant.

Using Payment as the Primary Action

Counterintuitively, tracking only whether users pay often provides misleading retention data. Users typically stop using products before canceling subscriptions. Most people have streaming services they pay for but haven't used in months.

A better approach combines payment status with actual product usage: "paying AND actively used a core feature this month."

Improving Retention Through Product and User Changes

Once you've identified poor retention, several strategies can flatten your curves and increase retention levels.

Product Improvements

Direct product enhancements often yield the most dramatic retention improvements. New use cases, reduced latency, and simplified workflows all impact retention curves visibly. You'll see newer cohorts performing better than older ones as improvements take effect.

Successful product improvements create a clear pattern in retention data - recent cohorts flatten faster and at higher levels than historical cohorts, indicating your changes are working.

Better User Acquisition

Sometimes you've built a great product but target the wrong customers. Acquiring different user types can dramatically improve retention without changing your product.

Google Photos experienced this when marketing executives decided to target Gen Z users specifically. Despite successful user acquisition, these cohorts showed terrible retention. Young people simply don't prioritize photo memory management and reminiscing - activities central to Google Photos' value proposition.

Enhanced Onboarding and Activation

Improving first-user experiences often provides the highest return on investment for retention improvements. Many products work well but fail to guide users into successful usage patterns.

Focus on understanding what users did before discovering your product and how you want to change their behavior. Bridge this gap through thoughtful onboarding that demonstrates value quickly and clearly.

Network Effects

Products with network effects - where additional users improve the experience for existing users - naturally see retention improvements over time. Social networks, messaging apps, and sharing platforms benefit from this dynamic.

If your product has potential network effects, focus on building dense, engaged networks around existing users rather than just expanding total user count.

The Ultimate Goal: Upward-Trending Retention

The holy grail of retention isn't just flat curves - it's curves that trend upward over time. This pattern indicates users find increasing value in your product as they continue using it.

Building Toward Compound Growth

Upward-trending retention curves combined with consistent user acquisition create powerful compound growth. Each month retains more users while adding new cohorts, building a "layer cake" of accumulated users from different time periods.

When visualized, this creates a beautiful chart where total monthly active users grow steadily, composed of thick layers representing users retained from many previous cohorts. This pattern indicates you're building toward a multi-billion dollar company.

Balancing Quantitative and Qualitative Insights

Cohort retention curves won't tell you what to change - they only indicate whether your current approach is working. Combine quantitative retention analysis with qualitative user feedback to understand both whether you're succeeding and how to improve.

If you look at your cohort retention curves and they don't get flat, the one thing you can be sure of is that you need to get out there, talk to your customers, understand what's going on, and hopefully make something that they want.

Conclusion

Cohort retention provides the most reliable quantitative method for determining whether you've made something people want. Focus on achieving flat retention curves rather than impressive absolute numbers. Most founders who master this analysis discover they haven't yet built something people truly want - but armed with this knowledge, they can iterate toward product-market fit with confidence.

The path from flat retention curves to upward-trending curves to sustainable compound growth represents the journey from early-stage startup to potential unicorn. Start measuring your cohorts properly today, and let the data guide your product development decisions.