Your Retention Is Lying to You
A single retention number masks huge variance. Some cohorts retain at 60%. Others at 5%. Some segments churn in Week 1. Others flatten and stick. US teams that drill into retention by cohort, source, and behavior discover the real picture, and often find their "good" retention hides a leak they've been ignoring.**
Your retention looks fine. 40% Week 4. Not amazing, but acceptable. You report it. You track it. You assume it's representative.
Then you segment. Paid cohorts: 18% Week 4. Organic: 52%. Mobile: 25%. Desktop: 48%. Activated users: 65%. Non-activated: 8%.
The aggregate was a lie. It averaged strong segments with weak ones. The weak ones were bleeding. You didn't know because you never looked.
Why Aggregate Retention Lies
The Averaging Problem
If 60% of your users come from a channel that retains at 50%, and 40% come from a channel that retains at 10%, your overall retention might look like 34%. Fine. But that 10% segment is a leak. You're acquiring users who don't stick. The average hides it.
The Cohort Problem
Retention improves as you fix the product. Newer cohorts should retain better than older ones. If you only look at "all users," you miss:
- Are recent cohorts improving?
- Is a specific signup week an outlier?
- Did a product change move the needle?
The Behavior Problem
Users who activate retain. Users who don't, don't. Mixing them into one retention number is meaningless. The "real" retention of activated users might be strong. The "real" retention of non-activated users might be near zero. The blend obscures both.
How to Get the Truth
1. Cohort Retention by Signup Week
Track retention by when users signed up. Week 1, Week 2, Week 3 return rates. See if newer cohorts improve. See if a specific week was an anomaly.
2. Retention by Traffic Source
Which sources retain? Paid vs. organic. Google vs. referral vs. direct. The source that retains best is the one to scale. The one that churns fastest is the one to fix or stop.
3. Retention by Activation
Compare retention of users who hit the activation event vs. those who didn't. The gap is usually dramatic. It tells you activation matters, and that improving it will improve retention.
4. Retention by Behavior
Users who use Feature X in Week 1 retain better. Users who don't, don't. Segment by behavior. Find the actions that predict retention. Optimize for those.
The Tools You Need
Cohort tables with filters. Retention by week. Color-coded or trended. Ability to segment by source, device, activation status, behavior.
SingleAnalytics includes retention cohort analysis. Filter by any property. See which segments retain and which don't. The truth emerges fast.
Real Impact
A US startup reported 42% Week 4 retention. They segmented:
- Organic, activated: 68%
- Organic, not activated: 12%
- Paid, activated: 45%
- Paid, not activated: 4%
Their "good" retention was driven by organic activated users. Their paid non-activated users were a retention black hole. They'd been scaling paid. They shifted to organic and focused on activation. Retention improved. CAC dropped. The aggregate had hidden the problem for months.
The Mindset Shift
Before: "Our retention is 40%. We're okay." After: "Our retention is 40% average, but paid non-activated is 4%. We're fixing that."
Before: "Retention is stable." After: "Retention is stable for activated users. For everyone else, it's collapsing. We're improving activation."
Segment. Find the lie. Fix the leak.
Ready to see the real retention story? Analyze retention by segment with SingleAnalytics and stop the leak.