Why Your Retention Curve Is Lying to You (And How I Learned This the Hard Way)
We had a problem at Sonic Linker that I didn't even know existed.
Our 30-day retention sat at 68%. For a new AI SaaS product, that felt incredible. I showed it to the team, we high-fived, and I started planning our growth strategy around it.
Then I dug one layer deeper and realized we were celebrating the wrong thing.
The curve that fooled me
When you look at an aggregate retention curve, you're seeing one averaged story. But products don't have one type of user. They have multiple cohorts with completely different behaviors, and when you blend them together, you hide the truth.
Here's what I found when I split our users:
Power users (20% of signups): 92% retention at day 30. These people used our AI linking feature 10+ times in week one. They got value immediately.
Explorers (35% of signups): 71% retention at day 30. They tried the product, liked it enough to stick around, but weren't deeply engaged yet.
Tire kickers (45% of signups): 31% retention at day 30. They signed up out of curiosity, poked around twice, never came back.
When you average those three groups, you get 68%. It looks healthy. But that number was hiding a massive problem. Almost half our users were never going to convert into real customers, and we were spending time and resources trying to retain them.
The aggregate curve made it look like we had a solid product. The segmented view showed we had an amazing product for 20% of users and a confusing one for everyone else.
What I changed after Finvestfx
I learned this lesson even harder when I joined Finvestfx. We had 20+ enterprise clients on our forex and treasury platform, and our account retention looked great at 89% year over year.
But when I started talking to clients, I realized something uncomfortable. Some were staying because they loved the product. Others were staying because switching costs were too high, or because their finance team hadn't prioritized finding an alternative yet.
Sticky for the wrong reasons is not the same as valuable.
I started tracking what I call engaged retention instead of just login retention. Did they use our core workflow (trade execution and reconciliation) at least twice a week? Did they invite new team members? Did they expand their usage?
Turns out, only 14 of our 20 clients were truly engaged. The other 6 were at high risk of churning the moment a competitor made switching easy enough.
This completely changed my roadmap priorities. Instead of building new features to attract more lukewarm accounts, I focused on deepening engagement with our best users. We improved our reconciliation workflow based on feedback from those 14 active clients. Retention among that group jumped to 96%, and three of them expanded their contracts.
The other 6? Two churned eventually. But I stopped pretending the aggregate number told me anything useful.
How to actually read your retention curve
Here's what I do now, and what I wish I'd done earlier:
Segment by activation. Split users into those who completed your core action in week one versus those who didn't. The difference will be stark, and it tells you whether your onboarding actually works.
Segment by intent. If you can, tag users based on why they signed up (referral source, use case, company size). At Sonic Linker, users who came from product hunt behaved completely differently from users who came from our LinkedIn content. Blending them together made both groups invisible.
Track depth, not just duration. Someone who logs in once a week for three months is not the same as someone who uses your product daily for three weeks then ghosts. Frequency and intensity matter more than tenure.
Look at resurrection. I started tracking how many churned users came back on their own (not because of a win back email). If that number is near zero, you don't have a retention problem. You have a value problem.
The aggregate retention curve is a starting point, not an answer. It's like looking at average revenue per customer without segmenting by plan type. Sure, it gives you a number to put in a deck. But it won't tell you what's actually working or what to fix next.
The real question isn't "are people staying"
It's "are the right people staying for the right reasons".
At Sonic Linker, once I stopped optimizing for the blended retention number and started optimizing for power user retention, our product decisions got sharper. We stopped adding features that appealed to tire kickers. We doubled down on workflows that our most engaged users needed.
Our overall retention actually dropped slightly in the short term (we lost some of the explorers). But our power user cohort grew from 20% to 34% of signups, and their retention stayed above 90%.
That's a much healthier business than one with a pretty aggregate curve and no idea who's actually getting value.
Next time you look at your retention dashboard, don't stop at the top line number. Slice it. Segment it. Ask who's staying and why. The curve might be lying to you, but the segments almost never do.