Every feature looks like a hit in its first week. The launch email goes out, the curious click through, the usage chart spikes, and the team moves on to the next thing, quietly convinced the last one worked. Most of the time, nobody checks again.
Marcus checked. He pulled usage for the six features his team had shipped that quarter, not on launch day, but ninety days later. Three of them were effectively dead. Two of the three had the biggest launch-week numbers of the batch.
The launch-day spike
A launch spike measures curiosity, not value. People try the new thing because it is new. The only question that matters is whether they come back to it once the novelty is gone, and that question is invisible on launch day. You have to wait, and most roadmaps never do.
So teams keep building on top of features that already flatlined, because the last real signal they saw was a spike, and a spike feels like a yes.
Usage is not adoption
Two different questions hide inside “is this feature doing well?” One is usage: did someone touch it? The other is adoption: did it become part of how they work? A feature can have high usage and near-zero adoption, a thousand people who tried it once and never returned. Signalpad keeps the two separate, because they lead to opposite decisions.
Does it stick?
Stickiness is just adoption measured across a window instead of a moment. Look at who is still using a feature seven, thirty, ninety days after they first touched it. The curve either holds or it falls off a cliff, and the shape tells you everything the launch number hid.
| Feature | Launch week | 90 days later | Verdict |
|---|---|---|---|
| Voice notes | 61% | 9% | Fading |
| Saved views | 44% | 38% | Sticks |
| Bulk export | 52% | 6% | Cut |
| Shared links | 39% | 41% | Double down |
Same launch, four different outcomes. Voice notes looked like the winner in week one and was nearly gone by the quarter. Shared links looked modest and quietly became a habit. Only the ninety-day column could tell you which was which.
Double down, or cut
Once stickiness is visible, the roadmap gets easier and a little colder. The features that held are where the next investment goes: deepen them, guide more users into them, build the adjacent thing. The features that spiked and died get sunset, or a second attempt with a real hypothesis about why they failed the first time.
“We stopped congratulating ourselves on launch week and started reading the ninety-day column.”
Shipping with a memory
The compounding win is not any single decision. It is that the release process gains a memory: every feature leaves a curve behind it, and those curves start predicting the next one. You stop shipping on vibes and start shipping on what your own users have already taught you.
That is the difference between a product that ships features and a product that accumulates the ones worth keeping.