In practice
You ship CSV export. Six weeks later 4 percent of accounts have used it. Before you decide nobody wants it, find out how many people ever saw it. If most didn't, the feature isn't the problem. The route to it is.
Discovery is the first encounter and first try. Adoption is regular, repeated use. A feature can be discovered and not adopted, or adopted by the few who found it.
Some teams use the same words for a different job: finding which features are worth building. That is product discovery, and it happens before you build. Here the term means what people searching for it usually mean, users finding features that already exist. The user-experience name for the property is discoverability: whether someone can perceive and understand a function the first time they meet it.
How to measure it
- Discovery rate = users who opened or saw the feature at least once ÷ users it is relevant to.
- Adoption rate = users who came back to it ÷ users who tried it.
- Time to first use = how long a new account takes to reach the feature.
- Compare discovery before and after you change where it appears or how you announce it.
You need an event for the first view or use of each feature. Without one, you can see only who used it, not who never found it.
Common mistakes
- Judging a feature by usage before checking whether anyone saw it.
- Announcing it only in a changelog or an email. Anyone who reads neither never learns it exists.
- Putting every feature into one first-run tour. People skip a long tour, and what they skip they don't learn.
Why it matters for product builders
Usage numbers alone can't tell you why a feature is quiet. Without a discovery number, you may cut a good feature or rebuild one people couldn't find.
People take the shortest path to what they came to do. They don't explore a product to see what else is there, so a feature that isn't in their path stays unseen.
What changes in AI-first products
Agents let you ship features faster than users can learn them. The gap between what exists and what people know about is wider than it was.
How people find things hasn't changed much: a clear place in the product, a prompt at the moment it helps, and a note to the people who asked for it. With Annsa, when a job ships, Share Back drafts a note to each person who asked, and you send it. Those people are the first audience for a new feature.
See how Annsa does it: Share Back →