What should the product show when it has nothing left to recommend?
The normal screen had another person to review. The empty state needed to handle the moment when there was nobody left.
Doing nothing was one option. The page could simply stop. That would also make the product look broken.
I wanted a clear card that said there were no more recommendations. It also needed to explain what the person could do next.
This was a small product state, but it affected trust. A recommendation system should be honest about the limits of what it knows. It should not repeat people, invent urgency, or keep producing weak suggestions only to make the screen look full.
The team also needed to decide what counted as a real connection. LinkedIn context was one input. Product activity was another. The private rules and implementation are not part of this article.
The safe public point is the behavior. A person could reach the end of the current set. The product needed to acknowledge that directly.
I asked for the state to be documented and designed before development. That gave product, design, and engineering one shared version of the expected behavior.
The work included the words, the button, and the next action. Connect another source. Return later. Review something already saved. The exact choice could change, but the state could not remain undefined.
The state also needed to persist after a refresh. If the same person returned a few minutes later, the application should still explain that the current set was complete instead of showing a loading screen or beginning the list again.
We also renamed the product channel to product and design. That matched the work in front of us.
We were not adding another recommendation. We were designing what happened when the list was done.