A note from a recent project where the second redesign performed worse than the first — and what that taught me about measuring the right thing.
Version one of the onboarding finished at 41% completion. We redesigned around the drop-off point, shipped, and watched completion drop to 36%.
The drop-off had been the wrong metric. Users who finished v1 were not activating. Users who quit v2 had self-selected out of a product that was not for them. The lower completion was a healthier funnel.
I do not trust a single funnel metric without a behavioural one next to it anymore.
Logos rarely matter inside the product. What matters is whether the brand has decided how the product behaves when something goes wrong.
What changes when your user has 90 seconds between rooms and one hand on a tablet — and why "minimal" stops being a style choice.
Generative tools have not replaced the discovery phase. They have collapsed the cost of being wrong, which changes how much you should explore.