Our first version of the dashboard prioritized information density, we wanted power users to see everything at once. User testing two weeks post-launch showed the opposite problem: people were missing time-sensitive alerts because they were visually competing with a dozen other elements. We revised the hierarchy to surface only 2-3 priority actions by default, with detail behind a expand toggle.

We should have tested that density assumption with users before building the high-fidelity version; we lost about a week and a half rebuilding what user testing would have caught earlier. The tradeoff we made (shipping fast to hit a client deadline vs. validating first) was the right call given the constraint, but it's not one I'd default to again without flagging the risk upfront.