The brief was to add three new charts to an existing dashboard. We delivered the three charts and removed eleven others, and that turned out to be the part of the engagement the client mentioned in every subsequent conversation.
How the dashboard got that way
Eighteen months of individual feature requests, each reasonable on its own — a sales manager wanted a breakdown by region, a founder wanted a trend line, a new hire wanted a metric they'd used at a previous job. Nobody had ever been asked to remove anything, because removing a chart looks like removing a capability rather than fixing a problem.
What we actually measured
Before touching any code, we instrumented which charts were viewed, by whom, and how often, for three weeks. Four of the fourteen charts had not been viewed at all in that window. Another five were viewed by a single user each, usually the person who had originally requested them.
The conversation that mattered
We took the usage data to the client rather than a design opinion. "Nobody has looked at this in three weeks" is a harder thing to argue with than "this looks cluttered", and it turned what could have been a defensive conversation about taste into a shared decision about evidence.
- Kept: the charts used weekly by more than one role.
- Archived, not deleted: charts used by a single person, moved to a page reachable in two clicks rather than the front screen, because removing someone's tool entirely without warning erodes trust fast.
- Added: the three new charts the brief asked for, now visible because the screen wasn't fighting for space with eleven others.
The technical side was the easy part
Cutting eleven charts and their queries reduced the dashboard's load time meaningfully, since several of the removed charts ran expensive aggregate queries on every page load regardless of whether anyone looked at the result. That was a welcome side effect, not the point.
Most dashboards don't get worse by having a bad chart added. They get worse by nobody ever taking one away.
What we'd tell a client earlier next time
That we plan to measure usage before making changes, at the start of the engagement rather than after we noticed the clutter, so it reads as a stated part of the process rather than a critique arriving after the fact.