Every organization I've worked in has a graveyard of dashboards.
They get built with good intentions — someone needs visibility, an analyst spends two weeks getting the data right, the thing launches to mild fanfare, and then... nothing. Three months later it has twelve monthly active users, seven of whom are the analyst's own team checking that it still works.
The tool isn't the problem. I've seen this happen in Tableau, Looker, QuickSight, and Power BI. The pattern is identical regardless of the platform.
The actual problem
Most dashboards get built backwards. Someone decides they want a dashboard, picks the metrics they think matter, builds the visualizations, and then ships it. The question of what decision this will help someone make never gets asked.
A dashboard that doesn't change anyone's behavior isn't analytics. It's a report with better colors.
What I've learned to ask first
Before touching a table or a tool, I ask three questions:
Who makes a decision based on this? Not who will look at it — who will act on it. If the answer is "everyone," that's a red flag. Dashboards built for everyone serve no one.
What would they do differently if the number went up versus down? If the answer is the same either way — or if nobody knows — the metric isn't useful. A metric that doesn't change behavior isn't a metric, it's a decoration.
How often does that decision actually get made? A daily dashboard for a decision that gets made quarterly is waste. Match the cadence of the data to the cadence of the action.
The fix isn't harder — it's earlier
The investment in getting these questions answered before building is small compared to the cost of building the wrong thing. One hour with the right stakeholder, asking uncomfortable questions about what they actually do with data, will save weeks of rework.
The best analytics work I've done started not with a data model but with a conversation about how a team made decisions and what was getting in their way. The dashboard was the last thing we built, not the first.
Get the question right, and the tool almost doesn't matter.