The opening question of almost every BI meeting: "which dashboard do you want?". Wrong. The right question is: "which decision do you need to make?". A dashboard is a means — decision is the end. But the BI industry spent 20 years selling the means as if it were the end, and the result is on every C-level wall: televisions with colored charts that no one looks at anymore.
This piece is about how Tableau (or Power BI, or Looker — the tool matters less than it seems) can become a language of executive decision, not a showcase. And when it becomes badly-governed self-service, it produces the opposite problem: each team with its own version of the same number.
The invisible cost of the vanity dashboard
A vanity dashboard has three trademark signs:
- It shows too much. Fifteen KPIs on the same screen, three colors, four filters, two time periods. All "important", none actionable.
- It recommends nothing. It points at numbers — sales, churn, NPS. It doesn't say what to do.
- It lives outdated. Built for a 2024 question. In 2026 the business changed, no one updated it, everyone pretends they still use it.
The cost is triple: time of whoever built it (visible), attention of who should decide (invisible) and — worst — false sense of governance. A board that looks at a dashboard convinces itself it is data-driven. It isn't. It's performing data-drivenness.
Three questions every dashboard should answer
The rule we use to review BI before any project. If the dashboard doesn't answer one of three, it probably doesn't justify existing.
- What is happening now that I need to decide on this week? Focus on actionable. Not "YTD sales", but "which 3 accounts need intervention today".
- What trend affects the next quarter? Focus on directional. Not "churn per month", but "this cohort is leaving in pattern X and demands a response" — and that depends on defining churn before modeling churn, a more common mistake than it seems.
- Where is my intuition wrong? Focus on counterintuitive. Not "show me my numbers", but "show me where my mental model fails".
A dashboard that answers none of this is decoration.
The best executive dashboard is the one that kills the next dashboard. Each view must earn its place — not occupy it by inertia.
Anatomy of a view that triggers decision
When we build executive analytics, we follow a simple pattern:
Numeric headline, not chart
The first thing on the screen is the number that matters — big, undecorated. Like: "3 strategic accounts at churn risk in the next 4 weeks". Not a bar chart. Not a time series. The number, in plain language, with timeframe.
One-line comparative context
Right below: "vs. 1 account in the previous quarter; vs. average of 1.8 in the last 4 quarters". Comparison is what gives meaning to the number. Without comparison, a number is trivia.
Drill that ends in action
Clicks that open details — affected accounts, likely reason, recommended next step. Not just data. Next steps. Who has to talk to whom, by when, with what offer.
Tableau (and equivalents) delivers this pattern well when you build it. But the tool alone doesn't — that's the point.
What Tableau does well (and what it doesn't replace)
Tableau is excellent at three things: rapid visual exploration over modeled data, distribution of views across the organization, and personalization by persona/role. The choice between Tableau, Power BI, Looker and Metabase by company size and cloud stack is a decision prior to the tool itself; any of the four requires a solid data model underneath to work.
It does not replace:
- Data modeling. A bad model makes Tableau pretty and imprecise. Invest in the warehouse/dbt — where documentation is the real win — before Tableau.
- Business conversation. The view only helps if there was serious discovery with whoever will decide.
- Automated recommendation. For that, ML/AI enters — Tableau visualizes, doesn't think.
The combination that works: clean warehouse + well-defined business model + Tableau as reading layer. Each piece in its place. Worth remembering: "clean" here means good enough for the use case, not absolutely clean — waiting for perfection is the fastest way to never publish the dashboard.
The final dashboard
The best quality metric for a dashboard: how many real decisions came from it in the last quarter. Not visits, not screen time. Decisions. If zero, kill it and rebuild. The same principle applies to product metrics that become "north dust" — the problem rarely is the metric, it's the system around it.
Companies adopting this rule cut 60–80% of their dashboard count and — not by chance — start trusting what's left. A decent data consultancy delivers this, not a report.
Questions that keep coming back
To close, the questions that come up most often when this topic reaches the meeting room.
How do I know if an executive dashboard is actually working?
Count how many real decisions came out of it in the last quarter — that's the metric, not visits or screen time. A dashboard that produced zero decisions is decoration, no matter how pretty, and the honest move is to kill it and rebuild starting from the decision the board needs to make.
A shortcut for diagnosis: does it answer any of the three questions that justify its existence? What to decide this week, which trend affects next quarter, or where the executive's intuition is wrong. If it answers none, the problem isn't design — it's purpose.
How many KPIs should an executive dashboard show?
Far fewer than the market default — and the screen should open with a single numeric headline, not fifteen indicators. Fifteen KPIs, three colors, and four filters on one screen is the trademark of the vanity dashboard: everything "important", nothing actionable.
The pattern that drives decisions is lean: the number that matters in plain language (like "3 strategic accounts at churn risk in the next 4 weeks"), one line of comparative context right below it, and a drill-down that ends in a recommended next step. Every extra view has to earn its place — not occupy it by inertia.
Is Tableau better than Power BI or Looker for board-level dashboards?
The tool matters less than it seems — all three deliver the executive pattern well when you build it. The choice between them (and Metabase) follows company-size and cloud-stack criteria, and it's a decision prior to the dashboard itself.
What none of them replaces is what's usually missing: solid data modeling underneath, serious discovery with whoever will decide, and recommendation — which is the job of people or ML, not the visualization layer. A bad model makes any BI tool pretty and imprecise; the combination that works is a warehouse clean enough for the use case, a well-defined business model, and BI as the reading layer.