The dashboard nobody reads is the one nobody designed

At Amazon, I delivered over 50 interactive QuickSight dashboards serving Ops VPs and Directors across 4 countries. Some of those dashboards saved 2,000+ hours annually. Others — honestly — collected dust. The difference was never the data quality, the refresh rate, or the visualisation library. It was whether the dashboard was designed around a decision.

Most dashboards answer questions nobody asked

Here's the pattern I see everywhere: a stakeholder says “I need visibility into X.” An analyst builds a dashboard showing X. The stakeholder looks at it once, says “great,” and never opens it again.

Why? Because “visibility” is not a decision. It's a feeling. Nobody wakes up needing visibility. They wake up needing to know whether to hire more drivers in Birmingham, or whether the warranty claims spike in Q3 is a trend or an anomaly, or whether the EV transition timeline needs to be accelerated.

The dashboards that worked at Amazon — the ones Ops VPs actually opened every morning — were built backwards from a decision. Not “show me fleet utilisation” but “tell me which stations are over-capacity this week so I can redistribute.” Not “show me compliance rates” but “flag me the regions that are about to breach threshold so I can intervene before it's a problem.”

The 3-second rule

Every dashboard I build now follows a simple test: can a VP glance at it for 3 seconds and know whether something needs their attention?

This sounds reductive. It's not. It's a forcing function for clarity. If a dashboard fails the 3-second test, it has too many metrics, unclear hierarchy, or no clear “so what.”

Here's what the 3-second rule looks like in practice:

• Top of the page: One or two KPIs that answer “are we on track?” Green means yes, red means no. That's it.
• Middle: The trend that explains why. A line chart showing the trajectory, with enough context to know if it's getting better or worse.
• Bottom: The drill-down for people who want to investigate. Tables, filters, regional breakdowns. But only after the headline has done its job.

Automating the reporting killed the busywork

When I joined the fleet analytics team, Weekly Business Reviews took analysts 20+ hours to prepare. Twenty hours of copying data from 8 different systems, pasting into slides, reformatting charts, and sending emails. Every single week.

The dashboards I built didn't just visualise data — they replaced the process. Automated pipelines pulled from every source, aggregated into a single warehouse, and fed into self-serve dashboards that updated in real-time. The WBR went from a 20-hour manual process to a 10-minute review.

But here's the part most people miss: the automation wasn't just about saving time. It was about changing what analysts spent their time on. Instead of aggregating data, they were now analysing it. Instead of building slides, they were building models. The 2,000+ hours we saved annually didn't disappear — they got redirected into actual insight generation.

GenAI changed the game again

The latest evolution was building GenAI enablement for analytics. We created internal tooling and documentation that let non-technical stakeholders query data using natural language. The result was a 3x increase in self-service data adoption.

This is where analytics is heading. Not more dashboards — fewer dashboards, but smarter ones. Dashboards that proactively surface anomalies. Dashboards that tell you what changed and why, before you even ask. Dashboards that are less like reports and more like advisors.