The Dashboard Nobody Opens: a Post-Launch Audit

Looker Studio and most BI tools will quietly tell you view counts if you know where to look. We checked, six months after shipping a dashboard a VP had requested with real urgency in a planning meeting. Three views total, two of them us during QA.

How the request actually happened

In the meeting, the ask was 'can we get visibility into this.' That's not a specification, it's a feeling. We translated it into a specific dashboard with specific charts because that's our job, and we built a genuinely good one — clean, fast, correctly calculated metrics. What we never did was go back and ask what decision the VP intended to make by looking at it, on what cadence, triggered by what event.

'Can we get visibility into this' is a request for reassurance, not a request for a dashboard. Build the wrong one and you get a dashboard nobody opens and a VP who still feels uninformed.

What usage data actually told us

When we finally asked the VP directly, it turned out the actual need was a monthly five-minute verbal summary in an existing meeting, not a self-serve tool. A dashboard sitting in a tab the VP never opened couldn't deliver reassurance on its own — it required someone to proactively look and then explain it, which a Slack message with three numbers would have done just as well, for a fraction of the build cost.

• We now ask 'what will you do differently based on what this shows you' before building any requested dashboard — if the answer is vague, the dashboard will be too.
• We added a 90-day usage check to our standard process, not as a vanity metric but as a trigger to go re-interview the requester.
• Replacing the dashboard with a monthly automated summary email, generated from the same underlying query, got read every month.
• The underlying data layer we'd built for the dashboard wasn't wasted — it just needed a different delivery mechanism than a dashboard.

Nobody opening a dashboard isn't a UX problem you fix with better charts. It's almost always a sign the delivery format never matched how the request-maker actually consumes information, and that mismatch gets built in at the request stage, long before anyone opens a BI tool.