designing risk dashboards people trust
Risk dashboards are routinely abandoned within months of launch. The failure is rarely visual.
Every risk team we’ve met has a dashboard graveyard: tools that launched with a demo and a memo, got checked daily for a month, then quietly lost to the spreadsheet they were supposed to replace. The post-mortems usually blame adoption or training. We think the diagnosis is simpler. People stopped looking because they stopped believing the numbers.
Trust in a dashboard isn’t a feeling — it’s a track record of the numbers surviving scrutiny. Design for that and the visual questions mostly answer themselves.
every number needs an address
The first time a risk officer catches a dashboard disagreeing with the source system, the dashboard loses. Permanently. So the single most important design decision is making every figure clickable down to its origin: this exposure number comes from these positions, valued at this timestamp, by this method. We call it giving the number an address.
This is more plumbing than pixels, which is why it gets skipped. But a plain table where every cell can defend itself beats a beautiful chart that can’t. In our experience the drill-down gets used rarely — its existence is what does the work.
People trust numbers they could check, even when they don’t.
show staleness, not just state
Risk data is never uniformly fresh. Positions update in near-real-time; counterparty ratings arrive weekly; some vendor feeds limp in monthly. A dashboard that presents all of it as equally current is lying by layout. The honest version puts the age of the data on the surface — a timestamp on every panel, and a visible flag when a feed has gone quiet.
This feels like admitting weakness. It’s the opposite. The dashboard that says “this number is four days old” is the one whose fresh numbers get believed. The one that never admits staleness gets all of its numbers discounted, all the time.
design for the second audience
A risk dashboard has two audiences: the person who reads it daily, and the committee that sees a screenshot of it quarterly. Most designs serve the first and betray the second — the screenshot arrives stripped of the context that made it legible, and someone spends a meeting explaining what the colors mean.
The fix is to make every view exportable as a document that carries its own context: title, date, thresholds, and the definitions of whatever is colored red. If the dashboard can produce the committee page itself, the numbers in the meeting always match the numbers on the desk — which is half of what “one source of truth” actually means in practice.
fewer numbers, held to a higher standard
The strongest predictor of a dashboard surviving its first year, in our projects, is how much the team was willing to leave off it. Twelve numbers with addresses, honest timestamps, and committee-ready exports will outlive sixty numbers with none of those things. Risk teams don’t need more information — they need information that holds up when someone pushes on it. Build that, and the dashboard stops being a reporting obligation and becomes the place the committee argues in.