Dashboards lie by omission

The most dangerous reporting I've encountered wasn't fabricated. Every figure was computed correctly from real data by people acting in good faith.

It was still wrong, because a dashboard is an argument about what matters, and arguments can mislead entirely through what they leave out.

Three failure modes, all of which I've shipped myself.

1. The average that hides the tail

"Median response time: 2.1 hours." Excellent. Everyone's happy.

P90: 31 hours. P99: four days.

The median describes the experience of the customer having a normal day. The tail describes the experience of the customer having a problem — who is, by definition, the customer whose opinion of you is being formed right now, and the one most likely to tell other people about it.

Averages and medians are optimized for reassurance. They tell you the system usually works, which you already believed. Every gauge worth publishing needs a tail number next to it, because the tail is where the churn is.

2. The rate that hides the denominator

"Conversion is up 12% this month." Real number, correctly computed.

Traffic was down 40%. The remaining visitors were disproportionately people who already knew what they wanted. Conversion went up because the top of the funnel collapsed and only the pre-sold were left.

Any ratio can move for two reasons, and the good one and the bad one look identical in the ratio. Never publish a rate without its numerator and denominator visible. It costs one line of layout and it eliminates an entire genre of confident wrong conclusion.

3. The snapshot that hides the trend

A number without history is a fact without meaning. "Churn: 3.2%." Is that good? Nobody can say. Good relative to what?

Show four weeks and it becomes information. Show a target and it becomes actionable. A snapshot invites everyone to bring their own prior about what's normal, and people's priors are anchored to whatever they last heard in a meeting.

What to add

Four things, every gauge, no exceptions:

The tail, not just the center.
The denominator, not just the rate.
The trend, not just today.
The target, so the reader doesn't need context you have and they don't.

That's four times as much ink per number, which is precisely why you should publish a quarter as many numbers. Eleven well-dressed metrics beat forty naked ones, every time, because the naked ones require interpretation and interpretation doesn't happen on a busy Tuesday.

The omission that matters most

Here's the one almost nobody includes: what isn't on the dashboard.

A short line at the bottom — not currently instrumented: partner-sourced deals, enterprise renewals, the migration path — does more for honest decision-making than any tile above it. It converts an unknown unknown into a known one, and known gaps get filled.

Without it, a dashboard implicitly claims completeness. Readers assume that if something mattered, it'd be shown. That assumption is almost always false, and it's the mechanism by which a company convinces itself it's watching the business while an entire segment runs dark in plain sight.

The law of clarity isn't "publish numbers." It's glass reveals — and glass that only covers the flattering half of the pipe is a mirror.