Data Analytics Toolkit · 15 min · 180 XP

Designing a dashboard

Pick a few metrics that drive decisions, define them precisely, and give every number something to compare against.

A dashboard is for a decision someone makes regularly: who looks at it, how often, and what will they do differently depending on what they see? If nobody can answer that, you're building a report nobody reads. Start with 3 to 5 headline metrics (KPIs) and put the most important one where the eye lands first, at the top left.

Define every metric in writing, because "revenue" means three things to three people. Revenue: the sum of paid order amounts in the month; refunds and pending orders excluded. Orders: the count of paid orders. Average order value (AOV): revenue ÷ orders. Refund rate: refunded orders ÷ (paid + refunded) orders. Put the definitions on the dashboard itself, or one click away.

A number alone means nothing. Revenue 613: is that good? Give every tile a comparison: the previous period, the same month last year, or a target. Show amounts as a % change and rates as a change in points (refund rate 0% → 20% is +20 points; "+∞%" helps nobody). When there's nothing to divide by, show a dash rather than an error or a zero.

March vs February, as tiles
Revenue   613     ▼ 58% vs Feb (1,472)   ← Feb included one B2B order of 1,173
Orders    4       — flat
AOV       153.25  ▼ 58%
Refunds   20%     ▲ 20 pts (from 0%)

Look at that revenue tile: down 58%, but almost entirely because February had one unusual order. A good dashboard annotates a movement like that, so nobody launches a rescue plan for a problem that isn't there. Avoid vanity metrics (page views, total sign-ups ever) that only go up and never change a decision, and show when the data was last refreshed.

Loading your workspace…