To quote a Japanese proverb, vision without action is a daydream.
I was working on dashboards within a logfile collection last week, when I was challenged how to determine the right components for the dashboard, and had several insights.
- There is not one "dashboard" - there are dashboards for each stakeholder. The data of interest to each stakeholder depends on his or her entry perspective.
- Dashboards are meant to provide a synopsis, not the whole story. A user needs to be able to identify issues at a glance, with as little involvement as possible. Pie charts, bar charts, and averages all provide a high level indicator that can be drilled down if something looks amiss.
- The level of information isn't the same across stakeholders. Trends over time may be significant to an executive, but an engineer may need specific details at the transaction level. A system wide response time number is relevant to a user who knows tolerances and expected values, but a global number may mask small components within the average. Drill-downs, trends and thresholds are all contextual to the user viewing the dashboard.
- It isn't enough for the data represented to be interesting. Seeing where in the world our users are connecting from may be interesting, but have no impact on me or my plans. I have to be able to act on the information, or I am just presenting trivia.
This leads me to my overarching recommendation on dashboards: