Michael Gaigg: Über UI/UX Design

Dashboard Design

Posted by Michael Gaigg

Example of a dashboard (Google Analytics)

I'd like to summarize my notes from the excellent talk titled "Designing Infographics for Web Applications" (watch the 54 minute video).

This talk was given by Des Traynor at the MIX 2011 and is a must-see for anybody deciding to implement a dashboard or even designing one. It explains the goals of data visualization and the benefits for the business.

While the original talk is about infographics I named this article Dashboard Design because this was my big take-away and I see you guys applying this knowledge mostly to dashboards as well.

Know your Audience

For which role is the dashboard designed for?

Typical roles are:

• CEO
  • strategic view
  • focus on long term
  • high-level overview
  • simple summary
• Analyst
  • query driven analysis
  • precision required
  • emphasis on trends
  • emphasis on correlations
• Operations & Logistics
  • focus on current status
  • issue & event driven
  • are things ok?

Know your Domain

What is the area (domain) of your audience?
Domains could be

• Sales
• Marketing
• Customer/technical Support

What are you trying to answer? Which questions are important?

• Average?
• Percent change?
• Totals?
• Popular?
• # sales?

Know your Data

Communicate single figure

Use when:

• context is obvious
• precision is required
• past & future is irrelevant to the user

Can have different states (red, green,...)

Single figure with context

Use to ask:

• How are we doing lately?
• Any problems on the horizon?

Sparklines help (trend up or down?)

Analysis of a period

Use to show:

• ...all the key moments of this month

Line chart needs lots of data points to be good, but then they are really good!

Analysis of a period versus target

Use to ask:

• Did we hit our sales figures?
• Are we fulfilling our quota?
• Is performance ok?

Actual versus target chart; focus on delta! Use bar chart.

Breakdown of a variable

Use to ask:

• What age groups are buying our stuff?
• What countries are we big in?

Pie chart becomes bad when too many entries; sorted bar chart is better; avoid decorating with colors.

Breakdown over time

Use to ask:

• How has the composition changed over the last year?

Line chart is good; stacked bar chart is really difficult to read.

Know your Visuals

Visuals are broken down into two categories: Quantity and Category.

Quantity

• Lines (length, width), that's the best
• Color intensity (shades of red,...)
• Size
• Quantity
• Speed (only useful in motion charts)

Category

• Line type (dashed, dotted,...)
• Color (red, blue, green,...)
• Shape (triangle, rectangle,...)
• Location (map)

Know your Style

Remove the junk (= stuff that doesn't change, e.g. gradients, noise,...). There is no reason to fight for impact, this is not a shiny print population, it's a dashboard.

Closing

The process of designing a dashboard should follow these rules (in order of importance):

1. They have to say something and be meaningful
2. Dashboards & Visuals evolve, i.e. a lot of data is necessary to make sense
3. Start with the basics: Rows from the database
4. Add insight as you need it: Sorting, comparison,...
5. Add a yearly view only after a year has passed
6. Include insights and actions
7. Consider adding projections