Displaying information in a dashboard: Advice or references

Question

I have to design a dashboard for managers who need to report to their managers on their areas of responsibility. In the past I have just put counts and line charts over time of what they are interested in but this is too overwhelming for some. This is for a case management system where there are new cases and new data for each case every day. At the end of the business cycle which can be years later a case is closed and the information is now tombstone material but still relevant for comparison to the new cases.

I'm looking for best practices or references on how to display information. My reporting tool does grids,charts, drill downs and the usual.

Should I start with an abstraction like percentage change over time and some images to indicate what is changing the most or start with counts over time?

Do users look at the first thing they see and then skip to the end of the page or will they look at the first image that grabs them?

The issues that I want advice on are:

• can you show too much raw data?
• showing what I think is a relevant trend and possibly missing other trends
• pictures first, data later or the reverse?

Answer 1

Perhaps the key challenge you're facing is that of making the information clear.

Buy, borrow or steal (in order of preference) copies of Edward Tufte's classic works.

The Visual Display of Qualitative Information should be your first port of call. You'll learn more from a casual read of this than from almost any other source.

Aside: The graphic on the cover comes from a railway timetable - time runs from left to right, distance vertically. The diagonal lines show the position of each train over time - steeper lines are faster trains. Every train, every stop, every time, shown in one diagram that takes just a few minutes to understand.

His other books are useful as well.

• Envisioning Information
• Visual Explanations: Images and Quantities, Evidence and Narrative
• Beautiful Evidence

Answer 2

If you want to cover your ass, do what most business apps do and include everything (bad UX).

If you want something to be useful, then you need to find out from your clients what the most important metrics are to them. Don't accept the whole "They're all important" story. Some are key and need to be clear on the dashboard, while others can be de-emphasised but available if you want to look. Then there is the final category of ones that just don't matter because they don't affect any decisions.

Finding them is the hard part. One way that you could go about it is by asking for all the metrics that they care about. Then allocate them 100 points and ask them to allocate points to the metrics according to how important they are to them or how much this metric would affect their decision making.

That should give you a better picture of what needs to be there. In general though, simpler = better.

Answer 3

Starting from the features and data available will open up too many possibilities that can end up not being very useful. To turn it around, I'd start by getting a hold of the reports that they use now and asking what they like best and what they hate about those reports. Also try to find out what decisions people want to make from the dashboard, and then provide the data and features to support them.

When you know what the objectives are for the people using the dashboard and what they like and don't like about current reports, you'll be able to decide which data and presentation features to build first.

To answer your specific questions:

• yes you can show too much raw data. A dashboard is ideally for orienting a person to see what has changed recently, how assessing those changes and direct attention to the things that need focus first.

• you can try to show what you think is a relevant trend, but you should be looking to the user base to tell you what trends are relevant to them in a dashboard scenario.

• I can't really answer the pictures vs data question. As JohnGB noted, pictures are a kind of data, and this would need more context.

It sounds like you have a mixed audience for the dashboard, so being able to customize what they see might be worth exploring. It also sounds like you're having to make guesses about a lot of what will serve those people best, so I'd say the best practice to start with is to get access to the people who will use the dashboard and find out what they can tell you.

Answer 4

Yes, there is such a thing as too much data. If you show too much, the important stuff gets lost. You need to figure out what is important and give that the greatest prominence. Maybe even hide the less important stuff away, available via drill-down or whatever. And put the most important stuff where the user will look first (top left).

Abstractions are good if they give the user what they need—that is, if they give them the answer they are looking for without making them work for it. But do they provide enough context?

As for relevant trends, how do you know what is relevant? The only way to know is do some user research to find out.

In addition to the excellent Tufte books that @Bevan recommends, you should check out Stephen Few's books.

I've only read Information Dashboard Design, but I can't recommend it highly enough.

Answer 5

I used to maintain a 70+ page monthly 'dashboard'. I thought it was way too much info and the report was increasing...

I'd go for quantitative graphs first showing the most important trends (where important depends on you client). Quantitative should be bar or line graphs.

Then, I'd show the most important (quality) issues in pie charts where possible.

We're talking about a dashboard. The most important data should be readily available and understandable on the first page. Use colors for clarity: green=good, orange=warning, red=bad...