Over the past year or two we’ve helped design maybe a dozen or so dashboards, for applications ranging from expert systems (internal staff managing hundreds of interactions per day) to apps (one person managing their fitness each week).
But across all of those projects there are a number of core lessons that apply to all. So we’ve gathered our key findings and top tips below.
1. Choose the right clothes
We all know that one size does not fit all. And we all equally know that what looks good on a model will often look somewhat different when we buy it, take it home and try it on ourselves. Oh, the shame.
Which is my way of saying, ‘know the customer’. Whilst there are many elements of data design and visualisation that will carry across most audiences, this is not true for all. Let me give you one simple example.
One dashboard we redesigned was for a fitness app (specifically around eating and calories). The initial data views were far too complex, and we worked with the client to simplify them. The end result had perhaps half the data of the initial design, but performed much stronger in testing.
Compare this to a second dashboard we redesigned, built for financial agents managing transactions and applications. Upon initial view the dashboard was way too crowded, with tiny fonts being used to display huge and complex tables. But upon reviewing the requirements and talking with those agents, this was exactly what they needed - no amount of stripping down and simplifying the data was going to work, as they needed all the data on display at different (and somewhat unguessable) times. So the dashboard remained somewhat cramped, but kept them absolutely happy with the outcome.
The right clothes for the right person.
2. There’s a reason Police sketch artists exist
Take a look at the following choice. Both describe the man who made off with the bank notes in a swag bag, which do you think would be easier to remember after a day or two?
Police sketch artists exist because it’s far easier to consume some information in visual form than in data form; and tables and graphs are a great example of this. Bumps and dips on graphs are far easier to visualise and understand at a glance than are rows of numbers with smaller or higher values.
But this is a generalisation and isn’t always true. For example we worked on a dashboard where we took out a number of visualisations, because they were simply not helpful - they showed pies and lines but these meant nothing to the audience at all.
So a sketch is better than a description most of the time, but sometimes you only need the headline.
3. Ogre’s are like onions - and it’s not about tears
When trying to explain that Ogre’s are complex beings in Shrek, Shrek tells Donkey that Ogres are like onions. They have layers.
The same is true for a good data dashboard. Try to show everything in one layer, and you’ll almost always get it wrong.
The rule of three’s is a great place to start:
Top level - overview of state of play
Mid level - checklist of items that make up the state of play
Low level - detail for the checklist items
Let’s take the example of a financial banking app. The CommBank app does this extremely well. The top level is visible as soon as you open the app, it gives you your state of play, how much money you have today, along with access to other features and the drill-down.
You can then select View accounts, and see the mid level. You see your accounts and cards listed with their own state of play information, showing how that makes up the overall picture. But it’s light at this stage.
When you select an account you then hit the low level, and you’re looking at the details that make up the state of play for one account - daily transactions and items. Three tiers, and the user has the ability to stop at any level. They aren’t forced to be a donkey, they can peel their financial data like an onion - or an Ogre. Not that you peel Ogres, but you get what I mean.
4. Look for the bubbles
There is a story about the first iPod that goes like this. Steve Jobs wanted the engineers to make it smaller, but they complained there was no way to reduce the size any more, they simply couldn’t deliver on that requirement. So Steve threw the iPod into an aquarium, and when bubbles rose to the surface he said ‘Those are bubbles, that means there’s space in there. Make it smaller.”
That may or may not be true, but it’s a great example of good dashboard design. Only with dashboard design we’re not making it smaller, we’re making it leaner.
You’re looking for what you can remove, whilst still keeping the dashboard useful. Most dashboards we see that are failing are doing so not from lack of information, but from a glut of it. The more information there is, the harder it is to find and seek the insights you need.
So, look for the bubbles - remove everything you can, pair it back, and simplify. Keep it clean and simple, lean and mean.
5. Walk the tightrope
As with so much in life, good design is about balance. Visual balance, goal balance, and a balance between simplicity and functionality.
In designing a great dashboard you’ll be finding the balance between:
Power users (who want everything up front and quickly accessible), and other users (who want simplicity and layered structure)
Generalists (who need a little of everything the system has to offer), and specialists (who only need a few things but more of them)
Task-based users (who want to only see what they need to know right now) and holistic users (who want to see the wider picture all the time)
And that’s just the start.
It’s not easy, and it often involves compromise - but there are routes towards happiness for everyone, too. For example:
Customisation - allowing the audience to customise what they see to their own needs (as long as you provide elements they can use to do this)
Personalisation - personalising the experience people encounter based on their behaviour, path, archetype or other information.
Mode-setting (switching the UI based on activity, to better suit the audience mode at the time)
None of this works without contextual user testing - or if it does, you’ll never know it. So, test, test and test again. Good dashboard designs will usually require several rounds of iterative testing and enhancing before you’ll know for sure that they are doing the job.
And if you want to know more - contact us.