There’s nothing quite as optimistic as the Photoshop mockup of a new dashboard design. Everything is perfect. There is the perfect amount of data, all of the labels fit perfectly, and each graph has the perfect number of elements.
And then, there is reality. Unfortunately, these “perfect” designs don’t always hold up when confronted by reality. Even when you use real data when creating your dashboard, things can change over time, turning those carefully constructed charts and graphs into a confusing mess.
1. Zero Data States
While too much data can quickly ruin a graph, when you first launch your shiny new dashboard, the problem is more likely to be no data at all. As you start your design, think through what you’re going to show people when the information simply isn’t there—or isn’t robust enough to provide any value.
Displaying a message like “No widgets were sold during this time period” can prevent people from hitting reload a half dozen times trying to fix the weird blank graph they’re seeing.
Remember, zero data states can also be caused by users setting overly aggressive filters or searching unusual keywords. So even if you think your dashboards will always have enough data to provide value, establishing guidelines about how and when to tell users why their report is blank is always a good idea.
2. Filtering by Time
Graphs and charts are often great until they’re not. Too many points of data, too many rows, too many contrasting colors—these things happen gradually over time. A chart that shows data by time may look great for a while, but after several months or years, it becomes crowded and unreadable. That’s why you need to specifically consider how you’ll let people filter by time.
This isn’t quite as straightforward as it seems. Users will want to filter some data by month. Other data makes more sense by day, quarter, or year. Sometimes what your users really want is the rolling 30-day average. And, of course, sometimes you absolutely have to allow customized date ranges. The filtering complexity depends entirely on what your users are trying to get out of the information.
The other thing to remember about filtering by time is that you don’t always need to launch a new dashboard with it. Filtering by month isn’t terribly useful when your users have a week’s worth of data. But you do have to plan for it, because someday that perfect graph is going to look like an eye chart. If you decide to add time-based filters later, make sure to set a date at which they’re going to become necessary. Also, make sure that you’re testing your dashboard regularly with real data (and real users!) so you know immediately when more filters are required.
3. Flexible Charts and Graphs
There are cases where even the most comprehensive filtering system won’t save you. Sometimes, the reality of your data simply doesn’t fit the graph you want to use.
Pie charts are perhaps the best example of this. They can be useful for showing how two or three different segments compare to each other, but what if the segments keep growing? How will a pie chart look if it’s split into five pieces? Or 10, or 30? Trust me on this, it will look awful.
Some types of data will naturally grow over time, not just in numeric value, but also in variety. For example, maybe a company only makes widgets in three colors this year. What about the holiday colors, though? Or the new summer line for 2019?
For data like this, make sure you’re choosing graphs that can scale more flexibly. I’m a fan of bar charts, personally, since they can grow vertically, and, as a bonus, it’s easier to show long labels on the y-axis.
Unlike segments in graphs, data in tables is guaranteed to get larger over time. It’s sort of the point of tables. Let’s say you build a table of all the big accounts who are buying widgets. Sure, it looks great if there are only five or 10. But hopefully, over time, the company will grow past that. What does it look like when there are 50 or 100 different accounts? It looks long!
This is an easy one, but it’s also very easy to forget when you’re designing with fake data. Paginate your tables.
Finally, you may have to admit that one size doesn’t fit all. Allowing your users to do some customization of the data means they can decide which information matters to them most. This can rein in the out-of-control growth over time of widgets on dashboards.
However, as with all customization, be careful that you’re not allowing users to create a terrible experience for themselves. It’s one thing to let users hide a few graphs or charts that might not be relevant to their jobs, or to let them splice the data by time periods or different attributes. That can be a huge benefit to them and you. But if you’re not careful, people may start tweaking every single aspect of the experience—colors, fonts, labels, and more—which will cause you enormous headaches when you want to make changes down the line or simply ensure brand consistency.
Great dashboard design is about more than a bunch of pretty charts and graphs. Great dashboards work for all your users and all of their data for as long as they use your product. That takes a little planning and a lot of thought, but the results are absolutely worth it.