17.3.16

Top 10 Dashboard Don'ts


10.  Don't build to committee  If a bunch of people sat around a room and drew out the dashboard on a whiteboard before you've even analyzed the data.... run screaming from the room.  This dashboard will end up being an '18 months to deployment product' and I promise you will want to slit your wrists with a spoon at some point, so pretend your hair's on fire and get out now.



9.  Don't build to spec  This is similar to the Don't above, but usually comes in the format of an e-mail with a list of charts someone wants built to put in their PowerPoint presentation. Example: 1. Pie chart of sales by state, 2. stacked bar chart of profit by state, 3. bubble chart of 10 million customers colored by height... (I'm using the most outrageous examples, but hey, this sh*t happens.).  So build the appropriate chart that helps to see the results and if you have to, build the chart they asked for to show the difference.

8.  Don't build on Excel pivot tables  If you are given a workbook with 18 datasources that are all basically Excel pivot tables, throw it out and ask for access to the original data source. Or pretend your hair's on fire and recommend the assignment be given to an analyst you don't like.  Let them try and figure out how to filter across all those sources and automate the updates.

7.  Don't build Excel charts  If you are given a bunch of Excel charts and colored tables to replicate in Tableau, ignore them.  This will only lead to frustration and you may end up missing insight that the data contains, but has gone unnoticed because it hasn't been explored. So ignore these charts, explore the data, build charts that are appropriate and build a real dashboard  that helps people see the results.

6.  Don't show off  Substance trumps Sparkle. EVERY. TIME. Keep it as simple as you can. That goes for color, fonts, chart types, images, interactivity and hacks.  Hacks are great, but if you can't remember how you did it three months later, it's probably too complex. I use the 'If I get hit by a bus' rule (or for you more optimistic types, 'If I win the lottery' rule) and  think about the poor analyst who will get handed this complex work when I'm gone. Unless you know and don't like that analyst, in which case, you're just evil.

5.  Don't try to use all the data in the universe and then complain about performance. C'mon.

4.  Don't title your charts by chart type  For example, a trend chart of 12 months of sales should not be titled Line Chart: Sales. Your title should just include the subject and perhaps the time dimension (eg. Monthly Sales) or take the opportunity to use the title to communicate with the user more directly (eg. How have Sales been, blah, blah, blah?).

3.  Don't ask too much of a chart  Don't try to answer the meaning of life, the universe, and everything with one poor chart. It's why waterfall charts are often a waste of time; people ask for them, but very few seem to understand them or use them. If you've thrown the kitchen sink at a chart - for example, dual axis charts with images, color, and size indicators - you'll probably just confuse people.  Even if you can make it look cool.

2.  Don't use red and green together  It's not Christmas and it's not a traffic light we're building.  Here's a question for you: do you know if any of the senior executives are color blind?  Because if they are, and they use your dashboard, all they are seeing is a sea of baby poo.  Besides, red/green is just so hard to un-ugly; you're whole dashboard can be ruined.

1.  Don't heed any of this advice if you don't want to.  If you like what you built and it works for you and your users, then bravo. Job well done. Sincerely, and with no snark, I truly mean this. Don't let the rules or other's advice impede your creativity.