1 2 3 Previous Next 44 Replies Latest reply on Aug 21, 2015 6:18 AM by Toby Erkson

    Let's talk about dashboard design!

    Rody Zakovich

      Hello Community!

       

      I love the Tableau Forum (Really the community in general). It is a great place for individuals to ask specific questions about "How do I do this in Tableau?", but we rarely discuss Information Design, and the importance of portraying data in way that tells a story.

       

      So I wanted to try and start a discussion that is a little bit different, but still beneficial for the community (I have seen similar posts, but a lot of them don't have much feedback).

       

      When working with Tableau, what are some of the questions you ask, or steps you take during the design phase of your visualization?

       

      Are there any individuals you study or best practices you try to follow?

       

      What is your favorite visualization of information? Why does it have so much impact on you?

       

      It would be great if we can get a lot of diverse opinions. I know there are a lot of blogs and articles on the topic, but I wanted the communities thoughts. Those of us who are down and dirty in the data everyday!

       

      Simon Runc , Mark Fraser , Jonathan Drummey, kettan, Shawn Wallwork, Pooja Gandhi, Tableau kumar

       

      I really hope we can get a good discussion going!

       

      Best Regards,

      Rody

        • 1. Re: Let's talk about dashboard design!
          Rody Zakovich

          Bump? Any takers?

           

          Just want to try to start a friendly open discussion

           

          Regards,

          Rody

          1 of 1 people found this helpful
          • 2. Re: Let's talk about dashboard design!
            Shawn Wallwork

            Rody when I have time later, I'll play. I'm thinking I should move this over to Viz Talk so it won't get lost in the forum flood. You OK with that?

             

            --Shawn

            • 3. Re: Let's talk about dashboard design!
              Rody Zakovich

              Hey Shawn,

               

              Absolutely! Anything to get the discussion going.

               

              Thank you!

               

              Rody

              • 4. Re: Let's talk about dashboard design!
                Shawn Wallwork

                It's moved. That should stay on the front page for about a month, so you won't need to bump it.

                 

                Cheers,

                 

                --Shawn

                2 of 2 people found this helpful
                • 5. Re: Let's talk about dashboard design!
                  Rody Zakovich

                  Thank you sir!

                   

                  Rody

                  • 6. Re: Let's talk about dashboard design!
                    Simon Runc

                    hi Rody,

                     

                    This is an excellent idea...In all my time on the community I've only even had a handful of Viz questions (eg. Best way to represent measure sums and their % / need help visualizing), and this is my fave subject!!

                     

                    I am a data visualizer, and Tableau is just the tool I use to gain the understanding (and communicate that understanding) of the data. Although I must admit to loving Tableau (and the logical challenges data and Tableau bring up), this is an 'aside' and wouldn't classify myself as a Tableau expert, anymore than a graphic-designer would classify themselves as an 'in-design' expert or an artist a 'paint brush' expert!! This is just the best tool to get to my goal of understanding how everything works!!

                     

                    Nearly all the data I work with is from the retail industry (although this can take many forms, from PoS data, finance reporting, space planning, location planning, supply chain processes, to customer loyalty data...), so my blurb will have this bias (but think it applies to most data problems).

                     

                    My first rule of thumb is to keep things simple, both in thought and visually. When I get a new data-set/project the first thing I do to get a 'feel' for the data...in retail, assuming I have no loyalty data, there are 2 main factors; What you sell (the products), and where you sell them (the Stores). So I tend to use a combination of low-level bar charts, Pareto curves and histograms to get an understanding of the shape of a business, such as 'how big is the 'tail' of products?', 'what is the average spend, and how much dos this vary by store?' (this bit is not that exciting!), but is necessary for the next (much more exciting) part...Informed by my general understanding of the business I can then start asking questions! This is the key to keeping it simple, I ask 'simple' questions (hypothesis) [generally one at a time] and build the Viz to confirm, or reject my idea. If I've asked the 'right' question, and built the 'right' Viz this should lead to the next question.

                     

                    This leads me to the next part of the discussion on Viz techniques...In simple terms I see data visualization as performing 2 main tasks. 1. Anomaly Spotting 2. Pattern Spotting. As such my tip is to build your Viz with very 'light' colour pallets, this has 2 main benefits (here's a link to some beautiful viz's using 'light' pallets Health Intelligence | Analyzing health data, generating and communicating evidence to improve population health). Firstly when you want to show something (such as an anomaly, pattern...etc.) you can employ a 'dark' color which will really highlight your point, and visually call out any anomaly or pattern. I also find this much easier on the eye (not just aesthetically, but less effort - a bit like a boldly painted room can give you a headache after a while!!).

                     

                    In terms of Viz hero's, it's Stephen Few for me! Mainly his insights on the topic of how we (humans) see (and process) light and shape, more than a 'strict' following of this type of data must use this chart type. Although I do defer to his better knowledge and I'm sure he'd win an argument on why he is right!...I like to go a bit more 'off-piste' (which is where Tableau is so good!). As an example, the below Viz (done by Bethany Lyons at last years Tableau Partner Conference, as part of her 'Moving beyond Show Me' presentation) demonstrates the point well

                     

                    American Election Viz.JPG

                     

                    This point of this Viz to give an insight into the 'evolution' of the american elections over the last 3 cycles. Each line is state, and the Y-axis is the %age from parity each state voted either Democrat or Republican over the last 3 elections. The genius here is that she's joined the 3 points, and then sized them by Year (which sounds very odd!), until you start to see that where you have an 'ice cream cone' shape the electorate has moved more democrat (as the year (used as an integer) is larger, the swing to democrat is greater), and the upside down version is the reverse. Once you get your eye in, you see this Viz really demonstrates the 'polarization' of US politics over the last 3 cycles, with democrat states becoming (generally) more democrat, and republican states becoming (generally) more republican. There is also a slight pattern that very democrat states became even more democrat, and vise-versa (i.e the further the lines from the center the longer they are). btw the var chart in the top shows how many states had their biggest 'swing' (so the Obama effect if very visible in 2008).

                    This made possibly by the way Tableau builds all Viz's from a 'consistent' pallet of Marks, Sizes, Colurs...etc. In traditional BI, you'd need the developers to create you a 'special' chart type (I wouldn't even know what to call this one!).

                     

                    I'll leave it there for now (I could go on all day!!)

                     

                    Would love to hear everyone's thoughts on their Viz ethos! and expand my knowledge from your insights and experiences

                     

                    Nice discussion Rody!

                    3 of 3 people found this helpful
                    • 7. Re: Let's talk about dashboard design!
                      Harley Ellenberger

                      My job is almost entirely focused on how to make sense of our internal data.  So every day I'm trying to come up with meaningful ways of presenting that data.  This usually results in some form of an interactive dashboard in Tableau - and I've built dozens of them.

                       

                      I refer to it more as Business Intelligence and I think it's a bit different from some of the more creative work - or Data Art.  I think Business Intelligence should be focused more on telling an insightful story to an audience using some of the more traditional chart types - bars, lines, scatter. 

                       

                      In my own design process, I try to think of the one question I'm trying to answer and I build around that.  I try to follow the 'Overview, Filter, Details on Demand' approach.  And I try to abide by all of the data visualization best practices, primarily data-ink ratio.

                       

                      I think Kelly Martin has a very practical and simple approach to dashboard design that applies very well to Business Intelligence.  And Ramon Martinez has some excellent examples as well.

                      2 of 2 people found this helpful
                      • 8. Re: Let's talk about dashboard design!
                        Rody Zakovich

                        Hi Simon, always good to hear from you!

                         

                        I really enjoyed your take on the subject. Like you, I feel that Tableau is merely the tool I use to visual information, but my main job is to present data in an insightful and understandable way.

                         

                        Generally speaking, the first steps I take when building a visualization is identifying, What questions are being asked and  who is the intended audience. These two can be interchangeable, as far as order, depending on the situation. This leads into the next step, which is always grueling, but basically it is taking a high level question, such as, How is Revenue Trending, and breaking it up into smaller, fine grained questions that match what the audience is indenting to learn. This way all stakeholders have a clear understanding of the reports purpose.

                         

                        After that is identifying how the report intends to be used, and who is responsible for taking action, based on what the viz tells us. This is very high level, but it is generally how I like to go about it.

                         

                        Finally I get to have the real fun! Which is figuring out how to design the dashboard/worksheet/story, etc. to answer these questions and make an impact. This includes using colors, shapes and even Borders, to draw attention to key areas.

                         

                        As far as "Best Practices", if there are 10,000 situations, only 2 of them call for a Pie Chart....but that leads into who I study.........

                         

                        I'm a big Stephen Few guy, he really takes science and applies it to business practice. The first book I ever read on design was "Information Dashboard Design". As such, I am also a big fan of Edward Tufte, and his militaristic approach to visualization. These two legends have really influenced me. BUT, I always felt emotionally disconnected from the styles of Few and Tufte. Though most of the time their wisdom and knowledge overrides my gut feelings, there are times I look to Nigel Holmes for the "Emotional" aspect of information design. But at the end of the day I think it all comes down to the audience.

                         

                        My favorite Viz was Nigel Holmes's Diamonds Were a Girls Best Friend

                         

                        Nigel Holmes.jpg

                         

                        Is this something I would make for a High Level executive of a Diamond company, not in the slightest chance. But Nigel showed us here that the emotional aspect of Visualization is important (Given the right context). I have looked at a lot of graphs, dashboards and reports. A lot of them fade from memory after a while, but this one never has, it has always stuck with me. Though I may not go to this extreme, I always try to make an emotional impact. I know that if I can reach the user on a Human level, they will be more likely to remember it. That being said, it is a very unique skill to perfectly balance the "Eye Candy" from what is most important (The data). 

                         

                        I may be different than most here, but I love being told I'm wrong, because it means I have the opportunity to learn. So if I have said anything that you think is incorrect, or could be handled better, please let me know! I will take it with open arms!

                         

                        Best regards,

                        Rody

                        2 of 2 people found this helpful
                        • 9. Re: Let's talk about dashboard design!
                          Rody Zakovich

                          By the way, as funny as this sounds, this was the First Question I have ever asked on the Tableau Forum! Ironic that it was moved over to Viz Talk, but I'm glad it did, cause it got the conversation going. Thanks again Shawn Wallwork!

                           

                          Rody

                          • 10. Re: Let's talk about dashboard design!
                            Simon Runc

                            Yes like you I love being wrong too!! (...fortunately it happens a lot!...so lots of opportunities for self-improvement!!).

                             

                            We do a mixture of consultancy projects and managed services, and I've approached the topic with my consultancy hat, where I'm creating the visualization for myself to provide insights to the Client (which is where I'm happy creating 'off-piste' Viz's for my own investigation of the data). I would, unfortunately, never be able to provide the kind of I showed in my original post to self-service client!. On this side of things, yes consideration of the end-user is paramount (their level of data/analytics experience, their job, what they want to get out of the models...etc...Although I must admit to constantly having to remind myself of this! in this regard I try to keep to a simple rule 'If I have to spend more than a minute explaining what you are looking at I've failed'...When I pose this question to myself this often kicks out the more 'extravagant' Viz's!! ('extravagant' is the kind word for them!!)

                             

                            Totally agree of Few and Tufte it's more the science of shape/colour/Visual-cortex...etc. that I adhere to, rather than a strict rule-set on which chart type (I think the Viz from Bethany above follows this, but obviously uses a chart type they wouldn't probably not 'exclusively' recommend).

                             

                            Another, non-visual, area of consideration for me is the study of Heuristic/Bias' (My fave book on this is 'Thinking, Fast and Slow' by Daniel Kahneman). As well as trying to avoid the pit-falls myself (especially the 'confirmation bias'), being aware of this for my end users also influences my design. In retail (and I imagine most industries) there are lots of 'myths' which build up over time (even though no-one has any idea where they came from, and what information they were originally based), and data visualization is very good at 'dispelling' these myths, so I ensure that the 'myth-busting' Viz is shaped/colured in such a way that the new insight (replacing the Myth) is very clear.

                             

                            ...just having a look at Nigel Home's site...already a fan just from the menu selector!! (you can tell straight away when someone knows their onions!!)

                            2 of 2 people found this helpful
                            • 11. Re: Let's talk about dashboard design!
                              Rody Zakovich

                              Ahhh Bias,  or the thing we wish to get rid of by analyzing data!

                               

                              This is always a tough one to deal with information design, especially when your stakeholders have a pre-defined notion of the answers. It took me a while to break people of the idea that 3D charts are a good idea. Or that maybe one of our products wasn't as "Successful" as some might have thought.

                               

                              It was one of my first designs (In the first month of being an analyst), that I realized the Galileo effect. If you design a dashboard that doesn't use Pie Charts, and disproves their Cognitive Bias, they won't always respond with the kindness, but it is up to us to keep fighting the good fight!

                               

                              Regards,

                              Rody

                              • 12. Re: Let's talk about dashboard design!
                                Shawn Wallwork

                                You're welcome. I find it interesting how this worked out. I don't 'follow' the forums (too much info), but I do follow Viz Talk some groups and individuals. I'm guessing others are like me. So putting it in Viz Talk got it emailed to this groups followers, which is why you got the response. Good to know.

                                 

                                Cheers,

                                 

                                --Shawn

                                • 13. Re: Let's talk about dashboard design!
                                  Rody Zakovich

                                  Well I appreciate it Shawn Wallwork!

                                   

                                  Also, Simon Runc, I am not trying to hop on the Homes's bandwagon, and say he has all the answers (My Few and Tufte fandom won't let me dive that deep), but here is one of my favorite short videos.

                                   

                                  Why so serious? | Nigel Holmes Explanation Graphics

                                   

                                  Personally, it is hard for me not to see his points. But then again it's all one's opinion/perspective.

                                   

                                  Rody

                                  • 14. Re: Let's talk about dashboard design!
                                    Nicole Edmonds

                                    Good thread!  I'm relatively new to Tableau, but I've certainly caught the bug as it were.  I come from a hardcore traditional BI background as far as my career, and spent more time in the wrangling/cleansing side of things vs. the design side, until recently. (Other than your super basic dashboards in Excel/SSRS)  A couple years ago, I was at the SQL PASS Analytics conference, and one of the keynotes was David McCandless.  His presentation really shifted my view toward the beauty of data, and the creativity that comes with exploring data in a visually meaningful way.  

                                     

                                    As far as the questions I ask - I generally start with - Can the user look at the viz, and be able to explain what they are seeing easily?  Is the information that is being built into the viz valid to the business question being asked?  (for my own edification here)  Is the interactivity intuitive for the user, and can they get to the information they need without too many clicks, etc.? 

                                     

                                    I will say that the blogs, forum posts, etc., from the people you tagged above have been an incredible resource for me.  Aside from Few/Tufte, I find Andy Kirk's Visualizing Data site, and Andy Kriebel's blogs to be very useful, and go back to them often. 

                                     

                                    Rody Zakovich - To your point, I recently put together what I thought was an awesome dashboard, only to have my users say, "wow, that looks great, can I get that in Excel?"  

                                     

                                    As far as my favorite viz's- This one is a front-runner:   http://guns.periscopic.com/?year=2013 

                                    3 of 3 people found this helpful
                                    1 2 3 Previous Next