The Logic of Dashboards – Video from Think Data Thursday

As analysts, we’re often tasked with designing dashboards for others. How do we develop with our end users in mind? This TDT looks specifically at the design of dashboards and provides practical tips on how to make better, more meaningful dashboards. You’ll learn tactics to work with a variety of users with considerations of design for corporate standards, ease of usability, and stability. This session will emphasize the little ways to tweak dashboards after charts are selected by showing where color, strategic ink, and arrangement can make a difference.

(Transcript)

Ciara:

Welcome everyone to Think Data Thursday. So today we’re delighted to welcome Bridget Cogley to our webinar titled, “The logic of dashboards: Designing for the whole picture.” So Bridget has a wealth of experience with Tableau going back as far as version 6.0. I think she’s a two-time Zen master and is passionate about meaningful design and ethics and BI. Bridget is a senior consultant at Teknion and is based out of Columbus, Ohio…so it’s an area famous for its ice creams, I’m told. So you can check out more about Bridget at tableaufit.com or you can follow her on Twitter @windsCogley. So we should have time at the end of the webinar today for some Q&A; and so during the session and if you can just pop your questions into that Q&A module and at the end of the session we should have time for Bridget to answer any questions that you have. So take it away, Bridget!

Bridget:

Well, thank you very much for the introduction and I will go ahead and stop my video now, but I do want to just do a quick introduction. So thank you for the introduction one of the other things about me is my previous background as an American sign language interpreter, so that really informs a lot of how I approach this work. The other piece is I did come from more of a business side so I’ve done everything from HR, management, and training prior to moving into this work as an analyst, so I did a lot more in the soft space and more in the business space and ultimately my core focus these days is dashboard design – so figuring out how it is we build a better dashboard.

And so without further ado, I kind of want to go into a process that we may be familiar with which is where we start with a chart and we kind of look at, “okay here we’ve got this chart,” – and I’ll leave discussions about this chart and you know are there good things to do with dual axis to Steve Wexler’s TDT that was a couple weeks ago – but one thing I may notice in my exploration is, hey what’s going on with profit? Profit seems to be a little bit low.

Oh actually sorry about that. I do want to do a couple housekeeping things. So the goals of this webinar are really to start building better. And so I’m hoping that you’ve come here with the hope to build dashboards that require less explaining, less revisions and back and forth, and less kind of those small little tweaks that we end up doing. Some of the other goals I’m hoping that you get out of this are that you improve usage and have better adoption. So that’s how I define a better dashboard.

One other piece I’m going to keep hitting these big ideas and so I’ll go over these in bit more detail. So one of them is the idea of setting the tone. Another is going from large to small. Third is the ‘single question’ and I feel like we’ve probably heard this a few times before, but I’ll go over this again. And the last is beyond the chart and this is one I’ll really kind of hammer home towards the end.

So, going back into our exploration this is kind of what this process has looked like for me historically, where you kind of start with some charts, you go, “okay, well this is interesting…tell me more about this.” So maybe I’ve taken profit and done it by category, and “okay, well this is interesting. What happens if i aggregate it and compare it to figure out, is the mix a little bit different? What’s going on?” And so then maybe I’m looking at trend of sales and I may notice things, like, “okay, what’s going on in March?Why is the 17th of November really hot day? And then, “okay, well everybody loves a map,” so I’ll go ahead and put that in there.

And then I’ll start this exercise, this kind of building of the dashboard. And I feel like this is a process where we’ve been trained. You know, maybe we double click, maybe we drag certain things out, maybe we put different things in different spots, maybe we put the map somewhere, and we end up with a dashboard that’s kind of something like this. Maybe we’ll tweak it, maybe we’ll rearrange things, but we’ll spend a lot of time kind of trying to figure out how we want things arranged and looking at this analysis, moving things around. Maybe we also look at this and say, “hey, I need some KPIs. My boss is going to want this and so then we kind of throw these KPIs out here, and – voila, you know. I know my very first dashboard probably looked a bit like this, scroll bars and all.

So this is often times kind of our beginning kind of process. We start building dashboards where we’ve made these charts. We’ve done this exploration, maybe we spend a bit of time kind of adjusting things to make them work out, but then we really start kind of moving from there. “Oh, okay well I’ve done this,” maybe I’m gonna tidy it up a bit when I start learning about containers, and then, “hey I’ve learned to float legends,” so that I’m gonna start floating the legends. Maybe one of the other parts I want to start doing is making this look nicer. And so this is kind of a loop where we go through, we make the colors nicer, maybe we color-code our dual axis a bit, we try to find nicer colors, and sometimes we push things back on the map. I don’t know about you, but this is a process that I went through. This was my exact path. I made very scary critters to, “hey I made slightly better critters,” but what I wanted to understand was how do I make a better dashboard?

I got to this point but then, then what? So a big part of this journey – it was a few years ago I went through this big discovery phase of how do I build a better dashboard? Who does dashboarding well? What are dashboards really? And I came up with this idea of a language and that maybe there is something to this. So I’ve got this big long theory, but today I’m not going to discuss that. What I’m going to focus on today is something a little bit more close to home, something that you can kind of put to action right away. So what I’m saying, this talk is out there. This journey is out there. I did it for TC16, I’ve got it all over my blog, so if you want to learn more about the journey and how I came up with these ideas and these kind of four rules I’m going to discuss today, it’s out there.

So, I’m going to change lanes pretty drastically and put this to practice for a second – and so this will have importance later  – it’s one of these kind of diversions I’m taking you on, but I promise you it has value. What I’m going to do is a quick exercise, so if you will humor me and just take a minute and describe the room that I’m going to put on. Think about – you can either do this verbally, if you got a sticky note, write down a couple words, but I’m only going to give you about a minute or two and I want you to describe this room as if I wasn’t there. (Pause) I’ll give you a couple more seconds to think on it, maybe jot a couple more notes down and then I’m going to jump to one question…

And my question to you is where did you start?  When I do this presentation live, I find people start with kind of a wide variety of places. I historically have found 50% of people go to the table, maybe 35% go to the window, another 10% will start with the lights. Again, I’m looking at starting not two or three layers down, but starting. Every now and then I get somebody that starts off with the painting or somewhere else, and I’ve done this a lot with a number of different pictures and groups and I find one thing to be fairly common – and I’m hoping a couple of you will tell me I’m wrong – but what I find is very few people start off with the words, “conference room.”

I do find that this has a linguistic bias, when I do this test with native signers or speakers of other languages sometimes I find that languages actually come back with different answers, so there is a language bias, so if you are a native English speaker, I expect you to hit the table or room first. If you speak other languages as your primary language, just know that there is a linguistic bias to this, so I did find this to be an interesting experiment – and this has the use later on – so I didn’t just take you down this path for no good reason.

I promise you, it has a reason, because it goes back to these big ideas where we think about setting the tone, and telling me that this is a conference room is a way to set the tone. From there we go large to small, so sometimes people start with the window, they start with the table – these are the two biggest things – then they work their way to the light, then they work their way to the painting. So we’re already doing this idea of large to small, and ultimately the question you’re trying to answer for me is what is this room? This room is not a table, this room is not chairs, this room is a whole composition, and so that’s what I mean when I talk about moving beyond the chart. This is an exercise I use as a test. It really gets people thinking about moving dashboards in that way.

So without further ado, I’m going to break this down into the four parts. The first is when we look at setting the tone. When we look at setting the tone, some of this is the look, the feeling, the flow, but it’s also providing the context and so a part of this is I am going to revisit that dashboard. And the first thing you’re going to notice I’m going to do is, instead of hitting tiled, I am going to hit floating. I’m also going to delete my reminder – this is there for me. I’m gonna come and I’m going to do a live build. Part of this is me thinking of how do I want to put these things out there and looking at when I build this composition and think about how my end-users are going to use this. I am setting a tone. I am trying to make a composition for others that they can understand immediately, and so you’ll kind of see me kind of play around with sizing a little bit, but then I’m going to throw on these other pieces and parts. I do think that the difference of floating and tiled – it’s – I think it matters. I think it affects how you analyze the data. I know for me when I’ve got things moving around, it’s very hard to see.

I’m gonna start here. I’m gonna kind of maybe put this guy up here on the top, because I feel like that’s a good spot. These KPIs I did intentionally make them light, because I’m gonna put them up here and I feel like this starts to give my work a bit more of a branded feel. It feels a little bit, maybe, like the corporation I work for. I’ll kind of move these guys up here, I’ll fuss with them a little bit more later, but part of this is looking at the analysis and beginning to understand. What is it I want to convey to my end-users? What is it I want them to do? So, you’ll see I move these pieces around and I just think about what am I trying to convey. Maybe the legend goes away. I’m gonna pull this piece out that’s hidden, and then maybe I’m gonna realize, “oh hey, I need one more piece,” because something seems a mess. You’ll see I’ll pull this other legend out, so rather than having a legend, I’m going to use a chart as a legend today, and so I’ll pull this piece out there. I’ll keep doing a little bit more work on this, and moving things around and kind of figuring out, how do I want this to flow? How do I want my users to interact with this?

We are going to do a little bit of kitchen magic here, in that I’m going to go ahead and hop over to the final version of this to where you can see ultimately where I’ve ended up. You’ll notice a few things and – setting the tone – I’ve added some “chart art,” if you will. I put in lines that really weren’t there. I’ve helped organize my KPI, so this just sets the tone here. I’ve really pushed these colors back. One of the things that most people don’t notice, but that’s very important to me because it does start to reinforce these relationships, is I’ve got this blue for sales. So, wherever I’ve used the sales metric, that item is blue, and then with profit, I’ve got this teal. I’ve kept things in kind of this teal and green family when it relates to profit. Lastly, I’ve got quantity, which is just totally different and I’ve went ahead and made it its own separate color. Just by setting the tone and playing with color very strategically, I help people navigate this without them even realizing it.

I’ve also added in some icons to help show people where to click. One of the other new features I’ve done is I went ahead and said, “hey here’s this lasso to select,” instead of having a drop down filter to pick a period of time. What people can do instead is, by taking some practices from d3, select this. You’ll notice the entire dashboard changes, and that’s done with a set action. That way you can actually see this last year and then, in the final version of this, I did take some time and put in some coding in the tool tips to help people understand what’s filtered and what’s not. So just by providing cues, pushing the color back, and really focusing on the image I want to project, I’ve really changed how this dashboard is understood.

 The next point I want to hit is this idea of large to small and any English majors know exactly what I’m talking about, where you’re building a main idea and then you’re putting in supporting items and then you get into the detail. I also call this approach faceting. I’ll get into what I mean by that.

This is an example of an exploration I’ve done and part of this is to support the idea of where do I want to work today? In looking at where do I want to work today, I can see that my master bedroom has a lot of change and it’s fairly volatile from a temperature standpoint. This is real data by the way, so if you look through here and you go, “man, your house is cold!” Yes, yes, it is. One part of what I’m trying to understand is where do I want to work? This is a great exploration for me, as the analyst. It’s not a good exploration to answer the question I have, which is where should I work?

What I’ve done in this dashboard is I’ve started very large to small. My high-level idea is what room is most comfortable? By starting out, I understand my master bedroom is actually quite comfortable. When I get to where I should be working, which is my office, I find it’s not nearly as comfortable, just that 5 degrees Fahrenheit makes a difference. What I can see is, when I click on my master bedroom, I’m gonna see how volatile it is per day. As I scan over, I can actually filter by day to really get exact feel for what that day is doing – this also helps support how I build actions – versus let’s say I look at my office. You can see over time it’s getting colder. You can also see that evening, not overnight, seems to be the bigger problem. I can look at that, I can see that there’s a lot more variation in that. Some of that comes from when we put the sensors down there, but I can also see that I’ve outlined my comfort zone. I did use this idea of that chart and I also made it so I could hover and work with these bits and pieces. You see that this faceting logic really supports this drill down to the detail.

Moving on to trying to answer a single question…I feel like sometimes when we’re designing these dashboards, we get hit with this idea of, “we must answer everything for everybody.” I really tend to push against that. I really push for having a single question, trying to do these kind of short answers, then expanding, and then doing sometimes what certain people call like the “five why’s” or “what next?”

My goal in doing this is clarity. I did this exercise for mobile, and this is extraordinarily stripped down. I show this sometimes and people tell me, “oh I could never put something this minimal in production.” This is a great way to start and, so one of the things when I’m designing is, I start either in gray or in a single color to really focus with laser-sharp precision on what it is I’m trying to answer. My question with this is simply, how is shipping doing? What should I know about shipping? Part of when I look at shipping, I’m looking at how many items do they send out? What product count did that represent? How am I? How quickly are these things shipping out? And am I discounting potentially because of shipping?

Starting with that high to low, and answering the single question I’m looking at, where is my standard transit time falling? The other part is does it differ from same-day, first class, second class, third class, and then lastly getting into more than nitty-gritty details of each individual item – just figuring out do certain items get sent a certain way? One of the things I’ve also put on this it’s not shown just yet, but will be in a second, is I put set action. What I wanted this to be was a more comparative dashboard and, part of the reason it’s designed the way it is, is because it’s optimized for mobile. I wanted to get this to be, so that if I put this in a warehouse or if I deploy this live, that somebody within a warehouse can use their phone and very quickly explore. This was also a test of the new future, so I went ahead and I left it on this new automatic layout to just see how could I design, knowing that it would automatically convert for me. This was a test of that as well, so there was some deliberate design choices made just to accommodate that, but what helped me to look at with this analysis and – again keeping it very minimal, keeping it very simple – was just trying to figure out where things fall.

In the final version you’ll notice that I actually changed this line from being a very literal transit time to being a difference. I found that what was more effective for me for this analysis was not so much the literal times, but I wanted to see how much variance I had. We’ll see that in a few more moments, but I just wanted to push this idea of thinking about this more from a standpoint of answering a single question, of keeping things pushed back, versus really trying to put everything on there. You get a little bit of that. You can add in some features, and you can see how you can get this information fairly quickly. You’ll also notice that the form factor really limited how much I put on there. Of course we’ve got drop down filtering as well.

The final piece that I really want to push towards is this idea of moving beyond the chart. This is where we explore, we build, but then we go back and we do this refine cycle. This is probably the point I think when I do teaching or, when I work with people, I feel like that they tend struggle with the most. Some of this is when I start building my explorations and thinking about how do I want to convey things, I’m piecing and parting things into a UI, so I’m setting the tone there for people to understand this is what I want to convey. Now I’m going to throw things out here: I’ve got these products, so that’s gonna be a smaller, more detailed item, and then maybe I want some trend lines, and so I’ll look at this and say, “okay well I’ve got this.” And then maybe I’m curious, maybe I want to know more about this particular, and so maybe I’m putting these category parts in where I want to kind of understand the relationship. So I’ve got these items.

You’ll notice I’ve kind of cheated a little bit and there’s a reason. I’ll play with these to get these labeled appropriately and move things around and really look at building an integrated piece. I want people to find this and feel like that it’s fallen – that they’ve kind of fallen into a world. I’m a big fan of certain photographs, I look at photography a lot, I look at movies a lot for that – do they build a world I’m interested in? Do they build things that I want to see? You can see now I’m starting to play with this a little bit. I may even move these pieces on top of each other just to start seeing, can I build something interesting that helps people understand what relates and what doesn’t? You’ll see I’ll play with this and I’ll finagle with this a fair back to really make this relationship clear. That’s a bit of thinking beyond the chart: looking at how do I make interesting things? I seem to be missing a chart but I’m going to move this guy up here and play around with this, but you see where I’m starting to make interesting things and to support this analysis.

The final version of this then starts to look more like this, where I really lined these pieces up and again this is one big reason why I personally tend to choose to design with floating. I find it makes it easier for me to make these kinds of connections to where, if I select California, I can see there there’s the shift where office supplies really makes up the difference in California. Maybe I want to push that more, maybe I’m finding my technology isn’t selling as much. If I look at a place like Ohio, where I’m from, I can see that SuperStore is probably going to move their brick-and-mortar out of my area unfortunately. That’s what this is helping me explore. From there, I may also start looking at what products need to stay or which ones get to go, but it helps me inform that analysis. I made some UI and design, and it looks like a cohesive piece.

The other thing is again I continuously keep pushing back those colors. If I take that same idea to that mobile dashboard and focus a little bit more on user experience and setting the tone and going beyond just the charts, you’ll notice I’ve added again more decoration. I’ve put in little dividers, I’ve put in little dots, to help people navigate. I may play with these and move things around, just to get them to look a certain way. That way when I take it to mobile again, I’m gonna kind of take those thoughts and considerations around experience. I may remove ink and – again think about the navigation piece – so moving beyond the chart, I’m also going to think about the UI background. If I’m using this on the phone, I don’t know about you, I’ve got some really fat fingers, so putting safe landing zones for mobile as well it’s kind of a good consideration too.  I’m thinking about with this particular chart, this is where people are going to click. I’m looking at where do I want them to look? Where do I want them to land? Making it so that they don’t have to scroll as much so these are all kind of different pieces.

One of the other experiments that I do with this is, and this one’s probably my more extreme example, but really figuring out what could dashboards look like in the future. I don’t know about you but when I watch sci-fi movies, I was wondering why is it black? Why is it when they design on glass, why is everything in the middle?  I have these questions, and so I seek to answer them through doing dashboards. They seek to say if I’ve got this HR data set and I want to imagine this HR data set in the future, what could it look like? So this is my little homage to Giorgio Moroder. One of the stories about Giorgio is that he knew he needed to click, so when he was making music and he wanted to design the sound of the ‘50s and the ‘60s and kind of push into the future, what could that sound like? He knew he needed something to help create that sound.

When I designed this dashboard, I knew everything ended up in the middle. I feel like circles galore happen in all these futuristic dashboards, so I said if I’m gonna do this, I need circles. You’ve got a donut chart, but then I needed a way to help people navigate, so I just put in various data artifacts. These little lines mean nothing, other than to anchor you over to this timeline. You’ll notice that I’ve got turnover (for HR). The labels are in weird spots. Again when I look at these futuristic dashboards, when I look at sci-fi movies, I feel like they look at this. I really wanted to understand why. I went ahead and altered some of these bar charts and then I did some survey responses, so how do people feel about their work-life balance? How do people feel about job satisfaction? I’ve got those color-coded based on the turnover, so that you can see people who turnover tend to be less satisfied, tend to be less happy with the work-life balance, and aren’t happy with the environment. I want to see does travel affect this? If I’m looking at people who travel rarely versus travel frequently. Non-travel surprisingly was the winner on dissatisfaction, so maybe people need a little bit of travel.

I wanted to play again with set actions and I started looking at degrees. Do certain people makeup a major part of our composition? Building this out and playing with this (chart), it’s a very odd shape and the lines are there solely to reinforce the relationship. This is playing with data art and playing with abstraction, because if you look at a lot of these futuristic dashboards, there is this abstraction. I went ahead and made a manual donut chart, so this is hacking with math, a lot of circle calculations, and just trying to make it so that I didn’t have two donut charts overlapping in the same spot. I wanted it in a weird spot.

You’ll notice there’s a ton of coloring this and when I looked at futuristic dashboards, I looked at works like Tron. I looked at the colors that they use, and so this yellow is really an ugly color on white. You can actually see it on the tooltips. It’s a terrible, terrible color when you put it on white, but it looks awesome on a black background. I’ve got a lot of noise with these white lines and that helps just kind of line things up, but puts a sense of balance and just makes it look a certain way. I’ve got a lot of translucency.  The population pyramid is really, really thin. it’s got a lot of color I’ve got these these bar charts behind it to help reinforce the relationship. I have a lot more male employees than female employees. Because I had that gap there, I was able to extend this chart in that space, so it’s not a very boxed in dashboard.

Now, people say, “well I could never get away with this at work,” and I agree. I could never get away with this at work either and so I decided to make a normal-ish version. Again, this takes into mind some of those practices of what are my big questions? When I look at this and kind of figure out what am I trying to answer when we deal with HR data, I feel like the first thing I end up having to answer is what’s my attrition? What’s going on with my employees? Are they staying or are they going? I can see the ones who quit called out with the brighter color. I’m looking at traveling frequently. Again, this is another area where I went ahead and just put in a set actions. I mean it’s still very modern. It’s still something that’s probably a little too edgy to get away with in the work environment, but it was something that I did as an experiment. You can see by using a set actions and playing with subtlety, being able to call out these particular things without it being too noisy – again pushing for more abstraction – looking at things like degrees and doing this composition and encouraging people to click and play, this is another way to draw people into our dashboards and get them thinking perhaps a little bit differently than they have historically while making a nice gentle balance on just color and what we’re trying to convey. You can see my population pyramid is here, you can see I knocked it down it takes up less space, and that I’m no longer moving into its space.

The final piece that I did with this puzzle was I know why futuristic dashboards are black now, because I did what I would call a corporate redact. Most corporations I know, they prefer lighter colors and they push you to use certain color guides, and so I went ahead and put this same type of futuristic dashboard on white. I feel like it lost some of the magic when I did this. Everything’s the same except for a couple of pieces. I had to push this red to orange. It just felt way too harsh and overbearing on the lighter dashboard. You can see just with this corporation redact I had to make noticeable changes again. I left this yellow as is. I probably should change this, but there’s something lost and then you’ll notice too, that I knocked out the color on this (population pyraind).  It started to get too noisy with all the colors on that futuristic style. I left these guys separate, in part so I could kind of go ahead and do this, and let you see these guys side by side to compare as well. It was just it was an interesting story to me was really fascinating to see these two side-by-side together to understand what is futuristic?

When I set the tone, when I put things on black, when I play with color, and I make these very branded looking items, how color matters and how the background makes a difference just the chart choices and how I place things. This is one big reason that I personally tend to go with floating over tiled pretty much any day of the week, because it does give me that control. I really wanted to kind of put these out there and let you see the differences with the design and the thought process. With that I’ll go ahead and wrap up and start moving towards questions.

Again I want to reinforce these big ideas, of the idea of setting the tone, of how is it that we think about when we walk into a room. How do we tell people what this room is, and you know same thing with a dashboard. As soon as people see it, they make a two second evaluation and I want that evaluation to be as successful as it can be when I’m designing. I’m not necessarily making the pieces larger but what I’m doing is I’m trying to go from a high-level idea, so whether I’m thinking about employee retention with HR data or I’m thinking about setting the tone with retail data, and saying here are my sales and profit. Now I’m going to start giving you details into this analysis. I focus my work with a single question and sometimes I will very literally even type that question up there, especially if I’m struggling. Sometimes I take that single question to a whiteboard, and I’m talking it out with sticky notes and figuring out what are they asking, what is it that they want to know, so that they can then take this to action? I really care a lot about moving towards momentum. The final piece is not just designing these individual charts and pieces – and again this is probably why I pushed the hardest against tiles and against a lot of data ink that separates my charts – I want people to see this as an integrated piece. That way when I’m looking at things like a conference room I’m not just seeing a window or a table, but I’m seeing a whole room and ideally that room is inviting.

It’s one of the reasons why I love to go to open houses because I like to see how people draw me in. I like to see what tricks they’re using. Are they using color? Are they using different little tiny features so a lot of times I put in lines or I put I group certain things together to tell people these pieces are meaningful together, so it’s not just putting in charts on a dashboard, but it’s also looking at this as a whole. These are the big ideas I would love to leave you with and when I find is that they all flow together. As soon as I set the tone, that really helps me move beyond the chart. When I start working large to small, usually I’m doing so in the pursuit of a single question. So, you’ll see that these guys really do fold into each other and almost become one at the end. So that’s me and now I will go ahead and move to questions because I’m sure you’ve got bunches.

(Q&A coming soon)