Talk Data to Me: Data Visualization Best Practices

hi everyone thanks for joining today's webinar visualization and desk dashboards design best practices this webinar is part of our webinar series top dated to me which is aimed to provide tips tricks and how-to to make you tableau savvy my name is Megan say and I'm part of the marketing team here at tableau I'll be your moderator also helping out on the call today are Hugh neuen and Patrick van der Heide from the tableau support team just a few quick housekeeping items before we get started all lines are currently muted if you have a question please use the Q&A window and we'll get two questions at the end if time allows a recording of this webinar and a copy of the slide deck and workbook will be emailed out after the webinar and with that I'd like to introduce our presenter today John Duggar John has spent the last 15 years working with big data with the more recent years as a team lead for global channel pre-sales team here at tableau he is a purist and practice artisan in the art of visual best practices and is an expert in marrying the story of data and explaining how the human brain works in this session we'll be discussing visual best practices you can use when designing your interactive dashboards in tableau John take it away Thank You Megan hey let's go ahead and get started pretty quickly I have a lot of content that I would like to walk through in a very short amount of time to cover it so with that we'll jump right in and begin our conversation on visualization and dashboard best practices but first we'll have a little blurb from legal but presentation is not complete without some words from our attorneys this presentation could contain some proprietary information and keep in mind you could be under a nondisclosure agreement or other agreements within tableau so keep that in mind if you do get a copy of this presentation please consider keeping that inside of the partner ecosystem all right that being said let's go through the agenda real quick what is a dashboard that'll be the first thing that we cover you may not have even thought about that you may think that a dashboard is just a lot of graphs or visualizations on a single screen in some cases that is correct but it's much deeper than that so we'll get into that definition of a dashboard will categorize dashboards as well there's three main flavors of dashboards we'll go through each one of those we'll talk about the common mistakes and kind of illuminate those in the tableau story point story and we'll talk about the principles of visual perception vision is our most powerful sense so we're going to learn today how to take advantage of visual perception to design dashboards that are easy to digest and create rapid perception and introduce outliers in a way that's very clean and clear we'll also critique some very bad designs you'll actually put some of this new knowledge to use and we'll show some very ugly dashboards and we will show you exactly why those are ugly you may have thought they were beautiful at one point but after this presentation I think you'll see that they all have mistakes that's a very end I have some recommended reading for each of you if you want to get good at creating dashboards you have to put in the time so it's practice as well as reading and learning from the experts all right so what is a dashboard rather than reinventing the wheel and creating my own definition I thought I would just steal one from Stephen view he is the thought leader on dashboard design best practices so there's some important notes to take away from this particular definition that's the dashboards of visual displays they do include text but the emphasis is going to be on graphics because graphs communicate data very very efficiently and we can take it in quickly and our brains can interpret the most important meaning from the data using graphs another important point in this definition is the dashboards display only the information that is needed to achieve specific objectives so dashboards should be objective oriented three to five objectives if you don't keep a dashboard or a set of dashboards tied to specific objectives they end up what I like to call boiling the ocean right so they're very confusing there's a lot of stuff going on lots and lots of different questions being answered in many many different ways but at the end it's a very confusing dashboard or set of dashboards so keep them tied to objectives dashboards should be you should be able to monitor an effective dashboard at a glance so really anything can be displayed on the dashboard but one characteristic that's common amongst all dashboards it is that they show abbreviated summaries of data both large and small it's very difficult to monitor all the detail needed to achieve your objectives but a dashboard must be able to quickly point out that something needs your attention or your action so really to sum it up at a glance a dashboard should be able to tell you that you should act on something rising dashboards is really three main categories one is strategic and these are typically your your high-level summaries of the most important information and typically these are static snapshots of data this would be something that an executive would want to look at or a VP would want to look at each morning they get to their office we want to take a quick look at their strategic dashboard and get a finger on the pulse very quickly operational dashboards are immediate by nature they're used to monitor activities that are constantly changing such as sales or profits or maybe a computer farm with memory and CPU and other statistics around those Hardware blasts is certainly not least is your analytical dashboards these offer rich comparisons drill down capabilities what-if scenarios these are 80% of the time this is the top of dashboard that you're going to create and with tableau software analytical dashboards are highly interactive let's take a look at one real quick this is the strategic dashboard that contains only the most important information for maybe a sales manager to monitor and if we look at the visual highlights of this dashboard notice that color is used sparingly and that the prime real estate has used for the most important data which is in the upper left hand corner which is our key metrics year-to-date anywhere you see the color red it means something is bad so if you look at this dashboard very quickly and rapidly you can see that the areas that are problems for your company and that's the key with a strategic cash flow the next example is an operational dashboard so this dashboard allows you to quickly see daily cells and profit activities you can see the top products today's orders in the last seven days cells you can also see performance metrics by region as well as cells or in profit over time and remember the goal of an operational dashboard is to immediately monitor activities and in this case it's our sales and profit activities that leads us into our analytic dashboard so in this particular case it could be a I don't know an analyst whose work supports the marketing effort so the companies hotels so maybe this analyst is booking is monitoring booking trends to identify seasonality and maybe opportunities to interest customers during slow periods so once again in this dashboard and what you'll see common amongst all great dashboards is that color is used very effectively and the data is organized on the screen in a very clear and concise way so the colors use sparingly on this dashboard as well so if you look at the color orange is telling you something is not so great we're looking at total revenue and in many different ways of it over time and a line chart we're also looking at rooms sold versus available in a bullet chart in the middle of the screen and then in a heat map at the bottom we're looking at that revenue seasonality so over the 12-month period January through December what does our revenue look like and you can see very quickly that in September through out through November revenue is not so great in fact it's really low so it leads you to think why why is revenue low if you look at the regions in this dashboard you'll see is Mexico the Dominican Jamaica the Caribbean so it's hurricane season and people don't tend to visit these places during hurricane season so maybe the marketing analyst could create some kind of a work with the the campaign people and create a campaign it would attract customers during these slow period so that our revenue doesn't take a huge hit during these slow months move on to the next screen all right so let's let's talk about some common mistakes we see a lot of mistakes in dashboard design and we're just going to illuminate some of the top mistakes that we see using a tableau story points so in this case we're fragmenting information which chapters would be mistake number one this is a very common thing where in this case we're fragmenting in a couple of different ways we have a scroll bar on the right side remember the scroll bars are the enemy and dashboard design you do not want scroll bars if you can at all help it they're very distracting and it really fragments the information we also have a scroll bar down here on the profit by category bar chart and in another way that we're fragmenting the information in this particular dashboard is that we're having to choose a metric using this parameter driven drop-down which is okay but maybe you want to be able to compare all of your important metrics on a single screen and the way this dashboard is designed you can't write you have to pick one metric at a time and panelize that information on this map and on the trendline so a potential solution to this problem is arranging all of your data on a single screen with no scroll bars and then presenting your most important metrics on a small multiple display which is essentially the same graph presented multiple times but is broken up by different dimensions in this case its regions so very quickly we can see those trends across all of our most important metrics and there's no scroll bars on the screen that will cause fragmentation of the data mistake number two is inadequate context and by inadequate context I have I have a some screenshots of gauges and now of course gauges don't provide great context they're really bulky they take up a lot of space on a graph and tableau makes it very very difficult to create gauges we definitely don't do it out of the box because it is not a best practice in terms of visualization some people like them but do avoid them at all cost so let's look at some solutions to inadequate context I'm going to offer you out three different solutions so you can use these in your dashboard design if you're not familiar with bullet graphs I recommend that you get familiar with them because it is absolutely the best radical display to compare two measures against each other in this case we're comparing actual versus budgeted profit so the bar itself is encoding our actual profit this year and then the red red reference line is encoding our budgeted profit so we can see that we're not quite to our budgeted profit factor nine thousand dollars short but we have some time left over and we're in a good operating range so the background color is essentially bad okay and good so the dark gray is considered bad so at the bar that black bar was inside of that dark grey reference ban and a background you know you're in a bad spot if it's in that middle gray shading then you're in an OK spot and if you're in this light grey area then you're in you're in good range so you're doing okay I'm also looking at data over time people typically just take a number like profits and they'll you know they'll roll it out over time and show it on a line graph and that's okay but a better and more contextual way of looking at profit is to compare that profit variance to the budgeted profit over time so this particular graph shows the profits variance to budget which you can see a huge decline in October so we have a problem there just a little bit of a different way of looking at the data but it really breeds a lot of life into the information that it's presenting to you so part two of three of inadequate context is adding a trend indicator to quickly breathe life into the number I saw that number by itself 780 K let's pretend that downward red triangle is not on the screen if you had a dashboard that had that number then you would probably think is it is 780 k good is it bad how does it compare to last year what's the trend you know all these questions are popping up in your head and they're all valid questions and a number by itself doesn't give you that information but just with a simple trend indicator a triangle you can see as you hover over that 780 K is in fact our overall profit but our month-over-month profit growth is is declining 97% so we do have a problem let's move on look at part 3 of inadequate context this is another example of of breathing lives into an ordinary basic number by just adding a quick spark line there's some kind of a visual cue to give you an idea of the context of the numbers so here you can see the 12-month trend for profit and the orange circles are giving you the high and low points for the profit over that 12-month period so we can see towards the end of the year we're spiking up around the holidays and then in the earlier parts of the year we're definitely lower in our profit mistake number three is excessive detail remember the dashboards are should be created to adhere to our creative thought processing and that numbers should be at a very high level you don't need precision on a dashboard most of the time when I say precision I'm talking about decimal points you'll need two three or four precision points after a decimal if you look at this - where we're trying to label the city state combinations with the profit numbers in the time line the trend chart we're actually looking at the exact date which causes a lot of data points to be shown and mapped on that line chart all these things cause visual confusion and by simply adding some dashboard best practices to this dashboard we clear it up immensely so we take off the labels in the map we truncate the labels on our bars we truncate the sales profit in the profit ratio numbers on our detail table to the nearest whole number and then on the sales timeline we're looking at the month year combination so that really clears up the time lines you can see the overall flow of this data as it runs through the time so very simple tips to clean up a dashboard so moving on to mistake number four inappropriate display media in this example we're going to pick on our poor friend the pie chart it tends to get no love at tableau but very few times does a pie chart really and effectively communicate information in a clear way so without the labels on the pie chart below if I asked you to rank order the customer segments by the percent of sales contribution you would get you would automatically know that corporate this number one clearly that is the largest wedge in this pie but the three remaining pieces consumer small business and Home Office they're relatively the same I would get hover over we can get a tooltip that will tell us what that is but this is not rapid perception so you'll be playing the guessing game trying to rank order second third and fourth place on this particular pie chart whereas if we encoded this exact information with a bar chart it's very easy to rank order these customer segments corporatism first place followed by Home Office small business and consumer we could also add a reference line to this chart maybe it would show the average percent of total sales and you can see very quickly which segments are above or below that reference line so bar charts cater to our pre-attentive thought processing they're perceived very very rapidly in our brains we don't even really have to think about it just happens we'll talk about pre-attentive thought processing a little bit later in the presentation mistake number four is inappropriate display media so in this case we're encoding trends over time with bar charts so we have budgeted cells as the yellow color and we have cells as the blue I think we able to all agree that this does not really encode the overall shape of the data as it flows through time it's not very effective as it would be if it was encoded with a line so really the rule of thumb when you're encoding information over time use a line chart a line chart doesn't a very effective job it's showing you this information as it flows through time another one that we see very often in dashboard design is meaning meaningless variety make people people think that folks will get bored for example if they have data in hand that is begging to be encoded with a line chart for example if you had nothing but data over time people think that you're going to get bored with that so they'll start creating or even inventing new chart types like this chart in the middle here this circle within a circle I don't even know what you would call that but it's not very effective right in this information be much more effective if they were all encoded with the line chart so you're not going to bore people if you're giving them the information in a way that they can rapidly perceive now if you give people information in this manner you just assume they're going to get bored because a line chart is quoted for boring then you're flat wrong people will get tired of seeing a dashboard like this where they have their brains have to work extra hard to decode the information that the graphs are trying to tell them so to solve this as I said just ink code the data with the correct graphs that the information is telling you to encode it with so if you're looking at data over time then it's perfectly acceptable to create a dashboard with nothing but line charts on it there's nothing wrong with that it's very clear and concisely to map out this information mistake number six is poorly designed display media and in this case we have a 3d bar chart which actually 15 years ago is probably one of my favorite visualization types I thought is really cool but the problem with this chart is if I ask you a question like hey what was the average monthly temperature in Boston in November of 1960 how you couldn't answer that because this chart suffers from occlusion and you can't see those bars way back there in the back so the way we can design around this with a tool like tableau is we can create a heat map that is encoding 100 years of Austin average monthly temperatures all on a single screen so very quickly you can scan through this and notice first thing is that it's pretty consistent actually there's some out lies it looks like in November and April but overall the last hundred years the average monthly temperature is being consistent and nothing is occluded you can see every record all a single screen very clearly mistake number seven is encoding data inaccurately this is something that I see quite a bit actually especially in journalism and it blogs with people who want to slay your opinion on a particular topic they can actually take advantage of graphs to minute and manipulate them in a way that will sway your opinion but if you look at this and I ask you what's wrong with this view you might not notice from the get-go what's wrong with it but if I said okay in January how much larger is sales when compared to the cost of goods sold you could kind of eyeball it and go okay so the cost of goods sold is one two three four five maybe six times larger so sales is six times larger than the cost of goods sold in January but that's not true notice where the y axis starts it starts at 20k it's not starting at 0 but notice what happens when we actually start that y axis at 0 the graph is telling a completely different story now now the cost of goods sold is roughly on a two and a half times smaller than cells so a very big difference and all we did was manipulate the start point of the y axis now by default in tableau the y axis always starts at 0 you have to manually go in and tell it not to do that keep that in mind always start your charts with a with an axis value of 0 mistake number 8 is arranging data poorly or bad visual design this is a very very common dashboard that we see out there in the wild nits and a tableau actually makes it pretty easy to develop poor dashboards like this because if you think about the workflow developing a dashboard and our software as you develop worksheets which are independent of each other and then at the very end when you want to wrap your analysis up you want to drop it on a single screen and that's all it is a simple drag-and-drop exercise and this is the end result so we have views that are placed rather randomly on a single screen and it's very hard to rapidly perceive what these views are trying to tell you and for the rest of these mistakes we're going to wrap them up in a overall solution so mistakes I believe it's eight through twelve will have a final solution for them so mistake number nine is ineffective hot highlighting and in this case everything is bright and bold so what catches your eye in this dashboard and the answer is nothing because everything is visually prominent prominent which causes sameness and same this is not a great thing on a dashboard you want that dashboard to show you the information that needs your attention quickly in a very effective and fast way now in this dashboard is absolutely impossible because there's so many different bright and bold colors that everything kind of looks the same mistake number 10 is useless decoration and as you can see if I didn't develop this one in tableau I just took a screenshot of a competitor's dashboard and threw it on here but there's a lot of mistakes on it I didn't do so thinking about decoration you know people like to dress up their dashboard with with quirky things like sparkles and glittered exploding pie charts and gauges and and pictures for example gradient backgrounds all of these things might be appealing to a normal user but it's going to eventually get boring it's distracting from what's what needs your attention which is the data and the big takeaway from is don't do it just stay away from any of these things that you see on this particular dashboard it's all pretty bad mistake number 11 is a very very common one that we see out there and that is the misuse or overuse of color and color is a very powerful way of highlighting data that needs your attention but it's also a great way of making a dashboard very difficult to digest and consume and in this case I'll just ask you a question would you want to look at this dashboard each and every day probably not because you're not getting anything out of it it's not really telling you a story so let's dig into the misuse or overuse of color just really briefly I'm going to show you three different ways to use color going from bad to best practice so bad is encoding each categorical element with a bar that the bar is encoding so small box is clearly the top bar you don't need to encode that with a color the user already knows that that is a that is its own independent element its own bar coloring each of the bars with a categorical categorical element just adds an additional layer of processing that has to happen in your brain and that causes confusion so a better way of encoding color to your bars is simply by using a natural color palette to the bars themselves add them all uniformed using the exact same color palette but an even better way of using color on a bar chart is to color encode something to need your attention with a very bright and bold color so in the in the third bar chart example you can see that the tan color is a profit margin is okay and then the orange is an alert we've got some issue with our profit margins so we're going to color that one a really bright bold color so it jumps off that page and really get your attention quickly so let's look at the overall solution two mistakes 8 through 11 so this is kind of putting everything together in a dashboard that's adhering to some best practices so in this case we've got the superstore data set we have our most important metrics laid out at the top this is what I call our 30,000 foot view and then the next layer down we've got our departmental metrics and then we can see that information on a map as well this is also highly interactive we can click on furniture for example lay down the furniture manager and I'm only concerned with the furniture metrics so now we're looking at a Furniture dashboard and I can see that from Missouri all the way up to Minnesota we've got some issues with our profitability as well as in Florida so I might want to drill in even further to analyze the next level in which is the product subcategory to see what subcategories are performing in these particular states and regions to really drill in to figure out what's causing this unprofitable behavior in these certain states but notice this once again the colors used very very wisely the data is laid out very cleanly on the screen and anywhere you see the color red we've got a problem all right mistake number 12 massive crosstabs with too many quick filters this is a mistake that we see from really from people with like a traditional bi background and in traditional biade products you know crosstabs are kind of the norm and what typically happens is you you create these crosstab reports people can download them they'll bring them into Excel or create some tippet tables and whatnot and it's a very clunky process and not only is it a clunky process but crosstabs by their very nature are not easily perceived so you don't get a lot of insight out of a crosstab and finally in tableau crosstabs can be very unformed non performant it's lots of barks on your screen you've got these quick filters that create unnecessary querying to the underlying data source it's just not a great user experience creating these massive cost apps but the good news is there's there's different solutions to the massive cross that problem and one of those solutions is what we call guided analysis so in this particular case we start at the highest level of aggregation in our data which in this case is our department and then we filter in to a lower level of information which is the sub category and then follow the individual item each of these views will filter the subsequent views that come after it so if you look at the title of each of these views on the dashboard as a number that number is simply the number of rows in that particular view so we can see in our crosstab that's fairly large they have about five thousand rows in our detailed view but that you can't analyze that I mean there's no insight that you can really derive out of this huge crosstab but by using visual filtering and identifying things that need our attention then we could filter that crosstab down to a much more manageable view and then we can kind of make smart decisions for that so for example furniture I can see that is the least profitable at one hundred and seven thousand dollars so I'm going to select furniture it's going to filter my other views so now my cross stuff has one thousand rows I can see the tables is by far our least profitable subcategory we're losing twenty eight thousand dollars on our tables I click tables it's going to filter our items our math and our cross time it's getting smaller and I can see that we're losing a lot of money on these bulldoze wood conference tables now if I click that now I've got a much more manageable detailed view we can see all the customers where they made the purchase of these chrome craft bulldoze tables and I can get their order details as well this is a much more performant and a visual best practice way of analyzing information okay so let's play a number game I'm sure you played this before but if you haven't this is it's pretty compelling actually so if I asked you to scan this table full of numbers and count the number of times the number five appears if you can do it but it's going to take you a while because it's going to involve what we call a tentative processing so this list of numbers does not exhibit any pre-attentive attributes that you could use to distinguish the number 5s from any of the other numbers on the front in this list but numbers are complex by nature their shape is complex so if I switch over to the next screen and all we did was apply a pre-attentive attributes to it one so we just bolded those number fives you can count these very very quickly because this is pre-attentive thought processing so changing the hue or the color of a number is a pre-attentive attribute and pre-attentive processing occurs extraordinarily fast in the human brain so this results in certain things standing out and a particular set of objects being grouped together all without conscious thought so attentive processing is sequential and therefore it's much much slower let's play another game easy little crosstab of information but it's going to be very difficult to answer a simple question from this crosstab and that's because our short-term memory is extremely limited so asking the question have we gained or lost customers over the last four years or which city has grown the fastest when you're going to be able to figure it out it takes a great deal of mental processing to arrive at that answer whereas if we encoded this visually with our line graph it's very easy to see the answer to our questions so by engaging our visual perception which is our most powerful human since it's easy to answer these questions of the data you so vision is by far most powerful sense but to really understand how to create dashboards to take advantage of this visual perception we need to learn a little bit about the the scientific side of visual perception and we're going to learn what works what doesn't and why next all right so short-term memory number one it's limited so we can only store around three to nine chunks of visual information at a time and our short-term memory and when that capacity is full for something new to be brought into our short-term memory that something there must either be moved into long-term memory or it's simply removed altogether and that means it's forgotten and we definitely don't want that to happen we want the key metrics to be very rapidly perceived and presented in a way that's going to enhance our thought processing so for example individual numbers on a dashboard or stored as discrete chunks of information but a well designed graphical pattern like the pattern formed by one or more lines on a line graph can represent a great deal of information as a single chunk and this is one of the great advantages of graphs over text you so there are four categories of pre-attentive attributes and that is color form position and motion I'm just going to go through I'm going to give you some examples of these so color hue is a more precise term for what we normally think of as colors so red green blue purple etc so we're physically changing the color of one of those circles makes it stand out from the other circles right it's different so that's on an outlier and the same with intensity intensity is really it really refers to both saturation and lightness it's not really a different hue but a lighter version of the same hue but you can see the end result is very similar in that that particular circle really jumps off the screen and graduate attention and that's the power of pre-attentive processing is that you can design dashboards that take advantage of these and your users will be able to scan a dashboard and get inside out of it very very quickly without their brains having to work overtime to get that information out it's just going to happen and they won't even know that you design your dashboard in this way it's just it works so in terms of form orientation line width line length shape size and enclosure are all pre-attentive attributes of forum so if you look at those those graphics you can you can easily see the outliers and that's because they are pre-attentive the next one is 2d location so that little circle that's kind of fallen off the table there very clearly that's different than the other three that are nice and and in line up in a row and last but not least is motion so in this case Flickr Flickr is an attribute of motion but you need to be very careful with flickr you know clearly this red dot will get your attention if i press the play button on this it's going to flash red and go away just flashing at you but if you can imagine you know 10 or 15 of these things on your dashboard blinking at you and a KPI is not mess it can get rather annoying but it's definitely a pre-attentive attributes all right so let's look at it before and after on a dashboard here so this particular dashboard was created to showcase our Amazon redshift connector it was new a few years ago and this is a dashboard that we had somebody had pulled together to to showcase that connector the problem of this dashboard is it really doesn't tell a story right there's three graphs on a single page there's a scroll bar on the color legend there's bright and bold colors all over the place but there's no cohesiveness to it there's no story to it so it's it's rather boring but applying some visualization best practices we could take the exact same data set and we can create a dashboard it is very very compelling so if you look at this one notice the title so the title is where it really starts so the sobering effects of drunk driving and how drunk driving impacted the victims in 2011 at the top it got your 30,000 side view which has your most important metrics displayed very easily at the high level and then you've got the map and if you compare the map from the before which was a field map it just doesn't have the same feel that a symbol mat has so each of these red circles is an accident or a fatality that involved a drunk driver the gray circles are accidents that involve the drunk driver but there was no fatalities so very clearly you can see on a map the the enormity of drinking and driving in the US and in fact this is just one year if we move over to the right you can see the most dangerous States and you see text this California Florida they're very dangerous when it comes to drink and driving when you're looking at the total numbers of fatalities and accidents but when you boil it down to fatalities per 100,000 people which is what you have in the bottom bullet chart it's a completely different list of states it's North Dakota Montana Wyoming South Carolina not necessarily the largest populated states but in fact the deadliest States Rd States these are the states need to watch out for late nights are violent if you buy the color red that those early morning hours late at night is when the majority of our fatalities happen drinking and driving you can see that the fourth of July and and around Halloween there are big spikes in drinking and driving and of course the weekends are killer people out drinking and driving most frequently over the weekend so applying some visualization best practices choosing the right visual displays that will really hammer home the story that you're trying to tell using titles effectively using colors effectively and arranging the data elements on the screen in an effective way all result in a dashboard that's easy to consume and is enlightening so let's look at the layout of a dashboard there's definitely a good and bad for placing elements on a particular dashboard and there's areas of the screen that get emphasized kind of Priya tentatively as opposed to other areas that aren't emphasized so if you look at a single screen and that's what you're looking at here the top left and the middle portions of the screen are the areas of emphasis these are the areas that you're going to want to place your most important graphs the graphs that are really going to hammer home the message that you're trying to tell you're going to place them in these pieces of real estate on your screen now the bottom left and the upper right they're neither emphasized nor they de-emphasize so that would be kind of your secondary graphs that support your your graphs that that provide that emphasis and then the bottom right is for your least important information and what we typically see in dashboard design is in that upper left-hand corner we see a big giant logo up there and that's distracting because that's the first thing your eyes see especially in the Western Hemisphere we read from left to right top to bottom so naturally when we see a page our eyes go right up there to the upper left and if you're seeing a big giant logo then you're missing out on the data so if you have to put a logo on your dashboard place it in the lower right hand corner you so digging in a little bit deeper on color we won't get into a lot of the details of color we could we could have a full-day session on this but here are some pointers so bright colors cause sensory overkill you reserve those bright bold colors from data that should stand out from the rest in other words a KPI that's broken or you know something that's bad should be encoded with a really bright bold color your brain naturally groups data by color so if you had a bar chart and you're encoding let's say cells by region you have four regions let's say the West region is red which is not a best practice anyway we're not supposed to encode the bars with a categorical element that the bars encoding let's say we were so the west region is red let's say your product category technology is also red on that same page your brain is going to try to group those two things together just because they're the same color even though they don't go together together your brain is going to do that naturally don't use red yellow and green on a dashboard and this is simply because you know 10% of our population in general or are red-green colorblind so they're not going to see the red green that you do and finally the definitely not least use colors that are common in nature for your overall color palette on your dashboard and then use those emphasis colors to really bring the light to outliers that need your attention if you can take advantage of these simple rules then you will not be you know colors won't be rampant and bright and bold and causing confusion under - would be very clear and concise so this is an example of two forms of red-green color blindness protanopia and deuteron Opia if you're looking at the chart in the middle this is what somebody with normal color vision would see so here's the green here's the red and then protanopia and deuteron obeah our two forms of red-green color blindness and this is potentially what they would see whereas you would see red and green they're seeing different shades of yellows and browns so just kind of an idea and there's going to be all types of you know the shades in between here depending on your your severity of colorblindness but this kind of gives you an idea of what somebody with the red-green color blindness would see if you're encoding things with red and green a better color palette for to get around that colorblindness imitation is like an orange blue there's also colorblind palettes in tableau by default so we'll get into some best practices here and the first one I would like to talk about is the small multiples so in a display of small multiples the same basic graph appears multiple times and by arranging multiple versions of the same graph next to one another you can show the entire picture with a nice pan and this is very powerful because it makes making comparisons very very easy so here we have a small multiple display of a scatter plot so no other display will show you this information in such a concise way where as we can see very quickly than in technology on the customer segment we have a very strong negative correlation between shipping costs and our profitability looking at the trend line makes that happen very quickly you so characteristics of great dashboards they're always well organized if you look at any great dashboard out there they're going to have these characteristics and an exceptionally well organized dashboard is one that is just it's clean it's clear it's really telling you a story it's very effective it's very very easy to see so always focus on organizing the information displayed on your dashboard it will pay off in that end result and how the pieces are arranged in relation to one another can make the difference between a dashboard that works and one it ends up being ignored all right so let's critique some very poor dashboard design and let's look at some ways that these could be improved so this particular dashboard is using red and green to encode in the bar chart which they have a good idea and that the color red means something is bad but using the colors red and green together is it could affect the that 10% of the population that could be consuming this dashboard and that they couldn't see those colors very effectively looking at the pie chart there's lots of wedges in this pie they look roughly the same shape same size it's very difficult to get any information out of that and if you look at the the most important piece of real estate on this particular dashboard which is this area up here you've got the filters and you have this crosstab with a scroll bar so that I highly doubt that this is the most important piece of information on this dashboard but based on where the Creator placed it it is look at another one so what's wrong with this one we've got a stacked bar in the upper left hand corner which is the most important piece of real estate that stacked bar is probably not the best choice for rapid visual perception and regional load is most likely not the most important graph on here it might be the average cost per mile I'm not sure but that upper left-hand corner is definitely the most important piece of real estate on this dashboard also the the organization of this dashboard is a bit lacking right this there's just a lot of views that are kind of randomly placed on it and that causes unnecessary confusion as well there's no flow to it you so what's wrong with this one well once again we have red green and yellow which is very very common in dashboard design I think the reason for that is I don't know people think of a dashboard maybe they think of a car you think of a car you think of a signal light I'm not sure the colors of signal light red green and yellow but definitely don't use those colors on your dashboard there's a lot of wasted space on this the use of gauges and trinkets are rampant throughout it it's not adhering to our pre-attentive thought processing and you can't monitor this at a glance and that's the point of a dashboard what's wrong with this one so once again look at the real estate being used the most important piece of real estate is a filter how do a ginge Ender affect the way you spend money there's scroll bars throughout the dashboard it's almost too condensed they need to get rid of those scroll bars colors not used very effectively their labeling the pie charts etc etc it's not very well laid out lots of problems with this dashboard so the recommended reading like I said earlier in the presentation if you want to get really good at creating these visual displays of information that we call dashboards these are the books that I would highly recommend Edward Tufte these beautiful evidence : where's information visualization and Steve infuse information dashboard design and if I would recommend one in only one then it would definitely be Steve infuse books this isn't the only book that he's written but he tends to ride in a way that anybody can understand and that you can take the information in his book and apply it to your everyday work very quickly it is very powerful and that's the way Steven writes so if you're like me you'll have to read the book three or four times for it to finally sink in but once it does you'll be creating very clear and concise dashboards the great news about this is that anybody can do it it just takes a little bit of practice so with that I will turn it back over to Megan and we can have a bit of QA thanks John great presentation um we do have time for some questions and answers so if you have a question please use the Q&A window and insert your question we haven't had too many come in and so I will give everyone just a couple of minutes to to insert their questions you you all right we have one question come in and do all great resources that you've given today some great tips and this recommended reading but you have any other suggestions of places people can go to ask more questions about visual best practices and dashboard design yeah actually tableaus website is full of best practices and different blog postings from knowledge experts on this subject as well as Steven whew Steve a few has a website called perceptual edge comm and that website is chock full of best practices and and white papers around this topic as well so I guess the answer is a quick google search and and you'll you'll get your answer great next question is if you put too many drilldowns in a dashboard is is that a good practice or not I think it depends on the overall functionality dashboard each dashboard is so very different than the other four drill downs they can be very very effective but but keep in mind the actual flow of analysis if you're if you have so many drill downs that the overall flow of interactivity is disrupted and it's probably too many my recommendation on drill downs or even in any kind of dashboard design is once you have it designed and you're happy with it send it to a few people and get their critique on it is it confusing to them because if it's confusing to them it's going to be confusing to a broader audience may need to bring it down a little bit and not do so much drilling down but do keep your dashboards objective oriented if you have them tied to specific objectives and and detail out what those objectives are somewhere on the dashboard maybe it's a help if you look at this screen I don't if you can see the little triangle here and this is actually not a this is very pertinent to this particular worksheet in tableau but this is a help menu so on a dashboard you could easily have like a little question mark in the upper right hand corner you hover over it will give you that overall understanding of what the dashboard is designed to do now this dashboard is that it's designed to answer questions a B C and D and it's interactive and that you can click on the bars etc etc or you can drill down by doing these things so that way a new user who's never seen the dashboard and kind of get that real high-level understanding of what the dashboard was designed to do so that's a little trick that I use for my dashboards just a little help icon and all this is another worksheet with a tooltip and I floated that in the upper right hand corner so I hope I answered your question yeah that's a great tip one question came in other things to consider when designing dashboards that will only be consumed as a screenshot so I know we like people to look at dashboards interactively being able to do actions and drill downs but you know there's those instances where people do a quick screenshot email it over so do you have any suggestions of when people are building their dashboards how to take that into account now naturally being a tableau pre sales person I would say tableau servers your answer and and it really a static screenshot of a dashboard is rarely as effective as an interactive one and tableau server a definitely bridge that gap but I do understand that there are cases where you just want to take a quick screenshot and email it over to somebody and say hey look our profit is really bad in this region check it out in that case I would make sure that on that particular dashboard if it is a dashboard that it was designed for screenshots that you have text in there that is used smartly that tells you exactly what these views are doing right because you can't take advantage of a tooltip on a static screenshot so maybe you have the title of the chart in like a 11 or 12 and then underneath that you would have some underlying information that gives you kind of a metadata of that particular chart another thing that you can do on a static screenshot of a dashboard is use the annotations so annotations are a great way to get get somebody's attention on you know something that is an outlier or write something that you wouldn't be able to find by interacting but on a static screenshot that annotation can be very very meaningful perfect thank you another question if you do have a colorblind audience how can you design your your dashboards with not compromising the colors that you're using giving recommendations for that yeah and I appreciate that question because that's something that most people don't even think about is you know especially if you're not red green colorblind and that is the most predominate other forms of colorblindness but red green is by far the most prominent but if you're not suffering from that condition you don't even really think about it so I appreciate somebody asking the question about that because you do need to keep it in mind so to answer the question though tableau has colorblind palettes installed by default but you can also use other palettes in tableau as well if I go to a new sheet here let's just build a chart real quick and I'll show you see market will do cost of goods sold I'm going to show you some of the the palettes we have available so maybe on this we want to color and coat it with profit and it's defaulting to the green which is okay but it's a little difficult to see maybe one encode it with an orange blue eye kind of default to an orange blue palette when creating these dashboards so that our colorblind friends can can decipher those also if you if you look at a discrete color palette we have a colorblind 10 right here so you can use any of these in combination with each other so that our colorblind people can can distinguish these from each other but the big key takeaway is to stay really stay away from red and green together red and green really have a they pose a lot of challenges to those folks are red-green colorblind I learned something new every day I didn't know that the red-green colorblindness was mussed I wouldn't want to say popular but that's the majority is what people are suffering anyways that's all we have time for today John I'd like to thank you for your time and this great presentation and I like to thank all of you for joining today as a reminder we will email out a recording of this presentation as well as the deck John used we hope you can apply what you learn today in your day to day work with tableau and with that have a great day thank you thank you

Channel: Tableau Software

Views: 86,716

Rating: 4.9015918 out of 5

Keywords: data analysis, data visualization, business dashboards, business intelligence, tableau, tableau software, Talk data to me, tip

Id: GnMSjSWDQNk

Length: 59min 18sec (3558 seconds)

Published: Wed Dec 16 2015