5 Expert Strategies for Effective Visualizations - Panel

hi everyone thank you all for joining our panel today my name is aaron stanton and i will be your moderator as we walk through data visualization and data for data storytelling i'll quickly introduce myself and then i'm going to hand it over to my co-panelists to introduce themselves so i've been involved with big data for about 15 years now and i have to admit that i used to think data vis was a waste of time up until about five years ago i had a real aha moment when i realized that a good data vis compelled my audience a lot better it told my story a lot better than a bad visualization so since then i've been just as addicted as hopefully the rest of our panelists so david i'm going to hand it over to you to give a brief introduction great thank you my name is david ciomo i work at humana as the data visualization principal i also lead our internal group called the visualization center of excellence which focuses on telling the right stories with our data but more importantly coaching educating and mentoring on how to do so so we are a self-service experience for our local associates and happy to be here thank you great and helen over to you hi um so i do mostly consulting work and i work at um the i do data science courses for cornell university's online program and i also create courses for linkedin learning that focus on data science tools to solve various problems using tools like microsoft power bi aws excel and r that's what i'm currently working on very cool and vigness my name is vignesh orion swami i'm a data science instructor at data society i teach data science classes data literacy classes uh in more hands-on data science classes related to statistics and machine learning in uh python and r um and uh really looking forward to talking more about data visualization with everyone cool so we figured it would be fun to kick off our panel with probably one of the most controversial data visualizations that exist out there the pie chart so give me one second as i go ahead and pull up my screen okay so on the right we have the dreaded pie chart and on the left we have the beloved bar chart obviously you can see where my skew lies but helen why don't you kick us off what are you thumbs up thumbs down pie charts and why so when we look at the bar chart on the left you can easily see that red is the largest and what we see is that blue um you know based on looking at those heights that it looks like maybe blue and orange are roughly tied and i typically will order them um if i'm doing categorical data and if i'm doing time series data i just do that in the order of where the time occurs um but the pie charts harder to measure out like if anyone's tried to cut you know pi so that it's in equal pieces it's hard and so pie charts are difficult it's difficult to tell the proportions apart because we have trouble um you know humans and we have trouble reading a pie chart because we can't it's difficult to look at the proportions and tell those apart parents everywhere i think know the struggle of trying to cut something equally for their children right trying to cut a cookie evenly for your children is very difficult so i liked that point uh david did you want to chime in what are you thumbs up thumbs down on pie charts i'll be honest i'm on the fence i generally say i don't like them and the reason being is because we nine times out of ten we use it inappropriately or we're trying to tell the wrong story with a pie chart you know it's a composition story parts of a whole but a lot of times people will actually use pie charts as a comparative story comparing size and pieces or percentages etcetera so i do tend to lean toward only using pie charts if the story really is appropriate for pie charts i love that and i think we're going to come back to a couple concepts you just mentioned there vignesh you mentioned that you're focusing on data literacy right so how do pie charts tie into data literacy and vice versa well pie charts are uh you know i would say that i'm probably three quarters thumb down on the pie chart i think the challenges that uh like helen mentioned the difficulty with comparing the segments and you know really when you're looking at the pie chart where you're comparing are the angles which is difficult to do uh whereas if you're looking at a bar chart uh on the left hand side it's easier to compare the heights so it's a it can be easier to compare the bar length as opposed to the angles um but one thing that i'll say about the pie chart that actually makes it three quarters instead of fully thumbs down is that it's a really common visualization and a lot of i'd say almost everyone that i've encountered is familiar with what the pie chart is doing so that that's one advantage is that it's very common and most people don't need to be instructed uh to in terms of how to interpret the pie chart and for a relatively small number of segments less than five as uh as a rule of thumb i think that it works pretty well if uh the difference between the slices are big enough that you can eyeball the difference the challenge is going to be when you go beyond five segments or um when the slices of the pie are so similar that it's hard to compare them that's a really great point around people are familiar with it right there might be more successful data visualizations out there i'm a huge fan of the box and whisker plot for example but i pretty much have to explain that nine out of 10 times that i show that to someone so maybe that's not as successful and i think that's going to lead i'm going to go ahead and stop sharing my screen and that's going to lead into uh my first real question here so um what in your all of your expert opinions makes a data visualization successful and david i actually want to start with you on this one because i loved you know your contextual answer around the pie chart the way i see it and again i need to take a step back from a moment and ground that in order to be a data-driven individual data-driven company data-driven mindset we have to talk about data literacy and data literacy is really just about understanding not how do we not just how we read and write with data which a lot of the folks in this conference is probably familiar with but it's also communicating and consuming it so when you talk about data literacy and you talk about that consuming and communicating of data you're talking about data storytelling when story you're telling you need the appropriate and proper data visualization to convey that story so the way i see it is data visualization is and i don't mean to downplay but it's a means to an end to the data story so as far as i'm concerned if you have the right data story and you know what you need to communicate with your data choosing the right data visualization is going to be critical but it can really make or break your data story as well because if again if you're mis aligning charts or misassigning certain visualizations then you're not going to get the insights and the actionable opportunities that you intended out of your data um so again the criticality i also want to just mention that because we do consume visuals shapes colors and pictures 60 000 times faster than words and numbers visualization is really important when you think about how we as human beings consume information every day of our lives whether you're involved in data in analytics or not that's that was a fantastic answer i love doing an exercise with our interns every summer where i introduce the concept of data visualization and how much better we are at depicting visuals compared to numbers so i pull up a whole screen of numbers and i say spot the trend and they're looking at it and thinking they're going to be judged on it and they you know it takes them forever and then i pull up a chart and i say spot the trend and everyone you know in a couple of seconds points it out so that was expertly put helen did you want to add to anything that david just said around what makes a data visualization successful so i think that one thing that i've approached that i've tried to take so i'm i wouldn't describe myself self as a user accessibility expert necessarily but i think that's something to think about um in terms of we're not creating our work for ourselves like i think it helps us to understand what we're doing but we're really creating it for someone else to understand and so it's the bridge between the data and the audience itself and i mean i'm always looking for examples of how to convey that and sometimes it comes down to thinking about what type of answers is someone in the business or your consumer trying to get out of it so for example there's a lot of businesses um in operations and organizations where they look at data on a daily basis and in that case they're going to look for you know something like a calendar visual is helpful um because you can see when the weekends occur month starts right because unfortunately we don't live in a world of the 30 the 3360 um accounting calendar um so it's putting it the correct context for how the end user is going to consume it and what they want to get out of it and the other thing is um i try to i go for the approach of i put enough instructions and um like labels on the visual so that it explains what it does without necessarily you know convoluting and uh you know cluttering it so it's like okay what's going on so it's all about you know using that as a communication using visualization as a communication tool so david was you know speaking around the story and you know what makes it successful is what conveys your story and helen i love that you spoke to both your audience right focusing on the audience focusing on what you're trying to tell them and making it as easy as possible for them to consume your data visualization vignesh what about you yeah there's actually a couple comments there that i wanted to highlight and when i think about good data visualizations i think about number one what is the message that you're trying to communicate and then number two who is the audience that you're trying to communicate that message to and the goal is to me to always communicate in the most efficient effective way possible so you want to communicate that message and you want to do that in the simplest way possible and what's interesting about that is that sometimes that means that maybe data visualization isn't always the best way to communicate maybe you use a sentence to communicate maybe use a single number um but like you said when you're dealing with a lot of numbers and it's difficult to identify trends a data visual visualization can be very effective so the goal is always to communicate in the most uh efficient way possible and depending on your audience the uh the visualization may vary a little bit a simple kind of rule of thumb that uh i like to teach the students is that your audience should be able to understand the visualizations in about 10 seconds or less as a rule of thumb i mean it may vary a little bit but if they don't then maybe it's worth considering a different type of visualization or even considering if visualization is the best strategy to communicate you've already given us two great rules of thumb kind of five or less for pie charts and

now 10 seconds or less for my data visualization consumption which this is great because we promised that we would give people strategies so um as someone who personally is stuck in what i feel like is dashboard land i'm sure maybe some of our viewers are also stuck in dashboard land where these don't feel like data story right data stories um these are not succinct right i have a lot of different charts and and and graphs and all the things in my dashboard how what kind of strategies are there that you all can suggest to take to elevate our dashboards to elevate them up to stories to make them more effective for our audience and vignesh i'm actually going to start with you you've given us two rules of thumb i'm seeing if you're going to give us a third for dashboards i will give you a third rule of thumb uh so we have a nice framework that we teach our students in our data storytelling class and the idea is that this five stage process kind of helps you think through what your data story is going to look like so it starts with just identifying what your insight is so what is the message that you're trying to communicate which believe it or not is oftentimes the hardest part of the whole exercise and it'll really force you to understand what you even want to communicate to your users uh and as uh as data scientists or data visualization experts sometimes uh sometimes actually that's not always clear to us so it's good to understand you know what is your insight once you identify the insight you tailor that insight to your audience so how do they consume information what do they care about what are their needs and wants what are the salient points of the the message that uh are going to really matter most to them once you do that the next stage is outlining the story uh and the idea of a story is that you're arranging visualizations narrative into this kind of linear sequence that has some logical flow to it once you outline that you plot it with a storyboard so you identify you know what is the how do you engage your audience you slowly build up to kind of the main theme of the story um you have a resolution at the climax of the story and then you kind of have some closing thoughts to tie the story together um and then finally after all of that that's when you actually format the story for delivery and the reason that comes last is that you don't want to start with the story at the very beginning you want to identify you know the outline of the story be as iterative as possible at the beginning once you identify uh the outline of the story and the main important parts of the story then you start actually uh thinking about what the best way to format the story for delivery is whether it's a powerpoint or a dashboard or you know whatever presentation method you want to use uh but this way you stay flexible at the beginning and as iterative possible at the beginning you've said the word irritative a few times throughout that and i think that's such an important um tool when it comes to building data visualizations is understanding that you're probably not going to get it right the first second third or maybe even fourth time and you need to keep evolving on it david what are your thoughts on dashboards and any anything that we can do to elevate those into stories yeah and again i kind of see data storytelling as you know coming to fruition as a dashboard or digital report i mean ideally that's what happens the reality is that too many of us have with the democratization of data the way it's developed over time and with the accessibility of bi platforms like power bi where everybody can gain access to it um everybody kind of has the keys of ferrari but not everybody's learned how to drive yet and so what do we do and uh we've in addition to what vignesh has talked about we've got a very similar kind of story sculpting process but we also add in there the process of emo executive managerial and operational this helps us kind of framework not only who our audience is but most of us have too much data we don't know what to do with it and a lot of us try to put all of that data on a single screen so it helps kind of organize your thoughts to say when you come into a dashboard the proper use of a dashboard is that almost high level aggregated view what are my biggest insights what am i really trying to say what are those extremely obvious aha moments and when a leader consumes that dashboard it's that person's coming into it with the intent of saying i'm going to consume some information but i have to leave that dashboard being able to either make a decision or be more informed than i was before when you get down to that managerial level view that's that executive level view this is the opportunity to do your drill throughs or drill ins where you're going into a kpi or chart to unpack it a little bit to see what's behind that what are those drivers what's moving those numbers up or down or changing it from green to red then you can drill through even deeper down to the weeds down to the minutia of the detail of the data that can be your spreadsheet that's your operational level view but the reality is your senior executives are not going to want those spreadsheets they're going to want it you to distill the information and then tell them why they should care about what they're looking at so as you're sculpting the story think about these layers involved humana has over 29 000 dashboards and digital reports in our bi environment and even though we're a very large company we don't need that many things we need fewer things and better things telling better stories and i think a lot of folks because they have access like i mentioned the beginning to the data and to the platforms people just get into this evolution of creation not really slowing down enough to think about what am i really doing am i helping or hurting am i muddying the waters with putting another dashboard out there or am i doing this in a strategic manner for the data to come across to leadership as being meaningful actionable and valuable i loved that emo concept i actually was taking notes for myself personally to uh take this away and take this back to vert 2. so one follow-up question for me and hopefully for the rest of our audience when you're doing this emo breakdown which is just so clear are you keeping that in the same dashboard or are you splitting that out into completely separate views so you know how we have um i'll kind of draw back to excel because it's the most foundational tool that we all use we have tabs or sheets in click view you also have individual what they call tabs or dashboards i like to think of the dashboard is the big thing but the dashboard is made up of reports underneath it so to me the dashboard is the experience that the individual is coming into now coming from the world of marketing and advertising i'm used to creating experiences for people that ui ux component of this very technical exercise of creating a dashboard so i think about when somebody comes in i want the experience to be clearly laid out they're coming into what i usually call either an overview or an executive summary page of the dashboard but then it's tabbed out to say maybe there's a financial tab or maybe there's a more detail tab but as you drill through certain metrics or kpis on that executive level entry dashboard page it automatically connects you to other pages within the dashboard i don't like to segment it out to say that i'm giving you a thing with multiple dashboards that just sounds really overwhelming it sounds confusing and it sounds really heavy for a leader to consume yeah and i even um i love the concept of even within those tabs almost saying executive management operational because it just i know what i'm going to get doing there that's absolutely and the reality is too a lot of times you'll have operational folks using the exact same dashboard that an executive uses and so rather than creating multiple tools right based on roles you can create a one-stop shop that everybody goes to but depending on who they are whether it be through a directory or something for that effect they're logging into the appropriate level of data consumption very cool and so um i want to transition now into the tools of the trade and helen i'm gonna bounce this over to you what kind of trends are you seen around demand in tool training you know who do you what tools do you think are coming out with new features we should all be aware of what i'll let you comment on the tools of the trade so they have features coming out all the

time i feel like even though i spend a lot of time working in all these applications that i'm still always trying to keep up there's always something new to learn um you know something else to integrate something else to think about um what my approach uh is to think about to kind of scale it up um because what i've found out is that when i create something let me see how they use it right and you could start in excel you know create a chart that we could use in our you know there's multiple ways you know to do that um and then seeing how that works and kind of scaling up i think that like so i would say for power bi for aws for these interactive tools right because um you can create our visuals in power bi but what power bi enables us to do is we can set up refresh schedules and we can share access so that we don't have to um you know keep an eye on the data on a sunday morning so you know making sure that those kind of questions are answered there's it's kind of clear what to do to hand it off to someone else um i think that a lot of the you know big data was kind of uh something it was talk a buzzword you know a few years ago but a lot of the tools are exploring options as to how to handle the data because um when you look at as uh david mentioned um is that often like executives are looking at a high-level aggregated view but you want to have it so that it's accessible for other people as well because if otherwise they may be looking at the different things different numbers and so putting it together um where that's possible and i think the other one of the analogies that i refer to a lot for data visualization is to think of it like a flat pack of furniture um so if you go to ikea and you buy you know you'll see you buy a dresser uh so it's already been measured they've already measured cut they made sure it fit together and they've packaged it you purchase it and then you have to put it together and that's kind of the quickest part because you don't need to measure the pieces figure out how much material you need but the thing is that that's the ikea proposition is kind of the um it's the golden key or the you know the ultimate goal of really all businesses is how to make the users feel like they did most of the work even though that's not necessarily the case so i would say um you know it's similar to uh like baking mixes where you add like egg and water even though they could put that in the mix is it gives people an ownership in what they do so they can change it and they can update it oh yeah you know that they are part of the analysis process they own the numbers and you know even if you know i'm the person that did a lot of you know a lot of the hard work and you know hard the lifting behind the scenes so you mentioned you know obviously you're teaching courses at both cornell and through linkedin but you mentioned yourself that it sometimes feels like a struggle to even keep up with the new stuff coming out are there any tips or strategies that you go about to try to keep up with the new features and functionality that are coming out of the available bi tools um yeah there's a lot of uh blogs to follow um updates you know i think that linkedin is a good place to track a lot of those updates people share a lot of material that's really great and i would say uh the other um thing is just be open to learning new things and i'm open to changing the way that i do things if i think there's a better way to do it and if the business you know so for example in power bi if you can download up to 30 000 rows of data from a visual if someone says you know we have more than 30 000 rows we're not able to download it then that's kind of um an indication to me that i need to kind of clarify what they're looking for um from their data what are their goals from looking at it um so that i can you know remodel that remodel what i'm doing so they don't have to download it so they can have easy access to it very cool so i want to remind our audience that we are open and accepting questions for our panelists so make sure that you get them in and i actually did receive one question here that i wanted to start with and um i think vignesh i'm gonna put this one to you what should someone do who's getting started in data visualization what tips do you have for someone who's just new to data vis i guess this would have been me five years ago when i had that aha moment that it's worthwhile to put effort into data vis but what can you what can you tell people who are just getting started yeah great question so there's two things that i would say number one is to

identify a project or something that you're interested in visualizing so that could be related to some work data set that you have or some personal data set that you're interested in maybe you really like the nba so much data is available and there's some uh you know a lot of resources out there there's kaggle there's ucr google has a search engine for searching for data sets and look for a data set that you're interested in and start visualizing with it and what i'd say is be tool agnostic at the beginning really it actually doesn't matter what tool you use whether it's excel or tableau or power bi or r or python i would say focus on the process of building the visualization first using a tool that you're already familiar with or honestly just try sketching it on a piece of paper don't be afraid to just start getting hands-on with the data and building a visualization in whatever way is easiest for you whether that means using a programming language or using a tool like excel it doesn't really matter uh the tools of all the tools change um but the concepts and the principles of creating good data visualizations never change so that's what i would suggest find a problem that you're interested in and use whatever tool or method is easiest for you to build visualizations with and go from there love it and i love that you said draw on paper my i just finished my masters and my dataviz professor said when you get stock draw on paper we get you know even as data visualization experts we have our own bias when it comes to selecting visualizations and i'm prone i already said i love boxing whiskers plots right and so he told us draw on paper first because it kind of removes you from those habits so i love that you mentioned drawing on paper here so i have another question and david this one is directed at you are do you run into any challenges convincing senior management to use data visualization and the dashboards that you all are building or how does that all work oh all the time

no i mean you know here here's the reality again and i use that word a lot because i do think that we live even though we are conscious of what's happening around us we live in this world of kind of old school technology as generations get younger and younger coming into the workforce things are changing but a lot of us deal with a lot of senior leadership that are of our generation or older and so yet everybody has an iphone everybody has an ipad everybody has some type of handheld device where they're consuming information visually in very short chunks whether it be social media or stock markets or advertisements etc psychologically from a neuroscience perspective we consume information without knowing it and so what i really kind of helped coach our leadership on is the visualization whether you know it or not is going to tell the better story um let me prove it to you and so i will do exactly what you mentioned earlier i'll show them the spreadsheet of the data especially our finance folks and then i'll show you the finance person them a report or a dashboard where it's visualized and there's this immediate like oh that feels right oh now i can see they don't realize it's happening because it's happening almost unconsciously or subconsciously right so i do think that a lot of uh here's the unfortunate truth as well a lot of folks think that data literacy data storytelling data visualization they're just industry buzzwords but they're not i mean the reality is this stuff is here to stay it is real and um we've got to coach folks that things like skill gaps within our technology forces um the data science folks the analysts the engineers the architects they all have to get up on visualization and color psychology and how to use data visualization data storytelling and our creative marketers and communicators and all the people creating the pretty stuff have to get smarter about data and how to use it how to access it and what to do with it we've got to find this middle ground between the two worlds in the meantime it certainly serves my skill set best because i've got a whole team of the data folks doing what they do best and i do all the creative visual stuff that i do best and we come together um it's their peanut butter and my chocolate right it's that old metaphor so uh yeah i think you just have to plug away at it you've got to keep showing that it works practicing the right philosophies principles best practices of data visualization prove it out eventually you will make them believers you use such an important word there practice that i've really come to appreciate i think with data visualization is it is a skill i think that does uh require practice that was fantastic um so helen you had mentioned power bi and i know vignesh khan had talked about not worrying about specialization but are you seeing you know folks within your community in the learning community trying to specialize become known as specialists in a specific tool or not and is there any value in being known as a is a power bi expert or do you think people should try to be more generalists um so i use power bi a lot but i use other tools as well i think it's not one tool it's a toolbox so to speak right um it's i'm not looking to i mean i think it's good to be experts in areas but i but the technology will change right five years from now you know what's what are these tools going to look like um you know how are they going to be integrated i mean power bi so only came out in 2015 so it's pretty new um you know and there's a lot it kind of joined with applications from microsoft that were much older like um sql server analysis services and they kind of built it you know it's kind of the platform on top in a lot of ways um and i think that there's a lot of way there are many different ways you can answer a question um another thing is that a lot of organizations use excel so there's a lot of comfort level to using excel and as david mentioned you know kind of how to implement and um you know promote using visualizations is often i will put a table next to a chart and then that's a way to kind of so people are comfortable with the numbers they already have and now they can see in a different way so i think it's kind of this gradual introduction um and examples of how they can use it to make them more comfortable you know going in that direction so you're kind of giving them training wheels and as someone so the whole reason i did not believe in data visualization until about five years ago is because i am a finance person and we love rows and columns and columns of data in excel and so i like that you kind of are slowly introducing it letting people get comfortable with the data visualization and then maybe at some point you move over exclusively to the data visualization so we only have a couple minutes here to wrap up and i do want to give each of the panelists just one opportunity to give you know vignesh you've already given us a couple rules of thumb but i want to give everyone the opportunity to speak to one tip that they can give away to our audience um and it's got to be short because we only have three minutes david i'm gonna go ahead and start with you what's one tip that we can all try to apply when we're doing data vis story first story yeah i'm gonna harp on that story first the tools doesn't matter what tool you use doesn't matter you know it does matter who your audience is but the story is first it doesn't matter where your data is coming from if your story isn't clear people aren't going to get it people you're going to be leaving people to have to make the conclusion of what about that data they should be caring about the visualization in the story helps them or helps identify for them why they should care so make sure that you focus on story first and then figure out how are you going to build it and what date use awesome big ness yeah one tip i'd say is uh one thing that really helped me get better at kind of understanding the principles of data visualization was finding good visualizations and also bad visualizations and forcing myself to not just know that they're either good or bad but try to understand what makes a good data visualization good what makes a bad data visualization and bad and once you kind of have that eye you know oftentimes we have a very intuitive way of understanding what's good and bad something just looks right it's easy to understand but forcing yourself to understand what makes it good what cognitive principles is making this visualization good or is taking away from this visualization can really take your data visualization skills to the next level i love that i actually keep a folder on my desktop of stuff that i love and stuff that i hate i mean helen i'm going to let you bring us home one tip you can leave with our audience um so i would say uh that data visualization um understanding what people are doing is part of the process and so i think listening to what other people want to learn what they're looking at finding out in every organization right every consumer they have something that they're trying to um optimize they're trying to uh you know maximize something minimizing what are they trying to answer i'm trying to get to the root of that so you know to orientate um you know the visuals in that way and to that extent also um you know i try to focus on one thing at once you know kind of asking and understanding one question at a time and building it from there kind of one piece by piece i love that so make it easy and approachable start small and then build on that so this was absolutely fantastic and i wanted to thank all of our panelists and the audience for joining us today i think dr joe i'm handing it back over to you hopefully yep i'm here hi oh my goodness this is like a uh this is a wonderful reunion to see all you guys again uh thank you so much uh aaron and big nesh and david excuse me and aaron uh you guys are just absolutely phenomenal touching on the things that i am just the most passionate about when it comes to uh to data visualizations uh i was supposed to have been on mutant i probably wasn't i really started chuckling tried to restrain myself when you guys were going off on that on those pie charts you know um it's a thing i like to say uh pie charts you never know about them some people think they're the best thing since sliced bread while others say they're the spawn of the temple themselves you know and it's not it's just a tool right it's in all of these things are tools right it's how you use them the tool is neither good nor bad hey everyone i'm todd in the hands of the person wielding the tools so um you guys just really took us through that in a very uh applicable very actionable very practical and very i don't know you've achieved both relevance and resonance uh with this panel so i'm very grateful to all of you for uh the wonderful insights that you um that you brought to the table this afternoon thank you very much thank you