 I run a company called FusionChart, which is a data visualization conference. I did not start out to build a company. It was a way for me to make pocket money. For the last 10 years, it has helped me make a little bit of pocket money. We are now 60 people across Bangladesh and Africa. Frankly, what we do is build JavaScript components, which other companies, anybody who wants to build a dashboard or a report, they use us for more than 20,000 times. My focus area is building such visualizations and setting it as a framework or a library to other businesses. I used to be a tech guy. Since 2010, most of the courtship by the company was hand-coded by me, but in 2010, things started growing. Since then, I have not had a look at the court. For the last three years, I have been at zero technical knowledge. So if you want to throw some tech questions at me, I will probably just be using my own experience. My goal here as a creator. My focus is how do you use visualization in business and how do businesses pursue visualization. So there is this whole, I mean, data visualization in the last couple of years has gotten a whole lot of traction, and you have seen different kinds of visualization. So what typically we are interested in is visualization. We are seeing the new farmers, infographics and data journalism. We are seeing some really cool data visualization, which it takes us to just understand what the guy was even trying to visualize. And all of these together are forming a very interest. They are converging at a very interesting point where there is so much more data available today and we are working on that. What I am going to be specifically focusing on is the business data visualization part. The good part is most of the learnings from here can be extrapolated to any visualization that you are doing. So I am going to be talking very specifically about charts and charts and how business use charts. But anything which is implied here can be abstracted, extrapolated and we put it into your data visualization where you are doing. So the way I have structured my talk is there are two parts to it. The first one is the gain part. Based on my interaction with clients, a lot of clients tell me this is the data, now visualize it. The problem is you don't just start with data and start with the visualization saying that you don't want to just convert this into visualization. It is more like the typical hacking process where you say that you know what I can code, let's go build a code without knowing what am I going to build, how my environment would look like, what would my ARL be like and you backtrack it. Data visualization is also a process. It's not like throw me some data and I am going to convert it into visualization. So the first part talks a little bit about the process. How to structure that process because this is still, this is not still something set in stone. So this is something which is my opinion, some parts borrowed from a few other industry people and which are important. And the second part is more about the visual part. I am going to show that this is what we think is right and this is what is wrong. And how does it relate to the human brain, the cognitive thinking and the visual function of the brain. So quickly I mean start, generating about 2.5 quintillion bytes of data that's 10 to the power 18 every single day. And right now, this is from just three. So right now we are at the data capture and organizing stage which is where the big data comes into play. Small efforts are going under in the understanding part but optimization and leveraging is still not happening and this is what will happen over the next few years because this part is already there. This is increasing with the internet of things with all the sensors capturing more data. Organizing is still happening at very thin level. More and more people are now becoming data scientists and they are recognizing that data is now something like an item of your corporate balance sheet. Till today data is something that is not really valued but if you look at some of the biggest companies like Google and Facebook say value for the data and it will eventually be a part of your corporate balance sheet. Understanding of data right now to understand what is to be understood is also very early. Like you have this past set of data what can you make out of that. That understanding is still not there and people are trying to figure that out. There is a very beautiful book on this one called Big Data by Victor Mayers. If you were to come pick it up. And Al Belian, who is Google's chief economist he said that the ability to take data to be able to understand it to process it, to extract value from it to visualize it, to communicate it that is going to be a hugely important skill over the next few years. So the way I look at it is data visualization is not just technology. It is technology mixed with art mixed with science and there is obviously a whole lot of maths behind it. So it is a multi-disciplinary stream where a lot of them can be learned vocationally. You don't have to necessarily be an academician to be able to learn that. And the way I define it is it is a visual display of measurable quantity which means anything you can measure using these six things. Points, lines and curves. You have a coordinate system whether it be 2D or 3D. You have numbers. You have shading. You have color. And you have symbols. There is one more thing on top of it which is in alphabets which you can put on the data visualization. Those are the four constituents of data visualization. And it has a clear purpose either to understand data to substantiate an already known or something which a hypothesis you are forming or to discover from data. So based on this there are three kinds of data visualization. Typically people say two but there is a third one which is up and coming artists using data for self-expression. So the first one is explanatory which is built around a specific thing. So anything which you see in the papers is explanatory because the guy has already figured out what he wants to show. He is pointing it as a story to you and you are just absorbing that story through the data visualization. The second one is exploratory which some of you guys are doing here today where you have the data and you are trying to figure out if there is a pattern in the data. And then the final one is the exhibition of the data which is a data as an art about changing gears. So in terms of business data visualization so most of the business data visualization today that is there is explanatory. Today there is right now there is a new breed of tools coming up which is called visual analytics or visual exploration which are exploratory tools that you must report of Tableau software of sorts click view also these form into the going to the exploratory. But irrespective of what you do there are three basic steps which we think are essential for a business data visualization. Know your audience and need for data visualization and this is a framework choose the right visualization and once you have done both of those which are functional things then go ahead and enhance the visualization which means you add insights to it where you make it more stylish or whatever you want to do with it. Now know your audience constitutes of three parts. What role is this information for? Is it a C level guy who is going to look at it? Is it analyst, is it operational guy or somebody else? What department does it belong to? Sales, management, marketing and third what is the metric that will interest him and help him achieve his goals. So going back the first step is knowing your audience there are three steps in knowing your audience. Now this is abstraction of data by role. So if it's a C level guy he is looking at very strategic level data he is looking at very top level data to look at the health of his business. He is not interested in your very minute data he is not interested in real time data he needs to know things like simple summaries and indicators. Then you have the analyst whose job is to figure out values of data and he might need real time data. So for him the data he would need detailed query analysis he would need detailed data with precision and then you have the operational guy who needs the data as is right now to complete the task in hand which means events, alerts and all of those kind of data. So this is data required by department. The sales would need leads, conversions average sale value, closure by marketing visits, acquisitions a lot of independent department based data. Now what is interesting is when you meet, when you match role and department and then the goal of that guy who is looking at it that constitutes a metric. So for example in customer support itself there are three different roles you have the support executor you have the head of customer support and then you have the CEO within customer support they need to look at different metrics so the support executive sees the number of tickets assigned to him his personal turnover on line. The head of customer support sees total number of tickets by department and issue type and the turnaround of the team and efficiency as well and the CEO is only interested about one thing what is the customer satisfaction index. So even when you are looking at support data segmented by a role so support is a department segmented by a role then match that role and department and then figure out what is that metric. Similarly in sales the sales associate sees the number of leads target allocated sales team head sees the number of leads for each team conversion rates, closure VP sales sees pipeline for all the team members and CEO sees projected revenue versus actions. All of this is derived from the same flat data but how they are sliced and diced for each role is very important because if you give the number of leads to the CEO he might not be that interested so when you are building a visualization even though you have the data first you have to understand your audience that role that department marry the role and department which forms a goal and that goal is equal to the metric. So there is a seven step process just before that so any metric which you are visualizing ensure that it drives the business and a dashboard or whatever you are doing which means it should be linked to a business outcome if it is not linked to a business outcome nobody really cares to see that data point because it is not helping him do his work better we should not be a vanity metric we all know what vanity metrics are and see just because you have some data does not mean you have to show it so personally I think the best visualization of planet earth would be one day when I wake up and see my mobile it shows me a red or a green green means everything in life is good red means something is wrong and only when it is red I tap on it to see what goes wrong think about it you do not have to worry about any visualization you do not have to worry about any data data is a means for data and data visualization is a means for you to get control over something if the control is so simple a red and a green and when it is green you do not worry about anything else and red is something when you click on it can you get simpler than that but to get to that simplicity it is a lot of work because then you have to figure out each NMP metric and how they tie it up into that single big metric to be able to give you red and green so just because you have some data does not mean you have to show it it is best to show the least amount of data which is linked to a business outcome and this is more of a seven step free work this is actually borrowed not mine so typically when you or when you go into organization and organization says we need to be data driven or we need to have a dashboard or we need to be driven by visualizations what happens is you just go and look at the data and you start turning on some dashboards and in the end you will have 10 different dashboards 3 for marketing 3 for sales 3 for the CEO and none of them converge to a point what is important is all of those data visualizations converge to a point by role and by department and to know what how to make them converge there are seven steps to it and these are the steps so first you define the company goal both short term and long run what is the company trying to do second what would it take to reach those goals which are your key performance indicators or key result indicators then what activity would you need to do to be able to reach those goals list all such activities then take up the top 20% because there is a 80-20 goal 20% of your activities will contribute 20% of reaching that take up those 20% find a responsible owner for each of those activities because he is the guy who is doing the activity monitoring it and tracking it he is solely responsible for that then you have to figure out how do you measure progress that it is going on so let's say when you are writing when you are doing a project how do you measure how close are you to completion is it based on number of lines of code, number of function points number of stories completed or whatever your metric is then one of those indicators you are going to divide it into three buttons which are already being measured and reported which can be measured and which cannot yet be measured because of data not being available or because you haven't figured out a way to quantitatively put it after this you will have a set of three or four metrics per role if you have more than four metrics it means we need to redo this process because at any point in time you cannot control more than four metrics four metrics means four different stream of activities which means you have to monitor those four things anything which goes wrong with that you have to go and fix it and get that metric back okay so if you see a dashboard with 50 different charts and metrics yes it may look cool it might look like a lot of work it's not the best dashboard is which can fit on your iPhone screen four very simple metrics tops not more than that now that the gap part is over so let's move to some part of visualization so okay I will quickly recap the thing we did three things here we knew we had to know our audience which is what we just covered the second is how do you choose a right visualization which is what we'll cover now so there's a whole lot of theory behind how human brain works and the human brain has much more power in terms of pattern recognition than your computer today think of it if you're in a room with 50 people the moment you enter you can recognize each of those 50 people today's computer cannot do that so 80% of our brain is visual function the rest is cognitive function the cognitive function gets kicked only when you force it to be active the visual function is always active you always are looking out and you're recognizing patterns something happens it triggers an alert in your brain that something has happened around you so when you're building data visualization it's very important to capture or leverage that 80% of the brain where we intend to maximize the pre-attentive processing pre-attentive processing is where you're not so think of it this when you're going out and having a cup of coffee suddenly a car comes towards you while you're still focused on your coffee but something happened there because of which an event or a trigger was raised in your mind that is pre-attentive processing you're not focused on that car you're not looking at that car but it triggered something when you're delivering a visualization part of knowing your audience is what is the medium of delivery is it the web is it a powerpoint presentation is it using it on its tablet and what is the skill set of the user is he an analyst a sea level your mom and dad or the mechanic and the design choice should support the comprehensive of data so that's what I discovered Edward Tuftly has how many of you know about of have read Edward Tuftly Stephen Pugh okay so Edward Tuftly is one of the I mean he has some seminal books on this he has a concept per data ink ratio which means what is the percentage of data versus ink on your paper and you have to increase that which means less of chart chunk or less of distractions and more value so that every pixel that you're using has a deliberate meaning to it and you're not just using it for fancy stuff or some user stuff so I'll just talk about this a visual function is extremely fast compared to the cognitive function 80% is visual function yeah so this is what I mean by pre-attentive processing when you're rendering a visualization it's always on two axes third perceived axis Z axis can be used for 3D but still it's a 3D on a 2D because computer screens are flat so we have another axis which we can use to showcase things which human brain can easily process let me give you an example the moment you look at this set of square what is the first thing that you look at the aberration here and you're not even thinking about it now if you look at a sequence everywhere where you're looking you're just looking at the aberration without even thinking about the aberration and this is because your brain is doing pre-attentive processing where it's not forcing itself to think on it but because there's something different it just catches your eye and goes into the visual function and then it processes it differently and these are things something which we will use which we use in data visualization to highlight or to point out a certain thing so the Z axis like in this the Z axis is change of color in this the Z axis is change of shape and this the Z axis is change of size we can use change of texture we can use a change of orientation where the square could be tilted by let's say 45 degrees and these things catch eye when you're building a data visualization so when you want to get the user to focus on something we use these tricks so one thing which at least based on our customer inputs we have figured out that we have 90 charts in our suite over 90% of our users use just 10 charts the rest of the 80 charts do not get used it's a very rare case that they get used and therefore very specific purposes because the world does not understand those charts if somebody has to put effort to understand the chart to be able to understand the data he might as well look at the data itself and not look at the chart the world is not for the statisticians so when you're making a which goes back to know your delivery medium if it's a one-on-one and you're talking to your stock broker you'll see you'll understand a candlestick chart but if you're showing the price of Apple for an average guy to understand why his stock prices have gone up with Apple or whatever it is he needs a line chart so here are your data visualization experts or your analysts or your domain experts who will understand any of the visualizations so if you're doing a one-on-one if you're doing a PowerPoint for these guys it's okay but if you're doing a mass story if you're doing it on the web for your consumer website or even for a business where you know that different roles will be involved you can't use any of the fancy charts it helps you have to stick to the basics something which people understand people adapt fast they learn fast but then they need to be guided so every new user who comes into your system to look at that visualization will need to be guided before you can actually build a visualization for that guy so fundamentally in business there are five types of basic representation either you have a single figure which is a KPI as we call it this can be represented as a gauge speedometer or whatever you wanted but this is the fundamental concept this is a single figure with historical context so you have a spark line which can also be a spark column and then you have an indicator of how it's going up or down then you have a comparison of data that's always a column or a bar chart so comparison is where you're not trying to find the trend you're seeing who's the highest and who's the lowest so never use a line chart for that it's always a column chart the reason being humans as humans we have a tendency to go left to right that is how we've been trained and we can track our line very well so if you give somebody a straight line he can tell you pretty much if it's a straight line or not if the painting is hung straight or not and most people are very good at it so when you go left to right your visual function automatically picks it up and says that this guy is the most when you have to see how a trend is changing a pattern is changing over time which is linear or sequential but connected as a set of events this chart is not good because then you have to first think where all the patterns are going here the pattern is already established for you so you can clearly say that it's the bottom most here and it's the top most here or relatively flat here and the third is composition so there are enough composition is when five items make a combined entry so let's say a number of employees across four cities that makes your total number of employees in a company so Python is one of the most used and abused chart in data visualization it is practically the worst chart to be used because if I ask him is this pie bigger or is this pie bigger there's no guarantee that you can tell me comparing both of them because we are not our brain is not prepared to comprehend angles and then compare angles we are not so good at comparing angles right in a pie chart alternate for a pie chart is actually a sorted column chart but still there are people who prefer a pie chart and they want a pie chart so especially when you're selling to clients you can't say no to a pie chart so I want to question on the bar chart your sense of vertical versus horizontal so vertical versus horizontal is actually not a question of sense it's about how big your data labels are if you look here all the data labels actually have to come in a straight line but each of your data labels come one data label for a line so if you have longer labels you can use a bar chart if you have shorter labels you can use a column chart that's a function of space I was asking about so human comparison I mean so there also you go in a straight line so when you're looking at a painting think of it on a wall you're looking at a painting if somebody tells you is this painting straight or not some people look at the top part of the painting some people look at the side part but both of those are straight lines and we can comprehend that pretty well it's good horizontal or vertical horizontal or vertical is a good try but people prefer this because it takes more horizontal space so if you look at any of the device apart from an iPhone we have more horizontal space on our screen as opposed to vertical space so and that's an interesting discussion altogether on a smart phone a whole lot of theories of data visualization are equal to PC and Mac does not apply that's a completely different paradigm today what we are seeing is most of those paradigms being transferred over to an iPhone and which is where there's a huge opportunity if anybody if any one of you are planning to build so we are already doing something in that space and we are kind of realizing the challenges but it's all completely different and so much fun's out there so we mentioned there could be innumerable types of data visualization I mean last count I had there are known 250 plus chart types and then if you go to let's say flowingdata or infoesthetics.com or let's say one of these data visualization that's CH every new set of data has a different kind of visualization so there's no there's no limit to what you can visualize how you can visualize but again remember that any complex visualization will lead to main knowledge and ability to understand that visualization to understand that data so that's two steps away from the data not one step away so do not build very complex animations or sorry complex visualizations unless you're absolutely sure about your audience so now what I'm going to do is show you some good and some bad visualizations some fallacies some mostly visuals again this is very simple business charts but the topics or the points which I'm trying to make can be extrapolated so if you want to stop me or just want me to skip fast just let me know so again communicate more than data to the user do not leave the processing to the user if the user has to assess the visualization he might as well look at the raw data this is an example okay so here this is the sales target and the sales achieved for it sales guy look at this and say tell me if this guy is going to get a bonus or is he going to get five and the last one in December I don't think he can get a bonus neither will he get a bonus so here what we're trying is comparing his actual numbers in a line chart with a zero axis which is the worst thing to do the first thing to improve the chart is not to have zero to fifteen thousand you have it from here to somewhere here which gives a wider scope of comparison this is white space here this is white space second is the chart is actually not the right thing we're trying to see did he perform on every single month which means we need a clear indicator generally did he perform which means what is the difference between this and this what is the difference between this and this does not perform this is a representation for that surely we're going to get five eight out of twelve months he did not come up to his mark so same data different chart and the thing is the way we've also plotted the data here we are plotting here we are plotting the actual revenue achieved by him in the target here we are plotting the difference of that data start difference and how you plot it so if I'm the salesman I'm going to use this chart but if I'm the manager I'm going to use this chart so back to the pie chart if your manager or your client says use of I have to use a pie chart and he has to use a pie chart for legit reasons and not to hide his data always use labels and give a legend to it now you can see this 23 and this 23 are same but if you look here this will look much bigger than this because of the way we comprehend angles at the front the visual function of that coming in does this look bigger to you than this yeah it's not both of them are 23 okay so okay but why is that like so the visual function so in a pie chart anything which is bigger to us in terms of angle perceived looks bigger to us because of the way our visual function works as opposed to something in the back if you look at even the perception of light poles the one in the front is the biggest because it's the closest to us the one in the back is smallest so that's the base so this is obviously going to be slightly different but if you have to actually hide some point that is very good even more make it a 3D pie chart yeah like isn't it a 2D one so how does the front back thing happen I thought it will happen for a 3D which are beautiful but the 3D one in a 3D pie chart it's actually bigger 3D is obviously bigger but 2D we have to ponder how it's closer to me and also the bottom one is you make it to be like how we perceive our tilted axis so if I flip the chart the black will seem less and the blue will seem more no so if you flip if you flip the chart the flip chart yeah so that's the way our brains perceive so I have to get more into research to know why that happens but I know basically that fundamentally when you look at I guess we can't comprehend I believe from where you are viewing it the top one must be looking larger because it also depends on the colors used that's all that's all so there's a very interesting most of you might have seen it take a red a gray circle same color right put that gray circle put behind that gray circle a rectangle of lighter gray and put behind that gray circle a rectangle of darker gray suddenly the same circle will look in different size because of the contrast around in the colors around and those are tricks which are used a lot in gaming and animation in pistol as well this is a pie chart this is a pie chart this is a simple trick this is a trick if you want to actually so this is actually a good replacement for a pie chart you do a sorted bar chart instead of a pie chart because then I can clearly see what is happening and another trick in pie chart is when you have a lot of so do not use pie chart appear more than six slices if it has more than six slices nobody can anyway this will start the way to do it is when you have more than six let's say seven eight nine plug them together three and put that as a single slice so that it's easy to compare cut down distractions on your chart so if this is your sales if you are selling let's say four million you are hardly bothered about your twenty three cents and fifty six cents so why you will tell that guy that this is twenty three cents and fifty six cents because the numbers here are huge and when it's four million it's hard to go four million five hundred thirty one at this point in time this guy is probably not even worried about that your chart looks so much more cleaner so this is very standard when you are doing anything for projections use dotted lines don't use the same line and write projected because the visual perception does not go to read projected only when you go and read it cognitively you read projected but anything which is dotted comes up as you know what this is different you can either use dotted or you can use different colors but different color is in an industry partner used for highlighting a column and effective I have also seen sometimes people using I don't know if it's right or wrong one is the dotted line and they use lighter colors I think the economics uses that like when they like and they have like an eight year data this first six would be like darker colors and then the other two would be projected and they would kind of fade out so just to show that visually that different so we have so there's again a theory behind this that when you use a dotted line some people cannot perceive dotted line and they see it as a straight line so there are color palettes which are color blind friendly so even if you use lighter and darker color people cannot actually perceive it but then there are palettes which are color friendly on the print so they have done their background on that part so it can be used but even then when it's a different color the perception of that depends on our English language when we put three ellipses that means what's after this I don't know that's like could we continue or could we not be there it's like a maybe kind of a thing so that is what people perceive it as in terms of dots in this example it could be ghosts how people show ghosts in films or comics it's always in a dotted or lighter palette which is to show that simple trick use annotations monthly revenue you know this since I joined this so here so what happens is then very clear you're making the guy first before looking at the chart he actually looks at it and we're building a data visualization so have you any one of you who has studied the paintings of Remrong so he uses a very interesting technique first he etches his painting and then he creates a painting with different light textures so when you look at a static painting the light is there on the painting it goes into an animated story you first look at one part of the painting then you look at another part then you look at another part and he is manipulating you through use of light on a black and white painting what happens is because your brain goes to the most active part and then it automatically figures out the visual function without even knowing that you're doing it you actually see the entire story on the painting that is something you can actually carry to a data visualization as well you can use colors you can use fonts you can use shading to make the journey of the audience on this chart the way you want it for example right now when you looked at this look at what happened my caption is actually smaller than this you did not look at the caption when you came to look at the chart you start looking at this or you start looking at the colors if I make this bigger you want to look at it if I can remove this to remove junk I can replace it to the right because then you want to look from left to right based on what I want you to focus in a data visualization you can do that it is a data manipulation that is another topic all together so you will go to the data manipulation part creator there is a book on it how to live with charts statistics statistics yeah that is enough so I will go to another article we have written tricks that don't teach you at business school lies, damn lies statistics yes so they are saying if you beat the data hard enough it will say anything you want it to interactive features so most of us are now doing development for web or mobile which means we can use interactive features the moment you have options for interactivity means you can cut down a static image on a print you don't have any other option because people cannot click but let's say here you have all these things you want to show these metadata what happened in October you can see it as a user when I go here I have too much junk I already don't know what to look at but in a chart like this I see that something happened in October which gave this rise so I just go and roll over it and I see additional details so unless the user really wants to see additional details don't force him to see all of that because then he might just not see the entire chart so your point of having a chart might be missed drill downs is a very common thing now in a chart the few rules or drill down are always start from macro data and go down to micro data have clear categories so like go from country to city to let's say within city around these minutes like smaller sections don't do arbitrary jumps and when you're drilling down second when you're drilling down use the same chart type always here we have this is actually error and this is the wrong thing which I wanted to show that if that is a column chart and you're drilling down from a column chart at all levels it should be a column chart it should not change to any other chart type on a drill down chart which allows drill down tell people that you can drill down have a visual clue so here let's say we have written a click on a column still it's not the best way because it does not appeal to the visual function so you want to have something on top of that chart or a visual representation that you know what come and click on this chart very important otherwise people do not expect drill down on a chart because they're not used to it as much today column chart so small things even when we do tables on a daily basis tables can be enhanced massively when you're presenting a simple table like this with data it's very hard for people to understand or look at part of the table but the moment this is kind of a similar to a heat map the attention goes here what happened here why is it so dark and then you over track and then you can even add drill downs to this part another modification of this is this part not just color but also the size which means it can handle multivariate data so the table is also data visualization by the way so how you can enhance those tables are limitless there are tables which has there are tables which have bubbles in it as well so very effective way to show large amount of data and then have a click action or a handler action or even hover actions on top of each of these cells to link it to something else isn't the other one actually like a lot of visual overload so on this screen you're seeing it when you actually see it on a PC because the colors are not reproducing property so this is a very light color palette when you actually look at it so you know on your screen it looks kind of hectic here actually when you put it up in a dashboard because of space because these colors are indicated with something else the entire dashboard is color driven so these colors represent each of these colors represent a state and there are multiple other charts which also represent that state in this heat map sort of visualization I mean so you were saying that this could be used for a multi variables sort of you know I mean data where you are also playing with the size right so there are two ways to do it so multi media data bubble chart this is three access x y and z there could be a fourth axis also based on color which is so this is your x y z is the size and then you have a z one which is the color so intensity of the color like what you see in a cool fusion maps table or this is another one for multi media data where color is already there so you have an x y value and color for example there is a vanilla motion that just like we are not good at comparing angles but comparing areas so for example if I ask you that circle in that circle what is the ratio correct so bubble chart is never for comparing individual data it's for comparing clusters of data so you look at the cluster and say so let's say this is this is cost per service and stickiness so let's say I'm a services company and I've got multiple services this is my cost of service and this is stickiness so I can see where am I doing where am I going right or where am I going wrong take this example as investment chart duration of investment duration of investment returns and the size of your bubble is how much money you have put in that investment so if you see that the duration of investment things which fall here which have where you've been invested for long but it has not given you return that cluster is something which you need to work on whereas things which are here means you are doing phenomenal whereas things which are here you are doing pretty much okay basically what you are saying is that the most important part of the the axis should not be in the size right because for example if you are caring about where am I investing most then this is a bad chart because you won't be able to visually compare so by just being careful about selection so position can be indicated relative size probably but so the most important data points here because you are looking at the cluster and you are looking at where it is placed the size is important the size once you have figured out this is the cluster I want to work on then you individually look at it the most important point out is like at least from personal experience we are looking at something like ggplot kind of which has like 6 5 or 6 even variables like color size everything but if on that if you have more than the one as you pointed out is the gradient like you know x and y and then the other things like color is probably the third that is important so that gradation is that and if you have more than three variables then it becomes extremely confusing for any user you have to go to the user and explain every time you know what that chart represents and there is a cognitive overload if there are more than three parameters okay so human mind cannot interpret more than two parameters if you give him on a table I will give you a table of x, y and size you can relate between x and y but in a tabular format also between x, y and size you cannot visualization helps you do x, y and size if you add a fourth parameter which is color that is a lot of cognitive overload you can still add two different tricks to add to the parameters let's say you can add a bar chart inside of this you can add another bubble inside of this you can do a hover over it to release another chart you can have spiky hairs on top of it to represent another thing altogether you can have a border around this the strength of the border gives you the gives you another parameter then I get that but what I am saying is that the more information you add in the chart then the one thing is as I said the visual part kicks in that's when your cognitive comes into the picture the whole point of having a chart is that you should not have to think twice or interpret it which is what I found from personal experience that if you have more than three parameters then it becomes that is when your brain starts interpreting this is unscientific anecdotal but absolutely this is scientific I mean on the lighter side of things when you are making a product for a popular and you are selling it cognitive overload can actually increase the price of the product you've seen that happen a lot so I can't understand it so it must be good so that's happening right if you have to pick up any of the Mackenzie report and look at their charts it just takes like insane amount of time even for me to understand what the hell are they trying to show and they'll put a stack chart not as one chart they'll put six different columns as six different charts and they'll put a column on the side of it showing projected and you just cannot understand then you think that yeah it's nicely presented it has so much of data should be right so to solve your problem actually Google motion charts has done something very interesting so what they do is you have a droplet of parameters here and you have a droplet of parameters here you can have twenty different parameters so let's say you are combining N parameters into N parameters so that's a matrix of N into N you select one from there you select one from here and then they also have something called motion so let's say this is for January how this changes in February and March you have a play button which smooths out and then you see we started by hands closing yeah I saw that actually we eventually gravitated towards that when we kind of saw this kind of happening especially the time always we had that for example we had like one of the matrix was CTR and fill ratio we saw it how it actually changed over like months of the year so yeah we actually eventually came to motion charts because without doing the animation kind of added another layer without adding like cognitive overload it was very easy to represent so the animation the step to cooperate that chart is you first look at the starting bubble chart you figure out the cluster you want to analyze then from the cluster you figure out one variable whose motion you want to see because we cannot track motion of two independent objects at the same time because we have to focus at one center point so the step is still the same they are doing it through animation using a static chart you can tell the user you know what click on this thing to just see the animation of that particular variable over time because if there is a whole lot of cognitive load other thing other observation was in the motion chart just computer conversation was in some cases you can use the trail so the trail actually can be useful because even once the animation is done still see the trend because the size keeps on varying so if you use the trail it can actually make sense for example revenue and maybe on two axis and then is it increasing is it decreasing so you can see it in the motion but if you leave the trail in some cases even that was like quite useful correct but then again with the caveat that there is alpha transparency as you was pointing out and there are not too many variables maybe like 3 or 4 but if they are like 6 or 7 it doesn't make sense you don't want to leave trail of only one the reason being because if the path enters across for two variables then it becomes very complicated and if three of them go so what happens even with the alpha variables go one on top of each because of the alpha the color actually changes so blue plus yellow is equal blue plus yellow is equal to green sure so the other one is heat map this is actually very effective and very underused chart this heat map if you actually were representing column chart this would be either 6 or 12 column charts you can compress 6 or 12 column charts across a multivariate data into this so we have a nice example of how to compare restaurant listings on various parameters from various people in a single chart as opposed to 50 different charts this is inspired by Stephen Pugh these are condense charts so Sparkline's was by Edward Tufty and bulletin graphs are by Stephen Pugh this is basically a singular data point with a target value and a range so you say like this was my revenue target bad average good this was my actual target and this is what I achieved so in a small space like this you can actually achieve so much information and it looks pretty neat so here you can see we are actually measuring 10 kph we are seeing historical context of 3 charts and here we are seeing stock price of 4 charts that is a lot of information for one dashboard this is just an example to show that how thinking beyond your standard chart and this is a chart which most people would very easily comprehend it's not very complex to understand there are the idea was to show how much data point we can fit in in a single screen using such charts so are you not I mean the tox is for for any road if it has every visualization has to drive you business outcome if it's not driving in business outcome and you're not owner of that and responsible for that you should not be seeing that visualization perhaps also relevant is that the difference between a dashboard and a visualization or your graph depending on your presentation is in a presentation you want instantly to get that in a dashboard you have probably already comprehended what the contents are and you are looking at changes over time or you know familiar with the dashboard already so that's why the you can potentially fit more into a dashboard but if this were to be flashed like it has been here it would be very difficult to complete so we didn't put as much as you want but the point is would you be looking at that and taking a call so let's say as a CEO of future charts I would be happy to look at the number of leads what can I do by looking at the number of leads it's not in my hand it's in the hands of my director of sales that's his ownership and the fact that if I look at leads that it means I'm not delegating it so I should not be looking at it my key metric is what is my total revenue what was promised by him what did he achieve if there's a disparity in that then I need to go and speak to him I mean it's such a completely accurate because we have built business at work so the first question which typically was thrown at us is what is the call to action so that it has for example this is the chart but it has like an indicator saying green or red what is the call to action about what is there otherwise it just becomes like a junk chart because it is information but especially when you're doing a dashboard but the first question anybody will ask is how is it relevant to me and then what is the call to action or what actionable insight that I get out of it so that actionable insight is like clearly the first question that any business guy will ask you this is an example of building a dashboard so I talked about a paradigm change right now till now we are building dashboards for this but the world will be on smartphones this will become your truck of the day and people will be driving their saloon cars today we are building for the truck tomorrow we will be building for the saloon cars how do you so there has still to be a link between that part and this part how do you make the experience unified so that your dashboard or data visualization or PC or Mac looks similar to what it is on an iPhone but the paradigm is different so if you look at some of the changes here this is sales and quantity trend there is a KPA here which is important which we have shown here KPA right there that means this is something which I want to look right away without another level of passion I might not be interested in this graph if I am interested it is one kick away this thing here average orders to deliver sorry this is weekly sales average days to deliver order this is two now a gauge here looks good because it is a filler whereas on a phone the most complex data visualization on phones have to be very simple numbers or as simple graphics if you look at so some of the good dashboards on mobile phones are Fitbits dashboard Fitbit is a health band like job on even job on is pretty good then you have a couple of financial applications where most of the dashboards are not built on column charts or line charts it is a simple percentage indicator they have a round thing showing how much of that is complete because when you are looking at your KPA it is on the move you are only interested if you need to get history on that then you do a tap on it and then you go for it so this is a different paradigm so at the cost of using the word responsive have you been able to do something which is which works across a PC an iPad and a phone so this is the concept of the exponential of responses is the same thing right or is it a different code being it is absolutely different with a different on a response the whole idea that you don't build separately for a mobile phone and for a screen like java right as a programmer you program in java we ensured I mean we would have ensured as a micro system it would run everywhere because then we build the aisles at the bottom of it so that it runs everywhere here it is not that complex but there is a whole lot of architecture behind it so that when you build something called a PC that is a different architecture when you build something mobile so think of things here we do not have to worry about latency here we have to worry about latency how many times the server gets ahead what do we have to cache because people are used to the app experience they need a tactile feedback right there that I am going to go load the data whereas on the PC and the browser people can see the status bar and they see that something is happening so the whole US changes, the architecture changes the data packets change because you do not want to do too many server hits then let us not get into the Android and iOS and everything I am just curious whether you are doing a single solution which can be seen across the device or do you make something catering to each kind of platform which is a solution for single mix but people implement it yeah so like implement it separately for mobile and so on how come HTML 5 works isn't free while the flash is flash also the previous version is free so any plans of making it a better pool paper maybe a barber support and that is bad job this is slightly outside sketchy this is something that a lot of people need in data realisation so we built a dashboard of patient monitoring system of the iPad for doctors now doctors are used to seeing their ECGs and they are used to seeing their devices in real world suddenly you give them very flat and very minimalistic look they were like what the hell happened this doesn't look real my patient will die you guys are not doing the right thing so we built something which exactly stimulates the patient monitoring system in real life so you can see the ECGs here 350 milliseconds of sorts these are actual thermometers and very realistic looking widgets any data visualisation guru will tell you avoid using this this is Q1.5 design where we are replicating things but there will always be a section of audience whom you cannot make jump from the current thing to the absolute modern thing this is the transition period and yes there is a whole lot of this big market which still needs this so we do it on HTTP but it can use a persistent connection we can even use a website so what we also do is we give you client side access to it if on the client side you can open a socket to your back end or you want to use a local app so you can either control at the server side or at the client side the choice is yours