image_path
stringlengths 19
21
| description
stringlengths 0
1.58k
|
|---|---|
images/image_0.jpeg | The graphic adds no value. To get any sort of information from this you need to read the (rather dense) text associated with each graphic. It’s another case of a graphic designer trying to be too clever and destroying the information in the process. If the use of the same shapes in different sizes for different bars is bad enough, here the designer has tried to use different shapes for each city. So there is no way at all that the size of any building says anything about the quantity it is trying to depict.Also there are two quantities being represented here (runs and average), and it’s not clear what the bar is trying to show.Now I’m starting to wonder if this is intended to be a graphic at all. Source: https://www.espncricinfo.com/story/_/id/27358389/which-top-cricket-city-win-world-cup |
images/image_1.jpeg | This is from Britannia’s investor presentation, sent to me by reader Varun Shenoy.Graph is all good, but where are the numbers? It shows that market share has increased, but without telling you what the number has increased from and what it has moved to, it’s incredibly hard to trust the data. Also - saying “largest competitor” obscures things a bit. It seems like there’s more than one competitor, and different competitors are the largest competitors at different points in time. Presenting each of the major competitors’ market shares as a line (with numbers / axes ) would have enhanced the graph! |
images/image_2.jpeg | Pie charts are bad, but they are at least okay if you’re showing parts of something that add up to 1. I really don’t know what the graphics designer here was thinking - maybe following a “rule” that if percentages need to be shown, a pie needs to be used. Needless to say, the data doesn’t add up to 100%. The sectors of the pie don’t represent mutually exclusive and collectively exhaustive parts of anything. This is absolute nonsense. Source: https://www.espncricinfo.com/story/_/id/27143430/kane-williamson-hand-steadies-new-zealand-ship |
images/image_3.jpeg | I spent a few minutes trying to figure out what the Y axis is here, and so far have no answers. Why would you distribute the points across two axes if the information is contained only in one? Source: https://wellcome.ac.uk/sites/default/files/wellcome-global-monitor-2018.pdf |
images/image_4.jpeg | This is a screengrab from Star Sports. This “graphic” just overwhelms us with numbers without giving us any information. I mean - what is the relevant data point here and what message is the broadcaster trying to convey? Couldn’t they have simply picked the relevant measure (either number of sixes or number of sixes per game) and done a simple line plot to show the trend? Also, issues with the data quality exist - there was no World Cup in 1995, bu there was one in 1996. HT: Siddhartha Vaidyanathan |
images/image_5.jpeg | This is a great example of “data mining”, where you look for some pattern that fits the pattern that you’re seeing now, so that you can extrapolate to make some random prediction. And if you look at the two lines that have been stretched and overlaid to make it seem like they overlap, they haven’t even done a good job of that - it’s hard to see how the two lines are similar. Now leave the bad data aside and let’s stick to the visualisation - What’s with the background? The shadows of the lines? The overall UX?The selective axes labels (0-10-30-50 on the left) and so on. Source: https://www.cnbc.com/2019/06/12/tesla-looks-like-netflix-did-in-2011and-it-may-see-a-similar-recovery.html |
images/image_6.jpeg | The representation using a choropleth isn’t bad, but the reason I’m putting this here is the choice of intervals. I understand it is the default algo of datawrapper, but why would you draw boundaries at 15.4% and 33.5%? There is no regularity to this.And when you use intervals like this, the choice of intervals can make a huge difference to the message in the graph. A better way would be to use a continuous scale, and do away with the intervals altogether. That would take out the arbitrariness, and help readers compare states betterSource: https://www.livemint.com/news/india/india-s-workforce-is-masculinising-rapidly-1560150389726.html |
images/image_7.jpeg | Cricinfo at it again. When will people learn that using non-standard figures in place of bars actually hides information? Minor points:1. The scale on the left is useless2. Not clear what the number in the red circle is. 3. Each batsman is playing a different shot. Again distracts from the graphSource: http://www.espncricinfo.com/story/_/id/26927142/how-india-beat-australia |
images/image_8.jpeg | 1. Why would you stack bars with quantities that have different units? 2. In the light of 1, this is minor but, why would you choose near-identical colours two quantities you’re stacking? Whose father what goes in picking contrasting colours? |
images/image_9.jpeg | If we could see past the colour scheme, this might be a good graph, but the choice of colour scheme makes it impossible to understand. Also the graph itself is not clear on what the Y axis is, and we need to perhaps turn to the accompanying article to figure that. The graph itself is simple - showing some measure of voting among men and women in India across time. But that atrocious colour scheme means it’s virtually impossible to see what’s happening. What prevents people from using simple graphs? |
images/image_10.jpeg | Check out the X axis. It’s really hard to find a pattern in terms of the scale that has been used. 1 and 1600 are as far apart as 1600 and 1913, and 1935 and 1953. It’s insanityI know what has happened - some bright analyst at BankAm Merrill Lynch made this in Excel. And instead of making an area chart (which is what it should be) they made some sort of a bar graph. So all the X axis data points they had got spaced evenly. And so resulted in absurdities like you see on this graph. Source: Bank of America Merrill Lynch |
images/image_11.jpeg | 1. As a rule, the Y axis of a bar graph needs to start at zero. This violates that. Makes it seem like Indian women are one-fifth the size of Latvian women or something. 2. Why are the figures scaled on the X-axis as well? Adds no information3. Generally not a fan of using random figures being used in place of bars. what’s wrong with bars?Source: Several people on twitter sent this to me |
images/image_12.jpeg | This is not all. You need to click through to this article on BBC, where you will find a few other such pictures. They use five pictures to convey the sort of information that one simple bar graph could have shown. And in dispersing this information across five graphs, they make it absolutely impossible to compare across these graphs.Source: http://www.bbc.com/capital/story/20170530-the-avocado-toast-index-how-many-breakfasts-to-buy-a-house?ocid=ww.social.link.twitter |
images/image_13.jpeg | Too many series on one plot here. And I don’t know why we have a mix of lines and bars - simple two-axis lines would’ve been good enough.Also, the extra set of green bars to show growth is a bit confusing, especially since the high growth period happens at a time when the base was really really low. And then the X axis labelling is atrocious (I guess this is Excel default). You don’t need to label each tick. I would possibly draw this graph with a logarithmic y axis. and dispense with all the bars. Source: https://twitter.com/MohapatraHemant/status/1120212609045655552 |
images/image_14.jpeg | The problem is with the left set of graphs. When you see a set of bars, you assume they all represent the same “property”, with one bar for each category. Here, while all bars have the same dimensions, they all represent different quantities, and so this representation is misleading.A sort of waterfall chart might have been better. The ordering of the bars doesn’t make sense as well - loss is defined as revenue minus expenses. So putting the loss bar in the middle doesn’t make sense. Also, expenses needs to be showed below the axis since it’s a negative number. I can go on but will stopSource: https://the-ken.com/story/grofers-bet-on-value-customers/ (paywalled) |
images/image_15.jpeg | I fail to understand why graphics designers still continue to use non-standard bars for their bar graphs. Without those legend numbers, it would be very hard to determine, for example, that de Villiers’s number is a bit more than twice of Russell’s. source: http://www.espncricinfo.com/story/_/id/26574308/no-complaints-andre-russell-batting-position-dinesh-karthik |
images/image_16.jpeg | The best description of this graphic, which has popped up on twitter multiple times today, is this tweet |
images/image_17.jpeg | It’s incredibly hard to say what is wrong with this chart, because it’s impossibly hard to say what’s right with it. Basically you use a pie chart (or a doughnut chart or a speedometer chart) only when quantities are mutually exclusive and collectively exhaustive. Here the data seems to be neither mutually exclusive nor collectively exhaustive. I don’t think it’s worth my time to comment further on this. Source: https://www.indiatoday.in/amp/elections/lok-sabha-2019/story/political-stock-exchange-find-out-narendra-modi-chances-of-second-term-lok-sabha-elections-2019-1497239-2019-04-09 |
images/image_18.jpeg | The X axis of this graph is misleading. January 2018 is not equidistant from May 2014 and March 2019. Seems like whoever made the chart assumed the date to be a categorical variable, because of which it appears this way. Might be a small mistake, but can have a huge bearing on communicating the data as it is. Source: https://twitter.com/MilanV/status/1114950062877155329 |
images/image_19.jpeg | 100 posts! |
images/image_20.jpeg | I don’t know what is the point of the Pizza represetation. Doesn’t add any value. And I can read only three names - Tech Mahindra, Infosys and Wipro. Really don’t know what the point of the big circles are. I would just simply use a bar or stacked bar, with some kind of overlay (colour or pattern) to show the country of origin of the company, if that’s something the designer seeks to highlight. in this form, the graphic is unusable. Source: https://observablehq.com/@mbostock/2019-h-1b-employers |
images/image_21.jpeg | It’s unreadable. Source: https://twitter.com/TGreenback/status/1108225060354035713 |
images/image_22.jpeg | My former newspaper is at it again. Pie charts are bad enough, but at least you should do them properly? The whole purpose of showing pie charts is to show proportions. So what’s the use of a pie chart that adds up to 33.52% (and that’s two significant digits too many). And wouldn’t it have helped to mention the shareholders (and percentage holdings) of shareholders who are clearly in favour of the hostile takeover? I would do this in the form of two stacked bars - one showing people in favour of the hostile takeover, and one showing people clearly against (I’m guessing there’s a large portion of retail and institutional shareholders whose views aren’t known). The total heights of these bars will also tell you who is winning. Source: https://www.livemint.com/companies/news/l-t-declares-its-love-but-mindtree-unmoved-1553016618238.html |
images/image_23.jpeg | Choropleth is not the best way to view this information. For starters, the legend is unreadable, so you don’t know which language is dominant where. Even if it were, for more than a handful of categories, choropleth doesn’t make sense. The information gets “lost”. I’d possibly make small multiples of choropleths showing dominant regions for each dialect. Then, there seems to be some missing data (which isn’t very apparent due to the black background), which isn’t explained. Finally, remember the old problem with choropleths - population distribution is not even. So colour covering big area doesn’t necessarily mean a large language. Source: https://twitter.com/ShamikaRavi/status/1038818007483645952 |
images/image_24.jpeg | Circles are hard enough to make sense of for the human eye. Why would someone make clouded circles? This is ideal for a slope graph, or even small multiples, where each country is shown through two data points - current GDP (PPP) and projected GDP. And ordered by projected GDP.Here, countries are hard to find, numbers are hard to read, and the graphics confusingSource: https://twitter.com/ehtoglu/status/1086633483550765056 |
images/image_25.jpeg | Concept is great but atrocious presentation because of trying to cram in too much data. As a tweet by the creator mentions, this data also includes the country (colour) and the player’s batting average (size of the ball). And the 3-d presentation of the balls and the black background mean that it’s hard to read anything in the graph at all! How would I improve it? Basically remove to improve. White background. Points + text for each player. Ignore country and average information. Get rid of fancy balls, etc. Source: https://twitter.com/guerillasa/status/1102960934157467648?s=12 |
images/image_26.jpeg | Regular readers of this tumblr may not require much commentary to know why this doesn’t work. It’s incredibly hard to get any information out of it. And it’s not even clear if the sectors in the two pies are the same or different! To recap, human eyes perceive lengths better than areas, so pie charts are not great. If you have up to three categories it is still excuseable but not when you have so many. How would I represent this? Slopegraph is the best option. Not sure of source. Found it on twiter. |
images/image_27.jpeg | Pie charts are bad enough. Half pies/doughnuts are worse - unless they’re showing parliamentary majorities. This one does an atrocious job of labelling. First of all you have the “tennis problem” where you need to keep looking back and forth from the chart to the legend to see what each segment represents. Then even the number are misplaced. I really don’t understand why such long and wonky lines had to be drawn to place the numbers. And they’re all so cluttered up. I would’ve done this as a bar chart, but if a pie/doughnut/half-whatever was mandated, I would’ve still used direct labels on the chart rather than resorting to a separate legend. And the numbers have too many significant digits - they could be rounded off to the nearest percent without losing much information.Source not clearly known. But I’ve seen this shared on social media by people from Zerodha and some other FinTechs |
images/image_28.jpeg | The human eye has a hard enough time appreciating regular areas (which is why pie charts aren’t great visualisation). So you can only imagine how hard it should be to appreciate a chart overlaid on an irregular area such as a map of a country. I mean it’s pretty impossible to get any information out of this map. The boundaries and shapes of the curves are rather arbitrary. The layout is extremely unintuitive (like why is Telugu placed in the extreme north?). Seems like some graphic designer learnt a new technique and wanted to show it off - that’s the best explanation I can offer for this. Alternative? Simply use a bar chart! With bars sorted in decreasing order of height. Source: MoneyControl, News18https://twitter.com/moneycontrolcom/status/1098469230498918400 |
images/image_29.jpeg | Really, what was the graphic designer even thinking? Each of the bars in this chart represents a different kind of quantity, so I have no clue what all of them are doing on the same graph! And look at the scale - how suddenly a linear scale goes to a seemingly logarithmic scale. Or maybe not even that - the distance between 250 and 300 is same as between 300 and 600 and then between 600 and 2500. Maybe I should stop over-analysing it. Source: Times of India. Via Bhaskar Dasgupta on LinkedIn |
images/image_30.jpeg | This is not a visualisation in the strict sense, but the way the data has been presented makes it incredibly hard to read and understand. The vertical format means it’s impossible to compare between 2017 and 2018.Something like a slope graph would have been useful in showing how apps have gone up and down the ranking list over the year. Basically 2017 and 2018 side by side, showing the list like it’s shown here, but with “match the following lines”. In fact this article is full of “datapukes” that are very hard to get information from. Source: https://factordaily.com/the-chinese-takeover-of-indian-app-ecosystem/ |
images/image_31.jpeg | This is mainly here for documentation purposes, and for “the sake of completeness”I don’t think I need to really explain what’s wrong with this chart. A simple bar graph showing top 10 would’ve done a significatnly superior job. Source: https://twitter.com/captn3m0/status/1077571211767345156 |
images/image_32.jpeg | This tumblr has come a full circle, as this bad visualisation appears in Mint in an article under my own name! As you know, pie charts (and doughnut charts, and half-doughnut charts like this one) don’t convey information properly because the human eye can’t perceive areas as well as it can perceive lengths. Moreover, you have the constant “tennis neck movement” between the chart and the legend, to understand which sector belongs to which party. That said, such half-doughnuts are a rather common practice in showing electoral data, since the half-doughnuts in some way represent the shape of the parliament house. The important difference, however, is that in conventional parliamentary charts, you build the two opposing alliances from opposite sides, and the middle of the graph represents the “halfway mark”.Here, by putting the two main rival parties on the same side, that representation is also rendered meaningless. Admittedly, I didn’t supply them information on which party belongs to which alliance, but had I known that the ultimate graphic would come in this manner, I would have added that information, and we would’ve possibly had a good half-doughnut chart! Source: https://www.livemint.com/Politics/3b1jc0XjgzQ2E0NE5T74HP/How-Lok-Sabha-will-look-if-people-vote-as-they-did-in-assemb.html |
images/image_33.jpeg | What’s this abomination of a chart from The Economist? (Frankly if I had to expect better from one publication, it would be this one)I mean - how are you supposed to look at that “protractor” and get any information out of it? So many lines overlap it’s hard to distinguish between them. And it’s hard to make out correlation between distance and fare change as well.A simple marked scatterplot would’ve done so much better! Source: https://amp.economist.com/graphic-detail/2018/12/08/why-ticket-prices-on-long-haul-flights-have-plummeted?__twitter_impression=true |
images/image_34.jpeg | The graph colouring scheme is, er, a bit academic? At first I couldn’t really make out that there were three different kinds of points in this scatter plot (and I’m not colour blind), and even after I figured it out (after seeing the legend), it was hard to make out any patterns in the data. Reading through the attached article gives you the context to make sense out of this chart, but ideally you would want these graphs to stand on their own. And it is not hard to show that there is a regime change after the year 2000 when there was a regulatory regime change. Or maybe I still haven’t understood it.. Source: https://www.motherjones.com/kevin-drum/2018/11/chart-of-the-decade-why-you-shouldnt-trust-every-scientific-study-you-see/ |
images/image_35.jpeg | Graph is pretty much unreadable. First you need to contort your neck to read sideways (lose significant digits and you can put the labels straight). Then 2017 appears before 2016, which is bad form (and you need to go to the legend find that anyways). Some kind of a slope graph would’ve made so much better sense here. Or even just a simple table (did you notice that the original article doesn’t even carry a table?) Source: https://dqydj.com/household-income-percentile-calculator/ |
images/image_36.jpeg | Why do we need a 2 dimensional graph here? How can we figure from this graph that “nationalism” is “29.9%” (too many significant digits, btw)? A simple one-dimensional stacked bar would’ve done the trick much better (Remember that the human eye is great at perceiving lengths, but not at perceiving areas)Source: BBC, via Twitter |
images/image_37.jpeg | OK this is an old post but I found it just today via twitter. Pie charts are bad enough, where did people get the idea of using concentric circles to depict data? In such cases, is the information encoded in the radius or the area? And the clearest indicator that the visualisation isn’t working is indicated by the numbers (rather inconveniently placed), which are mentioned to too many significant digits as well. Source: https://www.huffingtonpost.in/2016/06/14/how-india-eats_n_10434374.html |
images/image_38.jpeg | This is from an article that says “six categories of data scientists”. Maybe we should add one more - “those that don’t know how to represent data”? Source: https://www.datasciencecentral.com/profiles/blogs/six-categories-of-data-scientists |
images/image_39.jpeg | This is an interactive visualisation, so click through to the site. It’s supposed to represent length of discussions that led to deletion of wikipedia entries. There is an explanation on how to read the visualisation (right below it), but despite it, it’s impossible to get any information out of it. First of all, that an explanation is required is implicit admission that the visualisation is not intuitive. Secondly, even when such an explanation doesn’t help, you know that something is really wrong. I mean the concept is fine - trying to represent how editing discussions went before an article was deleted is fantastic. Just that the way they’ve represented the data means you can’t get any information out of it. Oh, and did I mention that the graphic is interactive? As if things couldn’t get any worse.. PS: the “how to read” graph talks about green and red lines. I’m not colour blind, but find it incredibly hard to find two different colours of lines in this graphSource: http://notabilia.net |
images/image_40.jpeg | This is an interactive graphic so I encourage you to click through to the original. Hover over the graphs and the numbers will show up, and you will see that they are not at all correlated with what appears in the graph. In fact, there is nothing to even tell you which line represents average and which one is the strike rate! This should’ve been an easy graph - simply have two Y axes, one for average and one for strike rate. And colour and label them appropriately so we know what represents what. And the mouseover data is not necessary any more. Source: http://www.rediff.com/cricket/2018/nov/01dhoni-performance.html |
images/image_41.jpeg | This is one case where a dodged bar chart would’ve actually made sense, since we’re comparing within group (support for each party among “natives” and “immigrants”).With the bars facing away from each other, there’s no way we can compare, and the two halves look pretty identical. Source: https://twitter.com/Yascha_Mounk/status/1056667980422152192 |
images/image_42.jpeg | I had to compress the image by a factor of 10 in each dimension to make it fit on this page. So this means you just keep scrolling and scrolling down the website before you are able to appreciate this. And when you get to this, you’re confronted with a three-dimensional figure! Two dimensional areas are hard enough for the eye to measure, so I don’t know what the designer of this chart was doing with three dimensions. It’s incredibly hard to get a feel for the relative sizes, A simple (one-dimensional) stacked bar graph, with humans being represented by an impossibly thin line would’ve done much better. It’s not like we’re able to easily measure the ratio of sizes in this anywaySource: https://www.vox.com/science-and-health/2018/5/29/17386112/all-life-on-earth-chart-weight-plants-animals-pnas |
images/image_43.jpeg | Four sets of bars are horizontal and the ohter one is vertical. Bars aren’t ordered in any scientific manner. The category labels have been banished to a legend, necessitating constant back-and-forth head movement to read the graphs. There is too much jargon which most vanilla readers of the newspaper won’t understand - there might be an attempt in the text to explain that but for that you need to get past this graph. There is no real way you can connect the heading to the information in the graph.Source: https://www.livemint.com/Money/4oVKgqr7DLC5AiRrBEVFmO/How-HDFC-Life-SBI-Life-ICICI-Prudential-Life-Q2-results.html |
images/image_44.jpeg | I’ve long maintained that India’s ruling party has a graphics problem. I mean there’s an art to using numbers to drive narrative, but having bars that don’t represent the data in question is not the way to do this.Check the last bar. The number is 2.08, and the bar is so much higher than the first three. It’s almost like the designers of this took some readymade growth graph, and then slapped on their own numbers on to this.And then you have the multiple bars (orange and grey) which only seek to obscure the data, and possibly take your mind away from the dishonesty. AtrociousSource: https://twitter.com/BJP4India/status/1054302026249646082 |
images/image_45.jpeg | I know this is Bloomberg’s standard format, but black backgrounds make graphs hard to read. And there’s something else about this bar chart that somehow makes it unreadable for me. Maybe the way the labels are placed far away from the graph? Or the font, or font size? And percentage increase is fine, but showing start and end points would’ve given far more information. A better way to have shown this information would’ve been through a slope graph. That shows start and end points on two separate scales, and then a line joining them - with the steepness of the line indicating the extent of increase. Source: https://t.co/0xzTS6OeoW (Bloomberg) |
images/image_46.jpeg | Three-dimensional bar charts plotted in 3 dimensions! And to add to that there’s a colour scheme and a legend. I don’t know how you can get any information out of this. Scaling means most of the data points can’t be seen at all (right half of the graph). And while I got this off twitter, and without context, this graph or labels don’t ell about what it’s measuring, whcih makes the data a bit suspect as well. Notice that one of the axes in this 3-dimensional plot is a time series - months of the year. So a line graph would’ve been appropriate here. And since there might be too many lines, the author should’ve considered small multiples. Source: https://twitter.com/AustenAllred/status/1054094811542708225 |
images/image_47.jpeg | These graphs are so atrocious I don’t think they need any commentary. Good luck to anyone who manages to get any information out of them.In fact, if you see the article, you find that a significant portion of it goes in explaining how to interpret this graph. That in my opinion is simple admission that these are bad graphs. Source: https://web.northeastern.edu/naturalizing-immigration-dataviz/ |
images/image_48.jpeg | These graphs are so atrocious I don’t think they need any commentary. Good luck to anyone who manages to get any information out of them.In fact, if you see the article, you find that a significant portion of it goes in explaining how to interpret this graph. That in my opinion is simple admission that these are bad graphs. Source: https://web.northeastern.edu/naturalizing-immigration-dataviz/ |
images/image_49.jpeg | Usage of two pie charts to show a before-after scenario completely obscures the data. Apart from the fact that colour scheme is inconsistent and pie sectors aren’t ordered properly (”Others” should be at the end in the top graph), usage of pie charts is simply incorrect in this case.A slope graph would make so much better sense here, showing how marketshares of each operator has changed over time. We might need some creative representation, such as clubbing (the yet unmerged) Vodafone and Idea in the first graph, and adding a zero data point for Jio for the first graph. But that will show changes very effectively. Source: https://www.livemint.com/Companies/rPju0LUcHUizRBbXH7ecdL/How-Mukesh-Ambanis-Reliance-Jio-shook-up-Indias-telecom-ma.html |
images/image_50.jpeg | Usage of two pie charts to show a before-after scenario completely obscures the data. Apart from the fact that colour scheme is inconsistent and pie sectors aren’t ordered properly (”Others” should be at the end in the top graph), usage of pie charts is simply incorrect in this case.A slope graph would make so much better sense here, showing how marketshares of each operator has changed over time. We might need some creative representation, such as clubbing (the yet unmerged) Vodafone and Idea in the first graph, and adding a zero data point for Jio for the first graph. But that will show changes very effectively. Source: https://www.livemint.com/Companies/rPju0LUcHUizRBbXH7ecdL/How-Mukesh-Ambanis-Reliance-Jio-shook-up-Indias-telecom-ma.html |
images/image_51.jpeg | What is this even! 1. Impossible to find your country here unless it’s large. 2. Not only two dimensional, but also weird shapes. So can’t figure out anything in terms of the areas. 3. You start wondering what the position i,mplies. Takes a while to realise it doesn’t mean anything. 4. Can’t even figure out the relative sizes of continents here, leave alone countries. this chart is so atrociousA simple descending order bar chart would’ve communicated so much more. But then I’m guessing the graphics designer would’ve billed so much less. Source: http://www.visualcapitalist.com/80-trillion-world-economy-one-chart/ |
images/image_52.jpeg | It’s not easy to find the message conveyed by the headline in the graph. Why couldn’t they have simply used coloured bars? The emojis are incredibly hard to read. Also, given the message, i’m not sure sorting by Tests is a good metric. Maybe they should’ve ordered by a crude batting order instead? Source: https://www.livemint.com/Sports/ayUjhOpp6KoDvphRBGExnJ/Do-the-young-starters-in-Test-cricket-play-long.html |
images/image_53.jpeg | The bar graph on top is extremely weird. The two bars are supposed to show ranges, as the data labels indicate, but the bars themselves are of a precise length. To repeat - the information in a bar graph is in the length of the bar, so it’s not an appropriate measure here since they’re trying to show ranges.Moreover, since data from 2016 and 2018 is being compared, it’s basically a time series. So the way I would show this is by inverting axes, and then showing how the range moves. Actually - this one is not to scale (I quickly made it using Paintbrush), but this is possibly how I would show this information - using two lines for the upper and lower limits. Source: https://the-ken.com/story/delivery-personnel/ (Paywalled) |
images/image_54.jpeg | This can be interpreted as a “sankey diagram”, where the widths of edges between nodes can be used to indicate the volume of the flows. When I saw this, my first question was “why are so many people going from south east asia to siberia?”. Basically it isn’t intuitive at first that each continent is a node (maybe due to the usage of the map rather than simple dots or circles). Having figured that bit out, I guess the point of this map is to show how big the self-loops are (within continent migration. at least that’s what stands out to me). And I’m not sure a Sankey diagram is the best way to show data where the point is to show the self-loops. How I would represent it otherwise would depend on the message I’m trying to convey. One idea is to simply present numbers in a 6x6 grid. This might be too much information, but we might be able to get around that with conditional formatting (either row wise, or column wise or overall, depending upon the message we wnat to show). Source: https://t.co/W0BUHF3ZAm |
images/image_55.jpeg | 1. Doughnut charts are bad. Human eye can’t perceive areas as well as lengths2. Legend away from the chart, leading to constant head movement back and forth3. What’s with the colour scheme? So many chart colours so similar to the background! It’s hard to say that a chart even exists there, if not for the yellow. Three stacked bars with nice colour schemes would’ve done so much better! Source: P&G annual report |
images/image_56.jpeg | 1. The stash of money on the left is superfluous and adds no value2. Graphics is way too “long”. takes too much time to understand3. Information in semicircles is in areas, which the human eye can’t easily interpret4. This achieves nothing that a simple horizontal bar graph with marked data labels couldn’tSource: Times of India, via LinkedIn |
images/image_57.jpeg | Apart from being a pie chart with more than 3 sectors (which is bad visualisation in itself), this visual has a serious problem in emphasis.The colour scheme used draws attention, in order, to “all others”, Google and Facebook, while the story is supposed to be about amazon and Microsoft. So bright colours for amazon and Microsoft, and muted colours for everyone else, is the way to go. Then again, Amazon is 4.2%, while Microsoft is 4.1% - hardly a difference that can be noticed in a pie chart unless you see the numbers as well. So you’re better off putting this as a bar graph, with kinks for Google and Facebook and eliminating All Others. Source: https://twitter.com/eastbengal/status/1042653062471577601 |
images/image_58.jpeg | Straight old bad pie chart, with a subdivision in the pie. And what’s with the practice of writing numbers in the sectors, rather than the categories? Makes you get a “tennis head”, looking back and forth from legend to graph. Also, too many significant digits. One less wouldn’t hurt. PS: Doesn’t this chart look like pacman? Source: http://www.ilo.org/wcmsp5/groups/public/—dgreports/—dcomm/documents/publication/wcms_625890.pdf |
images/image_59.jpeg | !. What’s with the hexagons? I’m opposed to using areas for sizes but at least squares/circles can be intuitive. This is bizarre2. Bizarre colouring scheme3. Impossible to find your country on this (unless this is US or Japan)4. And what do you even mean by “total debt”? Concept itself is ill-defined because one person’s liability is another person’s asset. Debt is a zero sum game. Source: https://www.weforum.org/agenda/2018/05/63-trillion-of-world-debt-in-one-visualization |
images/image_60.jpeg | Disclaimer: I have no domain knowledge at all about athletics analytics, so I’m critiquing this from first principles. My discomfort is with the choice of axes, with distance on the X axis and time on the Y. Which means the lower your graph is the better, which is counterintuitive.Also, I’m not sure if this is time on the Y axis or difference in time, in which case it should be labelled differently. Finally, black background is a no-go. Makes it hard to read. Use a white background and solid dark colours for best effect.Source: https://twitter.com/albertostretti/status/1041273207167361024 |
images/image_61.jpeg | There are way too many lines in this graph, and also a bar graph that lies on both sides of the axis. And also multiple Y axis (and it’s not clear which line/bar corresponds to which axis). So most people will simply “put well left” to this graph |
images/image_62.jpeg | The information in bar graphs is in the length of the bar. So it is recommended that such graphs be used only when all data points are positive and all are negative. If not, you’ll have bars on both sides of the axis, with lengths you can’t compare easily, and that subtracts from the visualisation. This, in that context, is a strange graph with “floating bars”. It is not clear if the lengths of the bars is the same, since they have been aligned somewhere midway. Also, with the stacked bar, it is not clear what the lengths are unless you read the numbers - and 25 numbers is too much information. Thinking about it, it is not an easy data set to show, since there are 5 categories, and 5 categories within them, and you need to show the distribution of the latter in the former. A stacked bar is possibly the least worst form of showing this data, but the only way in which people can figure out the proportions there is if the bars are aligned at one end - not somewhere in the middle.At least I’m glad they didn’t use pie charts!Source: https://www.linkedin.com/feed/update/urn:li:activity:6445045035439390720 |
images/image_63.jpeg | I encourage you to click through to the source, since the screenshot doesn’t do justice to how bad this interactive graphic is. First of all, I don’t see why we need an area chart to show this picture, when simple lines will do. This is a time series, and a line chart is the most intuitive way of showing that.Apart from being unnecessary, using an area chart here draws attention to the fact that the two categories don’t add up to 100% in most cases, something that isn’t explained in the data or the source. When the two categories shown are “broadly opposed” and “broadly supportive”, you would expect them to add up to 100%. Maybe the rest of the people are neutral, but the data just doesn’t show that. Also, you see an up-trend in the proportion of people “broadly opposed to homosexuality”. This is counterintuitive, and the article makes no attempt to explain it (the simple one being that the “empty” portion of the graph has decreased since 2002). Then, the tool tip on hover has too much information. You don’t need data series, X axis value and Y axis value. Just the last will suffice. This is a problem with using the defaults provided by your software rather than putting in a little thought to figure out how your message can be best communicated. Finally I don’t see why this should be an interactive graphic at all. The interactivity adds no value (you frankly don’t need those tool tips), and makes the graph “jumpy” by changing focus betwen regions. Source: https://www.livemint.com/Politics/nLQiPpl5UICajLDXETU3EO/Homosexuality-in-India-What-data-shows.html |
images/image_64.jpeg | This is a response by the opposition party to the previous graph posted here. Data here is not dishonest (though I haven’t really checked if bar lengths are proportional to the numbers they represent), but that doesn’t make this graph any less bad. Using bar graphs to represent one time series is bad enough. This one represents two time series, in different units (Indian rupees and dollars), in the same set of bars, without really telling you that two time series are being represented. And there are way too many numbers there, with all the data labels and changes and all that. And the ugly arrows showing changes still remain. In fact, this graph is so ugly that without seeing carefully you might think this is the same as the one it responds to. It might make a political point of having responded, but there is absolutely no information content. Source: https://twitter.com/INCIndia/status/1039134643922984962 |
images/image_65.jpeg | The text in this graphic is in Hindi but that doesn’t matter. The relevant numbers are in English.This is supposed to be a bar graph, except that the length of the bars have nothing to do with the quantities they represent. For the uninitiated, this is a political advertisement, and so it is not surprising that they try to use data to make a political point. But there is a good way of doing that, and this is nowhere close to that. How can the bar for 80.xx be smaller than the bar for 71.xx? How can the bar for 71.xx be only marginally larger than the bar for 40.xx? And if the intent of the graph was to show percentage changes, why even use a bar graph? A line would have done much better. Source: https://twitter.com/BJP4India/status/1039110217160478720 |
images/image_66.jpeg | Head hurts to look at this graph, with both dodged and bar graphs. Each category has two sub-categories it seems, and each sub-categories has two colours, and it’s not very clear what each of them is supposed to represent! And the headline figure is not to be found anywhere in the graph. I must admit this domain is not very intuitive to me (though I’m a capital markets guy), but this just looks like a graph for the heck of it, without wanting to convey informationSource: https://twitter.com/tracyalloway/status/1038946367991771137 |
images/image_67.jpeg | When you have more than 4 or 5 categories that you’re showing on a bar graph, it’s best if you can tilt the graph by 90 degrees (in R, it’s just one simple command - “+ coord_flip()”. The reason is simple - so that your readers don’t get a pain in the neck trying to read the categories. Also, the white spaces between bars seems too high here (but I guess that will by default get corrected when you flip coordinates), making it not as pretty as it could be.Also, given the large number of categories, it might make sense to do away with the axis, and instead plot the data labels directly on to the bar - right now your eye needs to stretch too far to see what the actual rates are, etc. PS: This goes beyond the visualisation, but having (weighted) averages for the blue and orange categories might help in the analysis - if it’s convenient for the message the author is trying to send out, of course. Source: https://twitter.com/rupasubramanya/status/1038989226149203969 |
images/image_68.jpeg | Dishonest chart. Stacked bar for time series is not a great idea. For starters, the X axis is off - the period between bars is unequal (1750-1800 in one case; 2006-16 in another), but he bars are shown at the same distance - which is dishonest. Then, you need to keep looking between the chart and the legend, which is again not a great idea. Then, it is hard to see pattern in bars that aren’t at either end. Finally, it’s a time series. And showing it all sums up to 100% is less important than showing the patterns for each category. So the choice of stacked bar is horrible here. This should’ve just been a set of line charts, with the legend placed within the chart at the end of each line. Source: https://www.nature.com/news/lessons-from-history-for-the-future-of-work-1.22825 |
images/image_69.jpeg | This is absolutely atrocious stuff - if dodged bars are hard enough to get information out of, why would you add a separate bar, in a different shade, which possibly has different units (percentage versus percentage points?)? The best way I would think of showing this information is a slopegraph, where for each category, you’d see the level in 2015 and level in 2017, and a line connecting the two, and the slope of that line (also possibly coloured) can indicate the degree of change. Even a labelled scatter plot would be okay to show this data. It seems like the maker of this graphic chose the worst possible way to show the dataSource: https://twitter.com/CathyYoung63/status/1038632265272705025 |
images/image_70.jpeg | What is this even? As if doughnut charts are bad enough, this one has nested doughnuts. And a colour scheme that shows data rather than just distinguishing by category. Congratulations to anybody who managed to get any information out of this. And since it is so hard to understand what is happening, I cannot even recommend an alternate way of showing this data. Source: http://www.visualcapitalist.com/worlds-ultra-wealthy-population-chart/ |
images/image_71.jpeg | A choropleth (coloured map) is wholly unsuitable for this kind of a visualisation, because differences in data points are large (making them impossible to capture on a colour scale), and because there is no real geographical pattern.A pair of bar graphs would’ve done much better in showing the data, insteadSource: Hindustan Times |
images/image_72.jpeg | I don’t even know what the map on the left is - just that some states are coloured yellow, others are black. The stacked bars on the righ are supposed to show the total number of registered foreign companies by state, and the number that are active. Except that it’s not clear which number is which.The legend suggests that red suggests total registered and green active. The bars, however, are the other way round. In fact, now I’m not even clear if the bars represent the information accurately at all. There is a simple enough way (using stacked bars) to show the information -with the “outer” bar showing total number of companies registered and the “inner” bar showing those that are active. So you first have the green bar on bottom showing active, and then a red bar on top that shows inactive, so that the sum of the two bars shows the total! As simple as that!Source: Not sure. This was posted on Twitter by a “third party” without attribution. https://twitter.com/KiranKS/status/1037365038166597632 |
images/image_73.jpeg | Why do they even bother using circles here? Makes it very hard to get anything out of this graph. Also it’s not clear if the circles in the US and Europe graphs are on the same scale. This is a time series, so could’ve simply used line graph for this. Would’ve provided all the information concisely. And while it is good that they have provided data labels (to compensate for the bad graphics, I suppose), they have too many significant digits - making it hard to easily understand the numbersSource: https://twitter.com/DriehausCapital/status/1037049205838295040 |
images/image_74.jpeg | i don’t need to say much about this. this post is purely for the purposes of documentation. If you can get any information out of this treemap/heatmap, please accept my congratulationsSource: https://twitter.com/NickatFP/status/1036651356382289924 |
images/image_75.jpeg | It’s a rule that bar graphs need to have the axis at 0. And the graph on the left is being dishonest here in not showing that. If it’s an 85% jump, the bigger bar should be a little less than twice the size of the smaller bar. And this is misleading.Even more misleading is that different bars in this graph have different scales - in fact, without the numbers being present there it is impossible to know what the number for each graph is. For example, the 140 in the second green bar is shown to be much smaller than the 137 in the second orange bar - the mind boggles. A good way of showing all this data would have been to use a slope chart. That is effective in comparing two data points across several categories across time. Source: https://twitter.com/BJP4India/status/1036682271317811201 |
images/image_76.jpeg | We didn’t need the three bars for each country, since two are complimentary, and the third is a “don’t know” which doesn’t add so much value. Instead, the “don’t know” could’ve been eliminated and proportions recalculated, to have one bar per country. And the bars should be laid out horizontally, so that the names of the countries are put out in full - you don’t want the graph reading exercise to also be a geography test. Finally, the arrangement of the countries in descending order of height of one bar is useful, but would’ve been better to highlight the bar corresponding to the overall EU average in a different colour, to draw attention to those above and below average.Source: https://twitter.com/spignal/status/1036636228450484224 |
images/image_77.jpeg | 1. Bar graph isn’t a great idea when you have bars on opposite sides of the origin. Idea of bar graph is to compare length of bars, which is impossible if they are on opposite sides of the axis.2. Dodged bar charts again aren’t a great idea, though in a sense they get mitigated by the bars being on the opposite sides of the axis, which means we can compare the bars on the respective sides. Yet, it is hard to understnad any “signal” in this graph3. In the worst case, this graph should’ve been flipped by 90 degrees, so that the categories corresponding to the bars could be readable. Now that’s impossible without breaking your neck.3. What are the highlights even supposed to mean? It seems to me that there is no real insight in this graph, and the purpose is mainly to convey data. In that sense, it would be best to simply publish it as a table, maybe with highlights! Source: https://twitter.com/Trinhnomics/status/1036452717730791424 |
images/image_78.jpeg | When you are using a bar graph, the axes should start from 0. This is a rule, not a convention. And this graph flouts it. The information in a bar graph is in the length of the bars, not the position of the top. Even though data labels have been provided, a reader who doesn’t bother reading those will think the quantity has dropped by much more than it actually has. Given that this is a time series, I don’t know why they decided to use a bar graph instead of a line chart - the latter could’ve been shown comfortably using this scale without any issues. Finally, what explains the colour scheme of the bars? Even if the axes had been drawn properly, this colour scheme alone would’ve merited the inclusion of this graph in this tumblr. Source: https://twitter.com/tEconomics/status/1035089817518964736 |
images/image_79.jpeg | This is not a “bad” visualisation. It’s mostly good, but they’ve got some decoration aspects wrong. And I’m glad they’ve used a scatter plot - which is the precise way to show this data. When you show discrete data on a scatter plot, you can have many overlapping points, and that can lead to a sort of grid like structure which makes the graph look useless. One way to get around it is by using “lighter dots” (in R, you can control this with the Alpha parameter). So when you have overlapping dots, it will automatically become darker showing a more even distribution. Also I have an issue with the labelling - they should’ve labelled fewer players and in a larger font. The rule followed for labelling players is not clear - but that’s a quibble. As for font, they could’ve used larger fonts for more “exceptional” players. So Messi would’ve been in a very large font, etc. A good rule for labelling dots in this kind of a graph is to label dots on the convex hull, and maybe the dots in the next layer convex hull. In other words, label dots that are “exceptional”. Source: https://twitter.com/VamosReva/status/1033192802681683968 |
images/image_80.jpeg | The use of two Y axes is confusing, especially since they show the same units (though strictly speaking, GDP is $Billion per year, while Market Cap is $Billion). And I don’t know why the data labels are marked only for India and China, but not for others. And “dodged bars” make comparison difficult. The easiest way to have shown this data would have been with two separate bar graphs, one showing GDP and the other showing Market Cap. Since India is second to China on both, none of the story would’ve been lostSource: https://twitter.com/banglani/status/1034252158236061696 |
images/image_81.jpeg | This picture on top gives very little info, especially in the Twitter size. Pie charts may be bad, but “doughnut charts” (like this one) are worse. It’s almost impossible to compare. Most labels are tiny, and with so many entities, the colours don’t make sense either. Interestingly, in the article (link below) this graphic is followed by a table showing share of top 10 online retailers, which clearly shows Amazon is at 49%. Now, i understand that including a graphic can draw readers’ attention, the format used isn’t great, since it’s hard to get much info out. I know this might be a surprise to you, but I might have just used a normal pie chart for this, but just showing “Amazon” and “everyone else”. The idea is that pie charts are great at showing 50% (or other well-defined proportions where there are only two categories). And since there is a table below any way, just this “binary” pie chart (with Amazon filled in bold) can be used to draw attention.Another option is to show the rest of the companies as well, but to show them with muted colours, maybe shades of grey. The important thing is to highlight Amazon being close to 50%. Source: http://www.visualcapitalist.com/chart-shows-amazons-dominance-ecommerce/ |
images/image_82.jpeg | Strictly speaking, this is not a bad visualisation, but the choice of the bar graph, and axes, sends a misleading message. The most straightforward way to make this comparison would have been to use two lines, since it is showing time series. However, as I’ve mentioned before, it’s a rather common practice to use bars instead - though one I don’t agree with. Here the issue is the choice of different scales for the two graphs. While it is useful to see the trend - US growth rate is picking up while China is stagnant, the choice of scales (which is hard to see) masks the fact that the China growth rate is much higher than that of the US. So I would put this down as a misleading graph - though there might be editorial decisions for showing the data this waySource: https://twitter.com/acemaxx/status/1033949349200515073 |
images/image_83.jpeg | One way in which graphics departments at newspapers try to justify their existence is by replacing standard graphics devices (such as bars) with graphics of items that are relevant to the story. Apparently, this makes the graphs “more relevant”. But standard graphics devices have been developed over time for a reason - for they are easily understood and easy to read. And when you replace bars with figures of whatever kind, you have the standard problem of not knowing how to interpret size. This example from the NYT is a beautiful example of how not to do graphs. Firstly, they use gavels (or are they rifles?) in place of bars. Then it’s not clear if the information is in length of stick or size of head. Then, assuming it’s the length of stick that has the information coded, there is the cardinal mistake of not starting the Y axis at Zero - remember that in a bar graph, the information is in the length of the bar, not the position of the top. In mitigation, it appears that an editor at the newspaper has recognised that this is a poor visualisation, and helpfully added the entire data set in the description. Source: https://twitter.com/tedfrank/status/1033897522908069888 |
images/image_84.jpeg | The last piece on Ishant Sharma’s lengths was on account of a graph sent by a reader. Now it turns out there is a much bigger chart crime in the same piece. Why oh why is this described using a stacked bar chart? First of all, both quantities are time series, so a line chart is what should be used. Even if we grant that bars can sometimes be useful, I seriously fail to understand why stacked bars have been used here. The two quantities (matches and economy rate) aren’t even in the same dimension! The downside of stacked bar charts is that while we can easily compare the total quantity, and the quantity represented by the lowest bar, there is no easy way to compare lengths of other bars. And so the story of the economy rate change over time (whatever that is) gets completely lost due to the way the chart has been used! Source: http://www.espncricinfo.com/story/_/id/24468047/how-ishant-sharma-got-mojo-working |
images/image_85.jpeg | This is a good old pie chart, supposedly showing the distribution of Ishant Sharma’s bowling lengths. The one thing you can figure from this is that about half his balls are at “length”, and I like it that the legend includes both the category and the percentage in the sector itself. Then again - if you’re mentioning both category and percentage in the chart explicitly, why have a chart at all? Why not a table? And this graph tells us nothing about where Sharma bowls his “average ball”, since there is no logic to ordering of the sectors. By now, you might know me and my view on pie charts, so I would show this information in bars. And the bars would be ordered (since the length of balls has an ordering). Source: http://www.espncricinfo.com/story/_/id/24468047/how-ishant-sharma-got-mojo-working |
images/image_86.jpeg | It is good to see that people are looking to show distributions of data rather some simple summary statistics. This graph (and the others on the same tweet) shows the distribution of a type of hypertension among districts in different states in India. And some states having multiple clumps shows you the value of showing the full distribution rather than summaries. The implementation isn’t great, though. The curves are all overlapping one another, and there are two curves for each state (male and female), complicating things further and making the chart unreadable. While the number of categories is large, one way in which this might have been salvaged is by using violin plots, which show distributions by category, and in a non-overlapping manner. With some cleverness (I know this is easy to do using Python and Seaborn), you can show two distributions (male and female in this case) on the same violin plot in two halves of a violin (like this). This one is just unreadableSource: https://twitter.com/anupampom/status/1033091730034315264 |
images/image_87.jpeg | This was rather easy - simply use a “waterfall chart” like the ones that strategy consultants like to use. The left data point (what it was in 2014) is irrelevant, and could have possibly been shown using a dotted line for reference. The “time series” (with 2 points) is irrelevant so the line is unnecessary. Should’ve started with a big bar of “projected population without interventions”, and then smaller squares showing reduction due to different reasons, and then finally ended with a smaller bar on actual jail population. Source: https://rikers.cityofnewyork.us |
images/image_88.jpeg | 1. It’s hard to search for your state here2. No idea what the colour scheme is supposed to convey3. Numbers are coded in areas, which the human eye can’t perceive as well as lengths. 4. Very hard to understand what the radius of the sectors is supposed to convey5. some of the sectors are literally invisible .Source: https://twitter.com/trevornoren/status/1032650674486501376 |
images/image_89.jpeg | No quibble at all about the choice of using a line graph here, or having multiple smoothed lines and also a trendline (they all add to the information). My discomfort here is that the Y axis (according to the label) shows the “change in mean see level”, and that confuses things majorly. Intuitively, when the Y axis label says “delta something”, you assume that it shows the rate of change, and that doesn’t sit well with a graph that shows a nearly secularly increasing line that goes from negative to positive. Almost like the sea level has gone through the bottom of a parabola - which doesn’t make sense. What is more likely is that the Y axis shows the mean sea level relative to the measurement on a particular day (sometime in 2004 it looks like). And if this is the case, the graph starts making sense. In that sense, this is a poor graph purely because it has been badly labelled, both in terms of Y axis and the overall title! Source: https://twitter.com/ZLabe/status/1032824717915242497 |
images/image_90.jpeg | OK I really don’t know what this cube is supposed to represent. I don’t know what the axes are supposed to represent. I don’t even know if the information is supposed to be encoded in the surface of the cube, or in the volume. Parallel edges of the cube are labelled differently, suggesting that the information is not in the volume. Or maybe it’s not a cube at all, and it’s simply a hexagon? In that case, the representation gets even more puzzling. To summarise, I don’t even know what’s wrong (or right) with this chart, but am putting it here because it falls under the umbrella of bad visualisation! Source: from a US Department of Defense report |
images/image_91.jpeg | This is indeed a crazy chart, as the accompanying tweet says. The problem is it compounds two errors - the use of “dodged” bar charts and the use of bars to represent a time series. Dodged charts are hard to read because it is hard to compare. Here given that the two series are at different “levels” it isn’t that much of a problem, but when you need to find a pattern in alternating bars, it is not easy to get information. And the problem with using bars for time series is that to paraphrase Edward Tufte, they “use too much ink”. The information in time series is usually the trend and not the absolute values, so showing the full bar is overkill. Just the top points of the bars can be shown, with lines connecting them to show the trends, and that gives us the line chart! Source: https://twitter.com/tbiesheuvel/status/1032169655996219392 |
images/image_92.jpeg | The chart itself is okay - just a pair of ordinary line charts, except that the way it is drawn the message is hidden. If the main message (as suggested by the sub-heading) is that the use of bitcoin cash is tumbling, it is not visible in the graph. In fact, it takes a second look to find that there is a second (orange) line right close to the bottom, which represents bitcoin cash and which is declining. This chart makes an excellent case for the use of multiple Y-axes (something that some visualisation gurus frown upon), if only to highlight the decline in bitcoin cash. If that is not feasible, then we might as well use small multiples and show the two currencies (bitcoin and bitcoin cash) in separate graphs, each with its own axes. Source: https://twitter.com/TheStalwart/status/1031570495479730178 |
images/image_93.jpeg | Data from maps is hard enough to read. So you can imagine the trouble with such maps that mangle existing maps. First, you would be hard pressed to find your country here, even if you know its approximate position in the world map. Second, information is encoded in areas. Pies and bubbles are at least “regular figures”, but how can you really understand anything from the areas of weird shapes that are countries? Fianlly the colour scheme is pretty confusing. Legend exists, but it’s not clear what information is encoded in the colour and what in size. Source: https://twitter.com/kumarsamit/status/1029991962827530240 |
images/image_94.jpeg | This is an impossible graph to read. Different parts of the graph seem to conflict and prevent you from getting information from other parts. It could’ve been simply done with three lines, maybe with points marked. Using bars for time series is a bad idea. And I don’t know why the two other lines have been “filled” to make area charts. And then you have the boundaries and contrast and gap in bars which makes the whole thing impossible to understand! YikesSource: Via teasri and pseudorasmus on twitter |
images/image_95.jpeg | This map shows cities where we need air conditioning to live, given current temperatures. There are multiple problems here. The first thing that you notice when you see this map is the differential density - why are there so many points in Eastern US, and so few for India? The most obvious explanation is that the makers of the graphic have been lazy - going by available rather than appropriate data. The least they could’ve done would have been to filter out the smaller “cities” in the US to equalise density.But then - why do we really need to restrict to cities - it appears that the paragons of liberalism at the Guardian seem to have decided that airconditioning is not important for people who don’t live in cities? Including rural areas would have had the dual purpose of making the map equitable and smooth.Finally, coming to the graph itself, the grey backgrounds and colour scheme (purple is too close to both blue and black) makes it really hard to interpret, or locate cities. At best the graph indicates regions, which makes one wonder why they’ve restricted to cities - the most likely explanation there again being that they’ve been lazy Source: https://gu.com/p/96xak |
images/image_96.jpeg | OK it is noble that you try to show nutritional information along with your display board, but why spider charts? First of all, this information on menu boards is not intuitive. Then, you don’t know where the centre is.Most importantly, spider charts don’t show information effectively because you assume the information is in the area of the web, but the area is purely a function of how you’ve ordered the radial lines. A spider chart can never show more information than a set of bar graphs! (one for each axis).Here, too, it would’ve been better if they’d used a set of (possibly normalised) bars for each nutrient class, one view of which would show customers what the particular food/drink is heavy in, and what it’s light in! Source: https://twitter.com/rahulw_/status/1028271201762734081 |
images/image_97.jpeg | First I award some points for not using a pie chart. And then I subtract some for using a two-dimensional area-based representation (remember that the human eye can’t perceive areas as well as lengths). And then I subtract more points for the colour scheme, so that the total points goes into negatives!I mean what was the chart designer even thinking? It’s good intentions to colour squares based on continent, but not if Canada and Mexico are adjacent! OK I’ll stop hereSource: https://twitter.com/wef/status/1027449409695633408 |
images/image_98.jpeg | First of all, I have no clue what this is supposed to represent, and the way that the data has been presented gives no information on this. There is some text around the edges, but it’s unlikely anyone can read them. And there are lots of lines!The tweet through which this came to my attention is about “5 awesome data visualisation sites”, and if a recommendation for these sites comes along with this graph, I would definitely not take these recommendations seriously! Source: https://twitter.com/42Courses/status/1027487125577588737 |
images/image_99.jpeg | Graph on left is okay. It’s the graph on the right that is weird. Message is “only old people are watching TV”. And when you look for old people you want to see the right most bar. And it turns out the bars are in reverse order (for whatever reason), making it that much harder to get the signal from the data!Also - speaking of the data itself - I’m curious to know how they measured the hours spent watching “Pay TV” in the 2-11 age groupSource: https://twitter.com/MylesUdland/status/1024838427001479168 |
End of preview. Expand
in Dataset Viewer.
README.md exists but content is empty.
- Downloads last month
- 39