image_path
stringlengths 19
21
| description
stringlengths 0
1.58k
|
|---|---|
images/image_100.jpeg | The reason this chart makes it here is due to the atrocious colour scheme, and the way the legend is placed. If you want to know which is the highest weighted stock in the S&P500 at different points in time (which is what colours in the lower line indicate), it’s not going to be easy.You need to keep looking back and forth between the line and legend, the colour scheme is bad (light colours don’t look good on white background) and the line itself is not bold enough. It won’t be easy but I would think of using darker colours, and placing labels right next to the line (rather than in a separate legend). Another way would be to shade different vertical regions in the graph (with rectangles), indicating different stocks being the highest weighted (again with the title right there rathe than in a separate legend)Source: https://twitter.com/lhamtil/status/1026262383512178688 |
images/image_101.jpeg | Intent is good - trying to show how much of land is used for what, but several problems1. Human eye can’t easily measure areas. So relative sizes of different uses (apart from that cow/pasture range is the largest), is hard to guess2. You don’t know where to look to find what. If I want to know “how much do roads occupy”, I have to look at several places through the chart to find something3. Presence of borders and coastlines makes the areas harder to measure, and even the inside boundaries are not always linear. Best represented as a bar graph with data labels clearly marked, and bars sorted in descending orderSource: https://twitter.com/business/status/1024238805249937408 |
images/image_102.jpeg | My big problem here is with the use of circles to denote size. Remember that the human eye can’t compare areas clearly? Also when you use circles, it’s not clear if the magnitude is encoded in the diameter or the area of the circle. So without those numbers there, it would be very hard to determine the size of Ant Financial. I would simply replace these circles with bar graphs. Will tell the story much much better.Source: https://twitter.com/vikasargod/status/1024210022299971585https://t.co/UV4eKeVcY3 |
images/image_103.jpeg | This map gives out absolutely no information. Maybe it’s the colour scheme (grey background) and the colour and size of the dots chosen, but I can’t make out much apart from the fact that there seems to be a lot of red. How would I do this differently? For starters, I don’t think the map is the right representation, unless there is a clear geographical pattern. I’d start with a bar graph showing frequency of each suffix. Then, if and only if there is a clear geographical pattern, I’d possibly show the dominant states or regions for each suffix in another table! PS: If this graphic was made in India, it’s extremely poor form to use a map of India that doesn’t show Gilgit-Baltistan and Aksai Chin as part of IndiaSource: https://twitter.com/avtansa/status/1023684468429742081 |
images/image_104.jpeg | It’s somewhat excusable when a “normal” chart made for corporate or journalistic purposes is badly made, since the maker of the chart may not be an expert. But then when a visualisation that seeks to teach what charts to use when does it badly, it is inexcusable. Leaving aside the recommendations for a bit, the “flowchart” here is impossible to use. The same kinds of charts appear in multiple places. The decision trees aren’t very well visible. The centre-outward format doesn’t work. And then, some of the recommended charts are rather ugly. I mean, the only set of charts in this set that make sense are the “distribution” charts, and even there there is an error, since you aren’t supposed to leave gaps between the bars in a histogram. Oh, and the “spider chart” is an abomination. Worse than pie chart. Source: https://twitter.com/ROIChristie/status/1023322139028586496 |
images/image_105.jpeg | There’s almost nothing right with this chart. What is even the point of the spiral? How can you find a country in this? How can you read some of the numbers? Why couldn’t we use a simple bar chart (with horizontal bars since there are so many)? And the atrocious colour scheme is the least of the problems. Source: https://twitter.com/NinjaEconomics/status/1022798759359471616 |
images/image_106.jpeg | One of the fundamental principles of visualisation is that the human eye can’t really see measure areas easily (lengths, on the other hand, are more intuitive). This is also the primary criticism of pie charts and bubble charts. While this graph is no doubt an interesting way to see how interesting a set of playoffs was (I really like how the sizes of the squares varies by round, and how they’ve been stacked), the problem is that the information is in the area - the amount of white in the picture. And that makes the graph hard to interpret. It takes way too much time to get information out of these graphs. Source: Joe Harris. https://twitter.com/WhiteBallStats/status/1022445673789247488 |
images/image_107.jpeg | 1. Too many significant digits in each graph2. Legends placed really badly. You need to hunt for them3. The two lines in each graph will add up to a 100%. So maybe they could’ve combined the two lines into one, simply showing the “percentage placed”? Then the two graphs would’ve been on the same scale, and could’ve been combined into one! 4. And the scales on the two line graphs are totally different - by an order of magnitude. While it isn’t dishonest per se, there might have been information that is not being conveyed because of this difference in scale. Again, simply showing percentage might have been betterSource: https://twitter.com/peepultree/status/1022684732138565633 |
images/image_108.jpeg | It’s downright dishonest to start bar graphs anywhere except zero. The information in bar graphs is contained in the length of the bar, not the position of the top. So truncating bar graphs gives a warped view of proportions, and is thus taboo. Source: https://twitter.com/pravchak/status/999512831254032384 |
images/image_109.jpeg | Choropleth again, but we need to move our head back and forth between the two charts (not to mention their respective legends) in order to make any correlations. And even then, there’s nothing clear from the two pictures. Much better presented as a scatter plot, with geographical information presented separatelySource: https://twitter.com/guywalters/status/1003618300427948032 |
images/image_110.jpeg | OK this graph is THE WORST. Where do I start? 1. Pie charts are bad2. Doughnut charts are worse, since it’s not clear if hte magnitude is in the arc length or area of the doughnut3. Legend placed away from chart means you need to keep oscillating between chart and legend4. The colour scheme!! Can’t they choose more contrasting colours? You don’t even know where one data point ends and the next begins.Source: https://twitter.com/ow/status/1010434421651136513 |
images/image_111.jpeg | This graph can be used to induce fear, that we are in the middle of a big bull run and the bubble will pop soon. But I’m uncomfortable with the way the data has been truncated in the past. Like only the bull runs have been shown, not how the bubble has popped. So it makes you think that the markets simply fall off a cliff at the end of the bull run! Source: https://twitter.com/6pranavk/status/1012132976375812097 |
images/image_112.jpeg | Decent graph, but information could have been represented much better with a scatter plot (with forward passes, and sideways passes representing the two axes). While this format shows a lot of data, the problem is that it is hard to look for a team in this, and nothing stands out. Source: https://twitter.com/shijith/status/1012598347117391874 |
images/image_113.jpeg | The scatter plot is fine, but the author of this graph has tried to force fit categorisations to the data points, giving all these weird amoebic shapes. The only thing the weirdness of the shapes conveys to me is that there is possibly not that much of a relationship between geography and where a country lies on this chart! Takeaway from this is that you should use simple lines and curves, and if some data points don’t obey your hypotheses, do be it. Source: https://twitter.com/simongerman600/status/1013098601441001473 |
images/image_114.jpeg | This is a decent stacked bar chart, and shows the right amount of information (like the percentage symbol only mentioned once). But until you see the legend, you don’t realise there are supposed to be four colours in the graph, not three. And that’s because the colours chosen are too close to each other (I understand that widowed proportion in 21-36 age group will be small, but it’s literally invisible here). Better to use contrasting colours for a graph like this. Apart from this, this is brilliant.Source: https://twitter.com/conradhackett/status/1014276459068841987 |
images/image_115.jpeg | First of all, using bar charts for time series is not a great practice, but my concern with this chart is with the use of default Excel format. First, the bars are too far away from each other. Secondly, the colours aren’t great. When you make charts in Excel, always remember to “flatten” the colours and remove any special effects.Finally, the value labels can use one less significant digit. If we can widen the bars and reduce the space between them, the values can be wholly inside the bars |
images/image_116.jpeg | Pie charts are normally okay when you’re showing only two colours, but you should make sure that the two colours chosen are contrasting. Like in this doughnut chart showing possession, you will have absolutely no information on which team had how much if not for the numbers mentioned there. Atrocious choice of colours! |
images/image_117.jpeg | Choropleths (data shown on maps) are a favourite of data journalists, but they have their own problem. Here, the choice of chart is appropriate (apart from the invisible legend), but it’s not clear that hte correlation pointed out here denotes causation (explanation here: https://twitter.com/page_eco/status/985896343536189441 ). In a geographical split like this, there could be several other reasons why the patterns are the way they are. |
images/image_118.jpeg | |
images/image_119.jpeg | While pie charts are inherently bad, and fail to convey information, this one is dishonest as well. For the market capitalisation shares of the 287 companies represented in this chart don’t add up to 100%. And the 282 small companies chosen to compare to the 5 largest - that 282 is a number pulled out of thin air. So not only is this graph ineffective, but also dishonest and sends out the wrong information |
images/image_120.jpeg | Banks continue to outdo themselves with bad graphics. This one is thoroughly unsuited for a stacked bar. First of all, there are too many colours, and similar colours. Similar colours don’t even represent similar assets. It is basically impossible to get any information out of this, even if you were to move your neck back and forth between the graph and the legend.This can be done much better in terms of a line graph, with one line for each asset. The other alternative, is to use “facets” (or small multiples) and have a separate bar graph for each asset. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.