Column charts and bar charts, they are absolutely the same. There is no real difference, the only thing I can think of is that you use column chart for time periods, this is a conventional resort. And you use bar charts if you have very long labels, or if you have very long lists of data and this is just convenience. For god sake please do not put your labels under weird angles, use bar charts instead. This is much more convenient, much easier to read. Now, that's that. The single most important thing about bar charts, or column charts, or just any kinds of charts in general, that you do not visualize data. You help people to arrive to certain conclusions, to make certain decisions and for that we need to formulate those conclusions and those decisions. And as far as bar charts are concerned, you would most likely need to sort your bar chart. Let me give you an example, this chart is a bit of a mess. It's all over the place, you'll probably need to pause the video right now to study it in detail, but let me just quickly try to explain you what it means. There are certain people on the right who got certain grades. Could be Excellent, could be Satisfactory, could be Good. For example Michael S., who is someone in the middle there, he got 25% of excellent grades, 50% of good grades, and 25% of satisfactory grades. Now, this chart does not lead to any conclusion. This chart, in contrast, does lead to certain conclusions. This happens when we start sorting this chart according to the average grade. Now, I do not know whether this is a good HR practice or not. That's beside the point, but this chart definitely leads us to certain conclusions. Some people are great as they are, some people need help. And we can focus this chart even more. That red corner stands out and it tells us that okay, attention is required. Go have a look. What's wrong with those people, we probably need to do something about them. And this is what I mean when I say we need to communicate conclusions and not just a visualized data. This was we communicating conclusion. Let me give you another example. And this time, I would actually ask you to pause the video, download the data, play with it, come to some conclusions, and try to draw a meaningful chart. Something which doesn't look like a spider. Okay, I do not know what your results are. Let me just show you quickly my thinking process. Now, what I'd do first, I would draw a simple stacked bar chart just to see what's going on, and I would need to sort it. And immediately I will see that Russia, well, lags behind in consumption in most product categories. I would probably remove bottled water, because there's just way too much of it. And I will see that, well, in some categories things are better, and in some are worse. But overall, this is the conclusion. Next, I will notice that with this design, it's okay, it's easy to compare chocolate. But it's kind of difficult to compare coffee and ice cream. So I would probably break this diagram in to four separate diagrams, something like this. And then I would remind myself that all of this should be about Russia, and I will highlight Russia on all of those diagrams. And this, I think, is a rather sensible design. We see what is happening, that the situation is sort of okay with chocolate and not so much with coffee. We're on par with Britain, we're tea drinking countries, and we really, really lead behind ice cream consumption and bottled water consumption. Which is probably not such a bad thing. Less sewer, less plastic bottles. And just to remind you that my software would probably produce something like this, or like this, or like this. This is my doing, this is my work, this is my added value. As an analyst, I need to formulate the conclusion and I need to highlight the most important things. As far as I know, no software at the moment is capable of doing that. This is me, human analyzing stuff. And this, I think, is what you should be doing with your data, not just creating random diagrams. But rather formulating conclusions and producing sensible charts, which make things clearer, which prove your conclusion because that's their main point. Thing number two, please remove all the unnecessary details that PowerPoint is so keen to add. Once again we can see here a chart and a table, which is a sure sign that there's something wrong with the chart. Typically, we wouldn't need the table to go along with the chart. The chart is deficient in one way or another. What happens if we just remove the table? And what happens is that we will no longer be able to see the exact data points. So the first thing that we probably need to do is to add those data points. But then as a next step, we'll need to have a look at all those chart junk is the term, that PowerPoint or Keynote added to our chart. Those labels we don't need them anymore, those grid lines, the axis title, and we can just remove all of this. And if we are really paying attention we're can have a look at the legends, and we'll see that there is a lot of words repeating each other and probably we can cut down a little bit. And, as far as I'm concerned, this is a much better design, much clearer, much easier to read and interpret. Thing number three, your axis should start at zero, but please do not be overly fanatical about that. Let me explain what I mean. By now, I think everyone have seen those kinds of manipulations. This is a difference between 35% and roughly 40% and I think that's like five times the difference in the area. So, yes, the chart from the left is the most misleading, the chart on the right has the most truth in it. And the chart in the middle is something interesting. The chart in the middle is actually something Keynote would produce, by default. So you have to watch those defaults. This is us unintentionally misleading the audience, we should go and correct Keynote in this particular example. And it does it all the time, and the same applies to PowerPoint. This is just a Reddit joke, 99% of the time truncating the Y-axis is misleading. But what's important here is that 1% of the time, it's not misleading, and let me talk about that 1%. Sometimes, for example, if we talk about the server uptime, the difference between 99.9% and 99.99% is crucial. In one case you get 53 minutes of downtime per year, and in another case you get 8 hours 46 minutes, it's 10 times difference. And I think you should be showing that difference to the audience, this is a slightly better chart, I think. Now if you remove the axis, however, this is a crime, you should indicate that you've truncated the Y-axis. Thing number 4, this is just a nice thing to do, this is no way a obligatory, try to keep the Y-axis scaled across different charts. Try to make multiple charts compatible, comparable. If you look at this example, chocolate bar and coffee bar had the same scale. However, the streak wouldn't work with ice-cream, therefore, I'm putting chocolate next to coffee, because they are comparable. While Ice-cream and bottled water occupy the separate row, because they are clearly not. Thing number five, please do not use stacked charts for trends. And this one is important because I see people doing this all the time. Now stacked charts are great, they are like two charts for the price of one. In this example you can see judges experience and you can also spot some bias, potential bias, this is a hypothesis of course. Some judges always rule in favor and some always rule against. Now why is that? That's a question to investigate. However, if you have time periods, it's really difficult to see trends in these kinds of charts. It's kind of easy with the bottom thing, but with other things, it's not that easy. I've not seen whether this thing is growing or falling because all the other parts of the chart obscure the view. So if you have more than five data points please consider a line chart. If you're trying to display trends, Line charts should be your choice. This is the same data presented in the form of line chart, and I think, it's slightly better, you can see something. And this is a way of keeping the Xs in tact, and this is a way of slightly truncating the Xs, which I think in this case, once again, we should be doing. Yes, it's more dramatic. And it should be, because we're trying to disentangle those things to make them visible to make our communication clear for the audience, which is our ultimate goal. So to sum things up, Bar and Column charts are the same. Use Bar charts if you have very long labels, use column charts if you have time periods, this is the only difference. Sort and help people to make decisions. You need to arrive to certain conclusions and then produce a chart which supports that conclusion, this is the ultimate goal. The Y-axis starts at zero by default, please do not be too fanatical about this. The question is, do you need the second axis? And if you did truncate the Y axis, then probably yes, but otherwise probably no. By default you should be removing the second axis. And finally, please not use stacked bars for trends, use line charts instead.
