Wednesday, January 25, 2017

Play Better to Work Better: Drawing Boxes and Using Vectors to Make Graphs, Charts, and Data Visualizations

by Steven S. Vrooman

I know Excel makes graphs for you when you click the graphy-looking images.

I know programs like Tableau make 


I know. Someone wants us to make dashboards. Someone wants us to process 

using the company's 

that had a very expensive licensing fee.

But I want you to relax and draw pictures in the sand with me for a second.

I want you to learn how to use vector graphics programs to PLAY with your graphs.

Graphs are often mistaken, ugly and uncommunicative because we DON'T MAKE DECISIONS ABOUT THEM. We just make what we can given our limited understanding of what our software packages can do.

What would you make if you understood what you could do?

This post is not going to be a comprehensive lesson in how to build graphics with Inkscape, the free vector graphics program you all need to download. This will simply use the program to show you how we can play with things. Vector graphics, by the way, are how designers make infographics and stuff. You can make things as big or small as you want without losing resolution as you save, which is soooooo important. (I know, you want to tell me that Adobe Illustrator has a graphing tab. Yep. And that's cool, too. But 1) if you are reading this and don't have/can't afford Adobe or 2) really do want to try a sandbox option, keep reading).

Only by playing with your options can you figure out the right choices.

Look, I know this is WORK and we are all supposed to be SUUUUUPER SERIOUS about our jobs and life and our own importance, but there are so many reasons we need to learn to play with our ideas more in order to both live and work better. This, guy, Practical Pig, who sang "work and play don't mix" in 1933, well, let's just say he's not up-to-date on productivity and management theory:

Here, for example, is me playing with my stylus and my tablet:

Have you ever done that? With a crayon or pencil or stylus or by dragging boxes around in software? Have you ever tried to figure out what you could do with a graph if you really tried?

Let's focus on bar graphs today as an example. They are ubiquitous, usually a decent choice, and really easy to mess with for our purposes.

You read my little comic up there. How wide should the bars be? Most of us will answer they should be as wide as Excel decides they should be. That answer stinks!

Let me take you through an afternoon I spent playing with a data set with one of my classes. Along the way, you will learn how useful it is to learn how to make graphs and charts with vector programs.

A note before we begin. Sometimes you do need real-time or pivot charts. Sometimes you need interactive stuff using a software suite that constricts your choices. Ok. Fine.

But sometimes you need to print. Sometimes you need to drop something on a slide. Sometimes you want to do what I preach in The Zombie Guide to Public Speaking and take your audience through a "multi-slide journey" and TEACH them your graph by adding and subtracting pieces to guide their focus. This lesson is for those times.

An example of how to do this and how to leverage the power of suspense and audience participation in that process is in David McCandless's TED Talk. Watch the few minutes starting at 6:15:

See? Isn't that effective? Well, it turns out that off-the-shelf software, which isn't deigned for those kinds of moments, isn't great at producing them. It can be done. But let's try something else.First I will make a bar by drawing a rectangle. Even those of you who are expert only in MS Paint can do this.

I have decided I am making skinny bars today. Why? Well, I want to stack a few sets of bars side-by-side, so I'm just going to go with it an see what happens. I need to copy that bar and paste it.

Okay. Now I can use those Y coordinates on the top of the screen to make them all end on the same bottom line. There are other controls to do this, but why not use the ones you can see in my screenshots?

Cool. Now I want to make the bars actually graph something. I can do this with all sorts of thicknesses and heights of bars, as long as I remember my units. In this case, since the data I am working with is percentages, I am going to make the max 1000 pixels in height, which will be my 100%. I will change the bars' heights now, and the smallest is 49.7%.

I also want to regularize the distances between the bars. I can do this in all sorts of ways, but if I want to do it using the X numbers at the top, which tell us horizontal positioning, since my bars are 20 pixels wide, I can place the bars with 20 pixels of space between them, for example, by placing one, say at X:100, the next at X:140, then X:180, etc. Or I can smoosh them together because I want to see how that looks like.

Then I will copy it all and even it up to the right so there is room for more stuff. Why? Well, this data was the racial breakdown of the student population at a large state university in Texas. Our class was looking at those numbers, which were presented on the school's website as one of those live, interactive graphical interfaces. We were trying to figure out how we might reimagine the presentation options. We need a block here for each large racial group at the school. Each block represents enrollment percentages in the last 5 years, with each bar being a year.

Now let's play around with it. If one of these things I make is useless and ugly, RELAX! I am just screenshotting things as I play with options to find the right choice. Live a little! Remember to play! 

What if we shared colored numerical bands across the bars, red for 60%+, orange for 50%+, etc.?

Kind of confusing? We'd add labels to this at the end to make sure people understood what the colors meant, but we should be able to make it clearer without the labels, shouldn't we? Hmm, not sure what to do there, so let me move on to something else. Sometimes a lateral move will bring the answer better than hammering on a problem.

Here I am playing around with the idea that I could try to help people visualize the size of the decline or increase from the first year in the sample to the last. A bit of simple arithmetic can tell me that number and then I can drop it onto the graph.

Is this a good idea? I don't know. Neither do you. Who is in the audience? Is this the information they really need? Will that help us make decisions? Is pink for a decline and purple for an increase too many colors? Probably, regardless of the audience, but maybe not. The important point here is that you cannot do this thing with the pink and purple bars if you are just using the options in graphing software. You are literally forced to think inside their boxes.

Again, what would you do with graphs and charts if you could do what the data was asking you to do for your audience?

Maybe I like those purples and pinks but want to jettison the percentage band colors? Well, it is easy enough to change the colors of the bars.

Then we can shift to representing only those changes.

This is my favorite trick for a slidedeck. I'll make part of the graph go away and just look at the part I want you to care about for a second.

And what if I want to zoom in on one of the bars and break it up into different factors, say men and women? I realize I need to make different color choices, once again, but isn't this an interesting way to do the "deep dive" people keep saying they want to do?

Okay. Back a few steps (CONTROL Z!!!!). Once we add some labels, does the percentage bands thing work better? 

Here is also where we can make choices. Do we need every year on the-axis? Clearly not. Is the race name label easier to understand underneath the way most programs would put it or like this, on top of the bars? How many numbers do we need on the tops of the bars? 

How abut if we shift the colors so we can make the race label a different color than the number labels, which creates some differentiation without adding more colors? 

And now we can see how this works with percentage bands with more like colors. Ah, good thing we are playing with options. That right-hand bar in the "White" group does not look like the same color as the three right-hand bars in the"Latino" group because of the way the grays relate to the black bars and the whitespace. Good to know. Would blue be any better?

Well, not sure about this. Also not sure about using analogous colors here. Maybe more contrast?

At this point, I decided I give up on the percentage bands idea. I can always, say, drop partly transparent horizontal boxes in different colors over them at some point in my slideshow to demonstrate those bands if I still like that idea after I'm done. But I'm just playing with adding numbers for now. How many do we need? I don't want to distract with needless detail. But I also don't want to drive people crazy. How about the endpoints?

I mean, we don't really need those internal numbers, do we? Not for the shape of this data. I could imagine a set where we would, but this is a pretty clear reasonably linear set of changes, so I think we are okay not knowing the exact number of white students in 2013, for example.

And then, if we make the rest go away and want to talk more about this fact, that a third of the campus is now Latino, well, maybe I will highlight that.

And then, even though I don't usually love slide backgrounds, I could make this bar a background to the next section. Maybe I could break the bar into different colors now to represent, say, different responses different parts of the university will be doing in order to met the needs of this changing student demographic? 

Or not.

I stopped there because class was over and we were just playing with ideas.

The point here is that with pencil or stylus on vectored boxes, you need to start playing with your visualization options. 

You can't communicate effectively if you don't understand your options. 

Tweet: You can't adapt to your audience if you don't know how to change your content. @morebrainzYou can't adapt to your audience if you don't know how to change your content.

So go experiment. Play around. Mess up some paper. Hit Control-Z a bunch of times in some software. 

Play better so that you can work better.

No comments:

Post a Comment