Graphs header

Your graphs suck!

In Tutorial by Morgan Bye1 Comment

Your graphs suck! Bold statement huh? I don’t even know you, Mr or Ms Anonyomous Internet Reader. Yet I know that your graphs suck. I know that when things get tough, and you really need a graph, you’ll suck the air between your teeth, quietly swear to yourself, before admitting defeat. Then, you’ll go open up Excel.

What’s wrong with Excel?” I hear you cry!

Ah-ha! I’ve got you! You’re already on the defensive. Trying to justify that Excel is a perfectly good piece of software used by millions of people every day.

And of course, you would be right. Ever since Microsoft first launched Excel in 1985 (ironically on Mac, not Windows), Microsoft has ploughed millions of dollars and man hours in it to make it perfectly acceptable.

But that’s just it, how many of us just want to be acceptable?

The truth is that you might be using the most amazing piece of software of all time. But, do you actually do anything other than the first few easy clicks and end up with something other than a basic template?

Why use graphs at all?

So recently, I was trying to make a point in a piece I was writing about how the world is becoming a better place. The average life expectancy, over the last 50 years has almost doubled. Meanwhile, the number of deaths in children has dropped and dropped.

I can just pull numbers out of the air like that – and it sounds convincing, right?

The problem is that if I just say something like that – especially on the internet – at best I sound like a politician. More likely, I sound like that guy down the pub that everyone tries to avoid.

To be taken seriously when I’m making a point – especially when I’m talking science – I need to talk numbers (and a really good scientist will back those numbers up with references).

The problem here is that talking about numbers is boring. Really boring. I haven’t even used a number yet and your brain is switching off.

Now I could write a big, long paragraph, with lots of commas. It would be very dense, very complicated and very tough to read. I’d have to use lots of “compared to”s and “less than”s and “in the year 19 [something-something] there was [some big number] per [some sciencey unit no-one ever explained to you]”.

Or what I can do, is say that fewer children are dying. Give you a nice, pretty graph and you can see if for yourself.

The x-axis (across the bottom) is time, i.e. a measurement for each year. The y-axis (up the side) is the number of children that are born alive, but do not make it to their first birthday. As this is big population data, this “death rate” is measured basically as a probability, i.e. for every 1000 births, so many will die.

Children are dying

As a scientist, numbers are really important. Some even go as far to say that numbers are the only real truth. But, numbers by themselves, without any context, don’t really mean anything. They are, however, the first stepping stone to significance.

4

See! The number 4, without telling you what the 4 is talking about doesn’t mean anything. If I tell you I have 4 cars, you’re jealous, if I tell you I’m giving you 4 gold bars, you’re really happy.

Every number, by itself, is only a single data point, but with enough data points we can draw trends. Human beings are really good at spotting patterns, and trends are better still. Doesn’t matter if it is the stock market, traffic on the way to work or the fact that red and orange leaves on the trees mean that winter is on the way. All these things I’ve just listed are patterns. Patterns that we ourselves learn by looking at data (i.e. experiences) and historical trends.

Fortunately for my point, the World Bank keeps really good data on what’s going on with peoples’ lives. When we head on over to their website, you have the opportunity to download all of their raw data about who is born, how long they can expect to live, and what is likely to kill them.

They even give you a nifty little web app so that you can view this data in (almost) real-time on their website (even if it is painfully slow).

But look at it.

World Bank graph

Just look at it.

It’s horrible.

It looks exactly like what everyone would expect to see from an automatically generated graph from a website.

A lot of people at this point would be guilty of just calling it. I need a graph. I got a graph. Job done. Click download, or if there isn’t a download button a screenshot is good enough. Paste into Powerpoint, get on with my life.

Persuasion

We as human beings, don’t just do work for no reason. There is always a reason, even if that reason is selfish. In this instance, we are generating a nice, pretty image because we think it will do a better job of presenting the data than a boring page of text.

When you think about it, we are always trying to persuade someone. It doesn’t matter what the reason is, but you are trying to convince someone of a point.

In this case, I am making the assumption that the reader is smart enough (they probably got into at least high school) to recognize a strong trend (or correlation in science speak).

Here’s my question to you. If I am trying to convince you, the reader, of a point I’m trying to make (doesn’t matter what it is) – shouldn’t I try to make the point the best way I can?

A bit of a rhetorical question, isn’t it? (see what I did there)

OF COURSE I SHOULD

But a large part of whether you decide to buy something is inherently wrapped up in the presentation.

If a homeless guy approaches you, telling that he will sell you a Ferrari at a great price, you would not believe him for a second. But if the same guy, with a clean shave, a $500 dollar haircut, beautiful suit, in an air conditioned showroom tells you that he can get you a Ferrari… Well, now you’re interested.

If I’m going to show you a graph, well now, it’s gonna be the best dang graph that I can possibly make.

The modern age

Given that we live in a world where I can get pizza delivered to my house without even talking to a human being, at any time of the night, I figured there must be a nice web app out there to make a pretty graph.

It seems in the last year or two, there seems to have been a big boom in companies offering infographics. Now I wasn’t going to shell out a $1000 dollars for one made by a graphical designer. But I thought I’d risk handing out my email address to a few companies to try a demo version or two.

Google docs

The most obvious choice to any Excel warrior is the freely available, Google Sheets. Advertised to many businesses as being able to take over from Excel at a fraction of the cost, it seemed like an obvious first choice.

Google docs graph

Pros

  • Selectable title – easily change font style and size
  • Editable axis titles
  • Good detection of year format, and then giving good axis labels (ie. ever 5 years)
  • Could change the colour scheme

Cons

  • Grid lines barely configurable
  • Axis titles cannot be moved, or changed size
  • No ability to “box” the graph, i.e. have a thick black line around it
Rating: PASSABLE

IBM Many Eyes

IBM stands for International Business Machines, but they gave up making computers that a normal person could use years ago. These days, they are much more interested in big data. Massive servers, networks and solving problems most people can’t even begin to think about.

Imagine then my impressed look when I found out that IBM has some graph making software. I was definitely going to have me some of that.

There was one slight catch, anything I made would be publicly available. Not public in a you-need-this-one-specific-in-no-way-guessable-web-address way, but public in a we’ll-put-this-in-a-gallery-on-our-homepage kind of way. So if this was confidential information, like business secrets or actual patients in a medical trial, this would be a definite no-go. But no matter, this is all public data anyhow.

And for giving away all of your privacy, you get this masterpiece

IBM Many Eyes graph

So let’s get into why this is a bad graph

  • There is no y-axis title available
  • You cannot get rid of that legend
  • There is no way to change that line colour from magenta
  • There is no way to turn the individual data points off to have just the line showing
  • The x-axis is a joke
  • There is no control of the grid – why do I want a line ever 2 years?
  • There is no way to zoom the axis – i.e. I cannot make the y-axis go from 0 to 125

RATING: NEVER GOING NEAR MY WORK

OK, maybe IBM was just having a bad day. Maybe this is just one of their products that they don’t care about very much. Let’s have a look at some actual graphs from genuine infographic sites – you know businesses that actually make money out of presenting data in convincing, pretty ways.

Infogr.am

Infogr.am is one of the biggest players in the infographic space, and they have themselves nice little drag and drop interface. The only downside is that they want $18/month to get started, and realistically to get most of the features you’d actually need to use them full time, you need to be ponying up $50/month.

Let’s see what, that extra cash gets you

infogr.am filled graph

infogr.am line graph

Now, I’d like to point out that I spent nearly an hour trying to get these as best as I could. I spent a long time on their support pages thinking that these could not legitimately be the only options available to me. There must be some hidden menu that I just couldn’t find.

But here we are. No control on the grid. But if I’m forced to find good points, there is basic axis start and end point control and… and… and it does have a title.

Unfortunately, the y-axis labels being over the data is unforgivable in my book.

Rating: ACCOUNT DELETED

OK, so maybe I hit a flaw with infogr.am. Maybe their thing is graphics, and maybe not many people actually try to make graphs there.

Maybe the competition will be a bit more “data-centric

Visme.co

Visme.co graph

Oh.

Oh dear.

That’s a no then.

(that’s a genuine download from a site that wants $7/month or $15/month if you’d rather not share your data with the world)

Piktochart

Piktochart graph

What is this? Is this some sort of terrible joke?

The good news is that they only want $29/month.

So that’s a no from me.

The end is the beginning

I started wanting to make a simple graph. After years of using specialized software like Origin and MatLab costing huge sums of money (as well as trying a few open source options like QtiPlot), I have come to expect a certain level of control.

Now that I don’t have the budgets of huge universities for site licenses, I knew that. I knew that I would have to make a few compromises. I did not expect to have the control that I was used to. But I was not doing anything too crazy. No trend lines, no fitting, no exponential axes.

I was expecting to give up some of the fine controls. I can live if my grid lines aren’t quite the colour I wanted. But I wanted a graph that looked good. I wanted a graph that I wouldn’t be ashamed to put on a website with my name in big letters at the top.

After an entire afternoon of trying to get a graph that didn’t suck, you know where I was.

Global Infant Mortality

Back in Excel.

But I still had one problem. I haven’t used Excel since it moved to the ribbon interface, so I then spent a good hour trying to find the menus that I knew existed, somewhere…

About the Author

Morgan Bye

Facebook Twitter Google+

Morgan Bye is a British science writer, out of Vancouver, Canada, dedicated to helping scientists better communicate their science. He has a Masters in Biochemistry, a PhD in Biophysical Chemistry and even spent time as a scientist at an Israeli research institute.

Title image is adapted from Tim Geers via Flickr under CC.