We work in an industry that is passionate about design. And for us professionals, there is no better feeling than presenting something we know has good design. Our design process goes a bit like this:
- research brief and identify objectives
- create questionnaire or moderation guide
- collect responses
- build reports.
Chefs also follow a similar pattern:
- get inspiration for a dish
- develop a recipe
- prepare the meal
- plating the dish to restaurant standards.
Unfortunately, our ‘plating’ of data is not up to restaurant standard. Errors are made in the analysis process, too much dressing is thrown on the charts, and there’s overkill in repetition. Data and charts are like wine and food – there are some fantastic pairings (I’m thinking red wine and pasta, champagne and oysters) and others that should be avoided at all costs (like chardonnay and chorizo, red wine and chocolate). If a chef’s dish featured these combinations, it would leave a bad taste in your mouth.
So what charts go well with certain types of data? How can we improve presentations so our results look as good as the cool infographics we find in a Google Image search?
Data visualisation assumptions to challenge
I’m crazy for data, and have spent several years working on data visualisation. I believe I have found three key principles you can follow to improve your charts. But before I introduce them to you, I need to challenge a few assumptions you may hold.
Assumption 1: No two recipes are the same
The first assumption that I’d like to challenge is the belief that all recipes are the same. Let’s say you want to impress your friends who are coming over to dinner and you decide to make macaroons (a fancy name for biscuits). You could download a recipe from the internet and make them, but they won’t be the same as Adriano Zumbo’s macaroons.
To make macaroons like Adriano Zumbo, you’d need to spend time learning his approach, gathering the relevant ingredients, understanding why he does what he does. It’s the same with data. We can’t just pull a questionnaire off the internet and expect it to produce dazzling insights to guide our decision making. Information for information’s sake is not very useful. We need to carefully construct our questionnaire first to ensure we are solving the right problems with our data analysis and subsequent charting. Data can show us many things: satisfaction, NPS, task completion rates, etc… but are they relevant to the problem we had in the first place?
Assumption 2: You should follow the recipe step-by-step
You think you know the drill: gather ingredients, prep work, cook and ta-da! It’s ready. But charting data is not just about following a sequence of steps. The first time I made apple pie ice cream, I gathered my ingredients, did all the prep work, made the base… then I had to wait four hours until it set in the freezer. Had I considered what was involved in the recipe – in its entirety – before starting, I wouldn’t have cooked my apples so early on (they ended up sitting on the bench for a long time!). Similarly, you may think that research follows the same steps: go out and collect data, analyse it, then chart. But the problem is that you’ve assumed you’ll have enough time after analysis to think how you can best approach and present your results.
Assumption 3: Plating is just being fancy
Another mistake you may make is to think that plating is just another word for being fancy; that it doesn’t really matter how things look, it’s the insights that pack the punch. (I mean, what’s the big deal about how mushrooms are presented on a plate – they’re good for us, so our kids will just eat them, right?) We, as researchers, can hold the same assumption about charting – we think we can just use tables and all this talk about data visualisation is not really important. Well, this is wrong.
I wonder if you’ve heard of Anscombe’s quartet? It’s a table of four datasets that looks like this:
As you can see, the mean scores for each dataset are exactly the same as are the correlations. So they are all telling the same story right?
But in fact, when we plot them, they each tell a different story. It’s so much easier to see the patterns in the data when they are charted compared to being tabulated.
Data visualisation principles to follow
So with two assumptions challenged, I’d now like to introduce to you my three principles of data visualisation.
Principle #1: Accuracy
You’ve seen it before on all the top cooking shows – a group of amateur cooks are given a difficult recipe and there’s 30 minutes on the clock to replicate what a Michelin-star celebrity chef does in half a day. Somewhere in between commercial breaks, nearly every contestant has ‘that’ moment – where they deviate from the recipe. It results in disaster (cue the ominous music and facial close-up) – and Manu starts asking, “Where is the sauce?”
The trouble stems from the fact that the contestant’s dish isn’t accurate. No matter how we decide to chart our results, we must ensure that the information we present is correct. Accuracy is fundamental; it will make or break our credibility (cue the ominous music and facial close-up). Without credibility we’d have a client like Manu asking, “What is your source?”
We achieve accuracy when we stick to the recipe given below:
- Ensure that the analysis has been done correctly
- Ensure that the chart used to present the data is appropriate
Here are three charts that would result in me being kicked out of Datachef in the first few episodes:
The percentages do not add up to 100% on the left pie chart (they equal 122%).
The axis on the two bar charts are not the same – resulting in a very small difference (.4) looking the same as a big difference (4.6).
There is a discrepancy between the data (25.8%) and the chart (100%).
I mentioned earlier that data and charts are like wine and food; there are some fantastic pairings and others that should be avoided at all costs. Some of those great pairings are:
- Bar charts used to make comparisons between groups
- Column charts used to rank categories from largest to smallest
- Line charts used to track variables over time
- Pie charts used to show how much (the percentage) of something in relation to the whole (100%)
Principle #2: Simplicity
Have you ever tried to replicate a Heston Blumenthal recipe? His roast scallop recipe (scallop tartare, caviar and white chocolate velouté) in The Fat Duck Cookbook (page 175 for the adventurous among you) contains no less than 34 ingredients including Chardonnay vinegar, Amalfi lemons, grelot confit and langoustine oil. Contrast that with Curtis Stone’s grilled scallops recipe in Relaxed Cooking, which has just seven everyday ingredients.
Heston’s a big inspiration to me and I am always amazed at his creations, but truth be told I’d make Curtis’ dish most nights. It’s not that Curtis’ recipe is better than Heston’s, it’s just easier to follow and still achieves great flavour.
Simplicity is not about making things so basic that your charts appeal to everyone. Simplicity means to reduce what Edward Tufte calls chartjunk – all that formatting stuff that isn’t really necessary to convey the message of the data. Consider the chart below (it is pretty much a standard bar chart from PowerPoint). Without changing the data, what, if anything would you change/remove to make it better? I can think of eight things:
Here is a before and after shot of the exact same data, just taking away the chartjunk (no extra analysis). Aside from looking better, the message is clearer: the southern cities (Melbourne and Adelaide) have an older customer base than the northern cities of Brisbane and Sydney.
We should keep only things that are essential to make the insights shine and the simplest way to get your message across. You don’t necessarily need 34 ingredients to make scallops shine, just seven can do it.
Principle #3: Innovate
Almost every week, we have fettuccine on the menu at home with my special sauce recipe. It’s a fantastic combination of tomatoes, basil, marjoram, oregano, curry, cumin, bush tomato, chilli, garlic, onion, spinach, brown mushrooms and a good splash of red wine. The flavour combinations hit the spot every time and we all love it. But every now and then we deviate from this earthy flavour combination and try something different.
The last time innovation took hold we made a Bloody Mary sauce with crabmeat, onion, garlic, tomatoes, lemon juice, vodka, tabasco, breadcrumbs and flat-leaf parsley. It’s been a hit ever since. We still make our special sauce, but now we have another family favourite.
Innovation doesn’t necessarily mean you have to scrap all your hard work and create something new. It does mean, however, that you counteract repetition with novelty. Instead of presenting the same chart over and over again (or the same pasta sauce), ask yourself: “Could this be presented any differently?” Or when you see something unique, ask yourself: “That chart looks interesting, how can I use it in my work?” But keep in mind the words of Randy Krum: “Including a chart doesn’t make a design interesting or memorable. The visualisation must be unique and impactful… just seeing a presentation full of similar bar charts can put the audience to sleep.”
There are two ways to implement innovation in the kitchen or presentation. The first is to keep a library of recipes (visualisations) that you like. When you have the opportunity, try to replicate them. You will generally find that you need to tweak the recipe a little because you can’t get your hands on cilantro – or it just takes longer for your oven to warm up (analogous to PowerPoint not having certain functionalities to replicate your chosen visualisation).
The other way to implement innovation is to really get to know your capabilities. When someone says they are allergic to dairy, you think you can’t make chocolate ice cream right? Well, that’s not true. Use soy milk instead of cow’s milk, use coconut cream instead of double cream, and use dark cocoa powder. Voila! You now have dairy-free chocolate ice cream that tastes better than normal chocolate ice cream.
Similarly, you might think that because you work with PowerPoint, you’re limited to only a handful of charts (bars, lines, area, pie charts and scatter plots), right? Again that’s not true. All the charts below can be created in PowerPoint. The most probable reason why people don’t do more with PowerPoint is because they just don’t know what its capable of.
The judges’ final say
Cue music and facial close-up. Zoom in on a few jubilant tears for making it to the end of the episode. Whether you won Masterchef by sticking to the exact recipe or trying something new, there was method behind the madness:
- You carefully considered every step involved in making your dish before diving in
- You acknowledged that presentation affects how people will perceive, receive and understand the ingredients
- You ensured all measurements (and steps taken to achieve the result) were 100% accurate
- You appreciated simplicity and kept unnecessary ingredients out of the mix
- You knew what tools were at your disposal that enabled you to innovate without disaster.
Go forth and good luck charting your data.