What type of visualisation should you use to tell your data-driven story?

The range of data visualisation tools is growing quickly, making choosing which one to use and which style to opt for even trickier.

Some tips before you start visualising

Remember that you don’t always need a visualisation to tell a data-driven story: sometimes words do it best. Otherwise, only select and visualise the key data that tells the story. You can, and should in the interest of transparency, always attribute and link to your full, raw data set.

Any data used must be cleaned throughly first, else your visualisation will end up looking amateur. Steps to take to avoid this involve: translating any jargon into terms a reader would understand; converting or appending data where necessary e.g full name to surname, postcode to constituencs; deleting mistaken repetitions; and checking spelling. If problems are hard to spot, visualising your data can make them stand out and alert you to them quickly, but cleaning first will often save time.

Regardless of the type of visualisation you make, consider the power of contrast and repetition, the latter of which makes the former work e.g. a change of colour, typeface, shape. Think about colour, particularly in charts, but don’t be tempted by the rainbow effect- changing the headline element to a different colour  from the rest can help it stand out. Certain colours complement others so look up colour schemes.

So which one to choose?

There are websites out there on hand to help you decide which type of visualisation to create. Juice Labs is user-friendly, plus they showcase a few of their favourites for inspiration. Andrew Abela’s chart suggestions map is helpful too, recommending a range of chart types depending on what story your data tells: relationship, comparison, distribution or composition.

From what I’ve learnt thus far at City, pie charts are largely useless as visualisations. The brain doesn’t interpret circles as well as bars, particularly in the likes of *shudder*, 3D pie charts. There are generally better ways of visualising compositional data.

Tree maps are great for displaying comparison and proportion, as you can create sub-categories and use bold colours. These are fairly simple to generate using tools such as Many Eyes or TreeMap.

If you want to show comparison across 3 axes, bubble charts allow you to add extra dimensions in the size and colour of the bubbles. Like a scattergram that points plots against X and Y, they can be animated, clearly displaying outliers and clusters of data. Take a look at Swedish statistician Hans Rosling plotting life expectancy vs. fertility rate. vs. population using a bubble chart here. As demonstrated by Rosling, this style of visualisation can look stunning when used effectively, but be wary of overloading and confusing your reader.

Avoid these like the plague! A 3D pie chart showing BBC broadcasting expenditure. Image: Wikimedia Commons

To display regional differences, heat maps are hard to beat. These can be created with Google Fusion and Carto DB among others. Choose bands carefully and ensure that gaps are properly identified- work out quantities in Excel to help, using the code ‘=quartile(cell:cell, 1)’ adding your own column range where it states ‘cell’ and changing the number according to the quartile you wish to calculate. Hot colours like red, orange and yellow are usually used for ‘hotspots’ of a certain thing e.g. casualities, shootings.

Network diagrams visualise the connections between people, while the more traditional idiograms use images to represent some sort of data. The latter suffer from the same pitfall as pie charts in that having to mentally compare different shapes only really works when large differences are involved.

Much maligned and cliched they may be, but word clouds can be useful for comparing and getting an overview of key themes in a speech, report, or document. These are, however, probably best used for finding leads as opposed to communicating more fully developed stories.

Journalism is unlikely to stop at the data

See your data as a brilliant way to unearth stories that nobody else has. If you can strip numbers out and bring people in it is often much more effective, as the human element is key to any great story. Think along the process to other sources and people likely to be affected, breaking down what you need into the smallest possible detail.

Check out Paul Bradshaw’s inverted pyramid of data journalism to remind yourself of the need to: compile, clean, add context, combine and communicate.

Don’t get bogged down with the stats when communicating your findings- making the story accessible to readers is crucial. Instead of writing ’48% of’, write ‘nearly half of’.

Visualising your data is but one part of the journalistic process- remember that while your focus is on telling a story in a new way, it is still first and foremost on telling the story.