«

»

8 Aspects to Avoid in Visualisation

Share

8 aspects to avoid in data visualisationIn my previous piece, I touched on the rise of data visualisation, and how it should be positioned in the organisation to achieve real insight from their data. In this piece I cover 8 aspects that should be avoided, in order to ensure that only clear and correct messages are conveyed to the business. Failing to avoid these will result in data visualisations being interpreted incorrectly, wrong decisions being made and providing visualisations that do little to further the understanding of the underlying data.

Data visualisation has become a commonly accepted way of displaying the stories contained in data.  It makes those stories easier to understand, where our “readers” do not have to spend hours analysing pages and pages of tabular data anymore.

However, in my experience, a poorly presented visualisation, can not only show misleading information, but it can also confuse business users into making the wrong business decisions. So together with the growing number of visualisation tools, the number of visualisation styles now available has also increased. But consequently, so also have the number of mistakes made when creating these visualisations increased.

Given this, let’s investigate the most important aspects that data visualisers should avoid.

Unpredictable visualisations

It’s vitally important for organisations to be consistent when presenting the data, otherwise users will need to stop what they are doing and figure out how to read each new graph or diagram before they can understand what it tries to portray. This wastes a lot of their time and hence they are not benefiting from using data visualisation at all. We always have to keep in mind that business users are only consumers of visualisations in order to do their “real” jobs.

Using standard colours, a standardised style sheet and consistent data visualisation practices can assist to avoid unpredictable visualisations. Proofing or sound-boarding new visuals with a small representative sample of typical business users is also a useful approach to ensure that any new visualisations will be appropriately received.

Displaying too much data

The more complicated your data visualisations are, the more likely you will turn your audience off – too much data displayed can easily lose your audience before you have even started. Trying to fit too much information into a single visualisation can mean that the image becomes too confusing to use. Any data unrelated to the topic under discussion gives your presentation an unnecessary cluttered look, which will ultimately result in the audience losing focus.

Remember to keep visualisations as simple as possible. Your audience want specific, relevant answers. My advice, only use the most relevant data that will actually make a positive impact on your users. The key of a good visualisation is not how artistic it is, but how effectively it can convey complex data to a wide audience.

Also keep Edward Tufte’s recommendations in mind that your visualisations should have a low ink-to-data ratio. This refers both to the amount of relevant data displayed as well as the formatting of lines, borders, text and the inclusion of images and other objects. What may seem a fun lively colourful graph to a developer, who only sees it for a relatively short while, may become a migraine-inducing bombardment of colour to a frequent user. A regular user wants the data to “speak” to him/her, not the overbearing bells and whistles.

Oversimplifying data

This point is basically contrary to the point above, in that one of best benefits of data visualisation is that we are expected to present data in a way that’s easy to understand. However, if you simplify or summarise the data too much, you could run the risk of leaving out crucial information. In some cases the data represents a good story at the summary level, but the devil lies in the detail.

So it’s best to interact with your users and colleagues and understand at what level of detail the data is meaningful to them. Then you can include all the essential data at that level in the presentation and arrange it into an appropriate structure that your audience can relate to and easily navigate and understand.

Choosing the wrong visualisation

I have noticed over the last few years that one of the most common mistakes in data visualisation is people actually tends to choose the wrong ‘visualisation type’ to present the information in.

There are many different ways to present and analyse data, and each of these requires different analytics and paradigms to use – this means that each presentation may need a different graph type to achieve a specific outcome.

This is where the teachings of Stephen Few are very appropriate, both in terms of graph types to use and on the styling and presentation of visualisations. Stephen advocates clarity and simplicity to the n-th degree, which reality results in very clear and useful visualisations.  Stephen’s website is at www.perceptualedge.com.

Messing With Convention

One of the reasons that visualisations have become more popular is that the general public now has a better understanding of how to read graphical representations of data. Excel graphs are being taught to kids in school, sport results are shown as graphs, and even popular survey results are displayed as infographics. You find these in Time magazine and even in the local newspaper. People have become accustomed to what certain types of visualisations are used for – like bar graphs to compare numbers and line graphs to represent measurements over time.

However, if you deviate from the norm, it becomes confusing and harder to read and analyse. So you need to keep the common conventions. For example, Y axes must start from the bottom and X axes from the left. These axes typically start at 0. For the most part, logarithmic and inverse scales should only be used for very sophisticated audiences, or where their purpose can be explained. Even so, it must be clearly annotate as well. Changing the conventions is going to have an impact on the simplicity that data visualisations represent.

Not Annotating Visualisations

This one seems like a no-brainer, but you will be surprised how many times necessary annotations are simply left off. This is one area where that extra little bit of ink is necessary to explain the meaning of the data.

For example, every axis needs to be labelled to indicate what it represents, and in which units of measurement. Having one big bar and one small bar representing growth may seem obvious, but without the labels to indicate the extent of this growth, can only be confusing.

Very important, if you are starting any axis at any point other than zero, it is important to make sure that the representation show this clearly. You cannot expect your audience to guess that you’ve changed the axis. As mentioned above, the same applies when you use unconventional scales, like a logarithmic scale.

Difficult Comparisons

Visualisations are often used for comparisons. Comparing this month to the same month last year, comparing actuals to targets, comparing performance against a competitors or the market at large, and so on.

When you need to compare two sets of data in a visualisation it is important to make the comparison process as simple and intuitive as possible. Creating an complex visualisation that resembles a “spot the difference” game is not going to be well received.

This means keeping comparable graphs close to one another, making them easier to compare by using contrasting colours and always make sure that the annotations are in the same places and in the same size and fonts. Helping people to understand which pieces of information they should be paying attention to and comparing is going to make this process much simpler. Side-by-side bars are easier to compare that two bar graphs next to each other. Different coloured timelines on the same graph are also good for comparisons.

Pie charts should never be used for comparisons – the human eye cannot compare the sizes of pie segments with any accuracy. In fact pie charts should never be used. Fullstop. You can read more about this here too. The same applies to 3D graphs – they simply are not good for comparisons – or for anything else for that matter. I always proclaim the only thing a 3D graph is good for is for a meaningless book, article or web page cover. In contrast, the human eye can compare the heights of bars horizontally or vertically very well.

Publishing Mistakes

As soon as your numbers do not add up, or there is any error on a visualisation, your credibility is instantly done. Even the validity of the underlying data will be questioned. Even if a single measure is out by a margin as little as 1%, it will bring doubt on all other figures and visualisations.

With the many visualisations being published to the business, the users are almost looking for reasons why they should not be using particular ones – especially new ones.

It is therefore crucial to test your visualisations and validate their correctness. Of particular importance is when you have table calculations across bars or rows or graphs, combined with filters. You need to make sure any totals and percentages are for the data shown, and not for the entire population. If it applies to more than what is shown, you should a) really question why you are doing that, and b) really annotate it very clearly.

Concluding remarks

So be careful – don’t go ahead and do great data analysis, but then make any of the 8 simple mistakes  listed here, that in turn can lead to your audience not actually understanding or being to make good use of what you are producing. Worst case is if you break their trust by providing inappropriate or incorrect visualisations.

Reference

Some of this material was adapted from the post titled 5 Common Mistakes In Data Visualisation.

 

 

Leave a Reply