A picture tells a thousand words.We’ve all heard this so many times that it has become a cliché. And we have perhaps started to believe that the picture itself doesn’t matter all that much. If you are finding it hard to explain the story in your data, why not just graph it? Everyone will immediately understand.
The problem is that many people will not understand. The real question therefore becomes which thousand words will your picture tell, and will it be the story that you intended? This has been an issue from the very first days of dashboards, and it continues to be a problem with the use of self-service visual analytics, and the rise of the citizen data scientist. All pictures are not created equal.
We already know that there is a role for IT and data scientists in ensuring that self-service analytics can be used effectively. They have to ensure that the data available is clean and in a usable format. They also have to supply a suitable analytics package, one that is relatively easy to use, and allows users to personalise their inputs and outputs to suit their needs. But is there a role for them in helping users to present their data so that it tells the right story? I think so.
Getting visual analytics right
Give users a great visual analytics package, and they can still struggle to present their data effectively. The first question that needs to be asked is ‘what story are you trying to tell?’. This reminds users that the output needs to make sense to the individual looking at it, and should therefore guide them through the ‘story’. Users should be encouraged to remove anything that does not directly affect the story.
It is also possible to make mistakes aesthetically when presenting the output from visual analytics packages. These mistakes include:
- Having too much information in the output: too many graphs, too many different formats, or simply too much information in one graph. All these are distracting and make it harder to see the main points.
- Overdoing the visual effects: people often joke about ‘death by Powerpoint’, and warn against using special effects, but too much colour or emphasis in a visual analytics output can have the same effect. Again, it detracts from the importance of the key messages.
- Putting information in the wrong order: in cultures that read from left to right, the first place the eye goes is the top left. It therefore makes sense to place the most important information there, so that it is visually dominant. It is surprising how often the most vital information is down at the bottom right.
- Not enabling easy comparisons: for example, it is much easier to compare two line graphs when they are placed one above the other, rather than side by side. Making the reader’s life easier is important in ensuring that they are fully engaged.
- Over-estimating attention spans: perhaps the most fragile of all elements, and often overlooked. Visuals that are too demanding in terms of engagement can perform less effectively than those which are kinder to the audience
IT teams and dedicated data scientists may feel that business users should know this already. But business users are not necessarily designers or even very strong on analysis. They may also find it hard to navigate the analytics package to change and amend the output. Help and support are likely to be both welcome and necessary.
Why go the extra mile?
This support goes above and beyond that usually ‘expected’ from IT teams and data scientists to business users. Indeed, many IT teams might argue that business users should even take responsibility for ensuring the data is clean and usable. And yes, in a way they would be right: the data belongs to business users, so they are best placed to ensure it is correct. But if they have no incentive to do so, why would they? The incentive lies in using the data, and seeing how important data quality can be.
And that is really my crucial point. Unless business users really adopt self-service analytics, they will not see its benefits, and unless they see the benefits, they will not adopt it. It is both a vicious and a virtuous circle, and only support from IT, and from dedicated data scientists in particular, can break it and/or build it.
This means that data science teams need to supply whatever support is necessary. Creating citizen data scientists is a long-term project. In the meantime, they need support to enable them to develop their skills. Getting it wrong now is likely to be very costly in terms of ‘hearts and minds’.