Using Visualization for Exploring Data – Part of the MEME Approach

Data Visualisation

Using Visualization for Exploring Data – Part of the MEME Approach

August 17, 2023
·
minute
read
Want to see more?
Register here.

To maximize the benefit of visualization, you need to focus on the task you are aiming to achieve; the visualization approaches for monitoring are different to those for exploring, and different again from managing and explaining – as I outlined in my post about MEME (Monitor, Explore, Manage, and Explain). In this post, I will look at using visualization for Explore, showing how you can use visualization approaches to identify and highlight the story in the data.

 

What do we mean by explore?

There are essentially two ways to explore data. The first is to seek to understand the nature of the data and the opportunities it provides. This is akin to an explorer visiting a new land for the first time and seeking to understand and survey the terrain. This exploring is a divergent approach; starting from one point, the process opens out to provide a wider view. The second approach is to search for the answer to a specific question. This second approach is a convergent approach, starting with a wide set of information and seeking to narrow down to the key point(s).

 

Note, exploring is a process achieved by the person exploring. It is not a spectator sport, the process and the visualizations used are not necessarily going to be seen by anybody else. The searcher/explorer is not seeking to create presentations or debriefs at this stage, they are seeking to gain understanding.

 

Dashboards for exploring

In many cases, a dashboard can be used for exploring data. The potential limitation of a dashboard is that it tends to be a collection of previously organized views of the data (sometimes referred to as widgets). If these previously configured views are sufficient for the exploration, then a dashboard can be an effective way of exploring data.

 

To illustrate using a dashboard to explore a topic and to use a dashboard to search for an answer I have created two short videos. Both videos use the FT’s Coronavirus Tracker dashboard. The first video (Explore with FT Tracker Dashboard.mp4) accesses the FT dashboard, with the aim to get an overview of what is happening globally in terms of the coronavirus pandemic.

 

The second video (Search with FT Tracker Dashboard.mp4) searches for an answer to the question, how does the UK’s performance compare with other countries?

 

A key strength of the FT Tracker Dashboard is that it provides alternative views of the same information, which makes it more likely that the explorer will be able to configure the option they are looking for. For example, there are two charts looking at the deaths from COVID-19 over time, one which looks at actual numbers and one that looks at percentages of all deaths.

 

Beyond Dashboards

If the exploration needs to utilize new views of the data, perhaps via creating new variables, then the explorer will need to go beyond dashboards. Examples of tools that can go beyond dashboards include Excel, SPSS, most cross-tabulation packages, and Hans Rosling’s Gapminder. The use of a non-dashboard approach to using visualization to explore data is also illustrated in a video (C02 and Income with Gapminder.mp4).

 

Friction-free exploring

The key thing that visualization needs to offer to a data explorer is a friction-free environment. In the context of exploring data, examples of being friction-free are:

·   Not having to cut and paste

·   Being able to ‘undo’ actions

·   Being able to save and annotate solutions

·   Being able to swap axes (e.g. swap brands by attributes to attributes by brands)

·   Being able to swap the base, for example, total deaths, deaths per 100,000 people, and percent of all deaths

·   Being able to use filters and look at sub-groups

·   Being able to look at time series

 

Exploring does not NEED presentation-quality outputs

When you are exploring, most of the things you look at (especially in the early part of the process) will not be part of your report or final presentation. You have to kiss a lot of frogs to find your prince, so you do not want to be thinking about formatting, labeling, and details like that during the exploration – they are a source of friction.

 

Another reason you do not need presentation standard outputs during the exploration stage is the difference in audiences. In your report or presentation, you will choose the types of visualization that work best for your audience. In the exploration stage, you will choose the types of visualization that work best for you.

 

As an example of this point, here is a chart that I looked at when I was exploring a recent study that I was involved in. In this case, I only used the visualization options that were available in the data collection/crosstabulation platform.

 

I find spider/radar charts are a good way for me to explore data efficiently. By running these charts in the data collection platform, it was easy to explore different sub-sets, using filters. However, I also know that these charts are not liked by most clients.

 

If this data (or part of this data) ends up being part of my presentation, then I will use a quite different visualization, one that focuses on the message I want to explain/communicate, and which is optimized for my audience. For example, I might focus on the average percentages that thought each topic would be automated, or I might focus on the different opinions about how much automation we might see for a specific topic.

 

Note, some packages offer an option to chart everything by everything and export it to presentation-ready PowerPoint charts for you. For data exploration, this is usually not helpful – because this separates the data from the outputs, which will become a source of friction in your exploration.

 

Another option that some packages offer is the ability to export a particular view or visualization to a ledger, folder or story – this is a really good idea. The reason that this is helpful in that it lets you build a collection of potentially useful items as you go. Once your analysis is finished, you can return to your collection, decide which ones you want to use, and then decide how you are going to design your presentation visualizations.

 

Summary

When you are exploring the data, you want to minimize the friction whilst maximizing your flexibility. If you are using a dashboard, it needs to have a good range of views, for example being able to see counts and percentages and to change the time frame and the base for the information.

 

If you are going beyond a dashboard, you want an integrated solution (you do not want to be moving data from one place to another), you want to be able to use the sorts of visualizations that work for you, and you do not want to think about finished outputs at this stage – most things you look out will probably not be in your final presentation.

 

You might also be interested in

Here are some other great resources on the topic of visualization:

·   MEME – your key to creating great visualizations

·   Using Visualization for Monitoring – Part of the MEME Approach

·   Webinar 27 July - Making your data dance | The 4 Pillars of Data Visualization | APAC

·   Webinar 27 July - Making your data dance | The 4 Pillars of Data Visualization | Europe

Download now

Download
Download

*Please note, Platform One was formerly known as Potentiate