Visualisation as storytelling: adding ‘process’ to the ‘pretty’
The prevalence of data visualisation has grown enormously in the last 5 years - in part due to the increase and availability of big data, open data and open source software and tools. Visualisations now, rather than supporting text-based stories with evidence, appear as the evidence themselves, telling the story with graphical representations of numbers and words. In this interpretative process, there is a danger of misrepresenting the data to fit the story, so a clear decision-making process needs to be in place to ensure the right visualisation techniques are used to accurately represent the data, explore your research findings and create a compelling visual narrative.
We recently used visualisation techniques to explore responses from an online survey conducted as part of IDS and CCHI research on Transgender peoples and livelihood options in Vietnam. This blog focuses not on the outputs of our research but attempts to share learning on the complex process and journey through visualising online survey data.
Why we chose to use data visualisation and present a ‘data dashboard’
The research looks to answer questions about different employment options and preferences for Vietnamese transgender men and woman and the links between stigma, education and employment. The research looked to engage with civil society, especially LGBT groups, around transgender people and their career options. For public discussions and engagement with transgender communities it is important that data are presented in an accessible and visual manner.
As this was the first time we had conducted such a survey we really didn’t know what kind of data we would receive, who would respond and what themes would emerge from the responses. We were also unsure of the quality of the data we would receive; without good data, a visualisation would be pointless. It would have been very easy to produce a number of unfocussed charts and diagrams that represented the data, but didn’t clarify or explore any of the research questions. As there was no historical data to compare, I decided to present the responses as a ‘data dashboard’, giving us a snapshot of the livelihood options for transgender peoples in Vietnam.
We conducted an online survey in Vietnamese, using the survey tool, SurveyMonkey.
Key stages in my data visualisation process:
- Clarify the research questions that we want to answer - Data visualisation is about storytelling, so I got involved at an early stage to be clear about what the stories were that we are trying to investigate and tell.
- Designing the survey – I had input in the survey design to make sure it provided the possibility for answering these research questions in visual ways.
- Define the purpose of the visualisation - Visualisation can be used for multiple reasons; to powerfully communicate ideas and concepts, to provide quick insights or as a method to analyse the data. It is usually a combination of these, and was so in this survey.
- Prepare the data for visualisation - Once the survey is published and the results are collected, I cleaned and normalised the data in preparation for visualisation. As you will see below, this was by far the most time consuming and difficult part of the process!
- Explore visualisation possibilities – I analysed the data collected, looking for patterns and stories and tested the assumptions and research questions against the data to discover which areas were of most interest to visualise.
- Visualise! - Having explored the possibilities for visualisation, I decided on the dimensions to visualise and the types of visualisations I wanted to create, applying visualisation principles and good practice as I went.
Learning from the visualisation process
There are many visualisation tools available and unless one is familiar with a particular tool there will always be a steep learning curve before acceptable results are achieved. An application I had experience of that is relatively easy to pick up, but also offers highly complex options to the more advanced user, is Tableau Public. This powerful desktop application is free to download and is available for Windows and Mac. As well as having lots of powerful data analysis options, Tableau also has the ability to host online dashboards, effectively creating interactive data visualisations which can be embedded on a website or blog, so was perfect for this research.
For me, preparing the data for Tableau was by far the longest part in the visualisation process and this did not go smoothly. Multiple times, data was imported, visualisations attempted and aborted (and frustrations experienced), before I realised that the data needed to be in another, different shape for the application to ‘understand’ the data enough to be able to visualise it. The data was imported so many times (in different formats), that I can comfortably add ‘Tableau Importer – Expert level’ to my CV!
Likert scales, very common question types in surveys, are difficult to visualise, especially when trying to represent overall sentiment across multiple questions. In our survey, we had a number of questions that related to how each respondent felt in their place of work; in terms of how well they were accepted and whether their colleagues, boss or clients were aware of their transgender status. I thought that this would be a key area to attempt to visualise, as it gave a kind of ‘temperature check’ on transgender peoples in the workplace. I decided to use a divergent stack bar chart which is very good at showing the spread of positive and negative values.
As you can see below, the visualisation shows that our respondents were on the whole very positive about their perceptions of themselves in the workplace (there are more response on the light and dark green side than the orange and red). This result was surprising for us and this kind of result was a very useful tool to provoke discussion in subsequent workshops and seminars.
As ever, with other deadlines looming, there wasn’t enough time to produce all the visualisations that I wanted to, but on the whole the project showed some successes. I was able to illustrate a snapshot of transgender livelihoods in Vietnam in a visual way and the dashboard visualisations worked well to do this.
As we didn’t know what data we might receive from the survey and had assumptions, based on previous research and experience, it was even more important to adhere to the decision-making process outlined above. The visualisations have subsequently been used in workshops and interviews, provoking interesting reactions, reinforcing the power of the image and the need for integrity and validation in the visualisation process itself.
IDS and CCIHP, a Vietnamese NGO specialised in research on sexual and reproductive health, are currently collaborating on the research project 'Transgender peoples and livelihood options in Vietnam'” funded via an accountable grant from the Policy Division of the UK's Department for International Development (DFID).