The White Ribbon Alliance (WRA) is a Washington, DC-based charity with the goal of making sure every woman survives childbirth, worldwide. WRA is supported by a range of celebrities from Naomi Campbell to Sarah, Duchess of York.
The WRA ran a survey, called What Women Want, of over 1 million women asking them about their top requests for maternal and reproductive healthcare. The survey focused on eight countries: Mexico, Cameroon, Nigeria, India, Pakistan, Malawi, Uganda, Kenya and Tanzania. The responses came back in text format (instead of categorical responses), in a variety of spellings and even languages, which made the survey difficult to analyse on a global scale.
The majority of text responses were manually translated and categorised by the WRA team and the remainder could be assigned to topics using deep learning algorithms.
Fast Data Science was then able to run a number of analyses of the survey responses, identifying common trends across countries and age groups. For example, Water, Sanitation and Hygiene (WASH) is a common request in some regions, and using natural language processing we were able to drill down to identify what kind of WASH facilities were lacking in individual regions.
The team at White Ribbon Alliance have been able to use the results of our analyses in their policy briefs and plan to present findings to the relevant national governments and the UN.
We have also been designing a public-facing dashboard, allowing anybody to drill down into the data and explore the individual responses, view them on a map, and generate graphs dynamically.
To our knowledge, the dashboard is a first for transparency at non-profits, allowing donors, policymakers and employees alike to analyse the results of a charity’s research in detail.
Until now, surveys have typically dealt with numeric or categorical information. This has been the case in market research as well as in the non-profit sector. Survey respondents are asked whether they “strongly disagree”, or “strongly agree” with something, often on a 5-point scale. These numbers are relatively simple to put into bar charts or pie charts, analyse, and visualize interactively. But this kind of survey has its limitations. What happens if somebody wants to say something that wasn’t in the original question?
In the What Women Want campaign, the White Ribbon Alliance and their local country affiliates asked over a million women an open-ended question: “What is your one request for quality reproductive and maternal healthcare services?”, and they responded in their own words, rather than agreeing or disagreeing, or choosing from a list of options. When we received the survey response data, we faced the challenge of displaying over a million free-text responses in a way that laypeople can view and understand them at a glance. Since nobody can read a million responses, we had to find a clever way to group the responses into topics.
We have used some cutting edge technologies from AI and data science, such as deep learning and natural language processing (NLP), to categorize the responses. Sophisticated AI models such as the state-of-the-art transformer neural network BERT, and a convolutional neural network called Spacy, learnt to recognize the 39 topics and have automatically assigned the women’s responses to the correct category. This was complemented by extensive human coding and quality control by White Ribbon Alliance.
Now, for the first time, members of the public can see and understand the WRA’s front-line work, wherever they are in the world. Are you interested in how many Cameroonian women above 40 are interested in blood, or what Nigerian women see as important within the broader category of Water, Sanitation, and Hygiene (“WASH”)? Do you want to see what words were frequently mentioned together (co-occurring) with “condoms”? The answers are at your fingertips. We think that the dashboard is setting a new bar for transparency in the third sector. When White Ribbon Alliance and their international network embarked on the What Women Want campaign, policy decisions in the focus countries had been usually taken and influenced by dignitaries and officials. Women in local communities have often been given very little voice. In patriarchal societies, systems such as democratic voting can result in women’s voices being a conduit for what men want.
The caseworkers on the ground in countries such as India, Pakistan, Nigeria and other countries set out to ask local women about their healthcare needs, and expected initially to hear a relatively narrow range of healthcare complaints. They were surprised by the variety and diversity of requests. Stories came flooding in about disrespect, domestic violence, FGM, and other concerns. In some regions, the main issue was a lack of access to clean water, which can lead to typhoid fever and cholera. The women also shared positive stories about what was going well. The open-ended format of the question meant that the variety of issues raised was much larger than what we would normally expect with a survey of this scale.
The local workers on the ground sent the survey response data to the global headquarters of White Ribbon Alliance in Washington, DC. Advocates from WRA have brought the findings of the campaign to policymakers in the respective countries. To date, the What Women Want campaign has resulted in more than 20 policy changes across the globe, especially in sexual health issues, reproductive issues, and water, sanitation and hygiene.
Surprisingly, the What Women Want campaign has had an impact reaching beyond the third sector. White Ribbon Alliance received enquiries from market research agencies and commercial ventures who are embarking on marketing campaigns targeting women in the countries in question, and who have seen potential in the rich goldmine of data that has been laid open in the dashboard.