Techniques charities can use to get insights from their data
Tue 30 Jun 2020 | Giselle Cory

DataKind UK is a charity that supports social change organisations to use data and data science, through a large network of pro-bono data scientists. In this article, Executive Director Giselle Cory discusses some of the methods used by social change organisations in the UK and around the world.
The UK social sector is vast. There are around 170,000 charities and nearly 500,000 social enterprises in the UK. These organisations support millions of individuals and countless communities – yet most of them aren’t making use of data science.
They are limited by a lack of resources, the high expected salaries of data experts, and fears about misusing data. However, some brave organisations are leading the way! Below, we highlight projects by organisations at the vanguard of social sector data use.
Computer vision to find villages in need
DataKind worked with non-profit GiveDirectly, which provides financial aid to people around the world. They identify some of the poorest households in rural Kenya and Uganda and send money via mobile phone transfer – a simple, radical idea that has been proven to work.
But it is both difficult and labour-intensive to identify which villages are most in need. So GiveDirectly turned to data science – and DataKind! – for help. From experience, GiveDirectly knew that the material each family uses for their roof is a proxy for their level of poverty. If families have the money to do so, they invest in a metal roof. Those who can’t afford metal make thatch roofs.
DataKind worked with GiveDirectly to use satellite imagery and machine learning to identify, on a village-by-village level, the proportion of thatch and metal-roofed homes. The result was a powerful proof of concept, showing the potential of these techniques in development and humanitarian work.
Using predictive models to ensure the most in-need to get support first at a food bank
The Welcome Centre is a food bank based in Huddersfield, UK. They provide support to people in crisis, offering practical help in the form of food, toiletry, and household support packs. For those who need it, a support worker can provide advice on underlying problems, and help them avoid becoming dependent on the food bank. Identifying those most in need of support, who are therefore most likely to become dependent, is challenging, and currently done manually by the support worker.
DataKind UK and The Welcome Centre partnered to build a system that could identify a client’s likelihood of needing additional or longer-term support. The aim was to create a probability score that would aid the support worker to decide, in conjunction with other information, whether a client is likely to need extra support. Using this information, The Welcome Centre improved the accuracy and efficiency of the targeted work that the support worker undertakes, enabling them to make earlier interventions before a crisis escalates.
Finding insight from text data to better understand the experiences of vulnerable families
Buttle UK gives out around 10,000 small grants a year to some of the most vulnerable children, young people, and families in the UK. These grants fund basic goods that most of us take for granted, such as beds, cookers, and fridges.
Buttle UK holds a huge amount of data on the grant applications they receive, including short written reports by support workers describing the families’ circumstances. Using NLP, DataKind assessed common sequences of requests for support, with a particular eye for which requests tend to lead to others. For example, families who receive grants on the grounds of domestic violence often also apply for grant support with child health and development problems later on.
Analysis of text data has traditionally been off-limits for many social change organisations, due to the technical expertise required to get insight from the data. As the social sector increases its access to data expertise, NLP will become more commonplace. For more on how the social sector can use NLP, see here.
Segmentation to understand outcomes among young homeless and vulnerable people
Computer vision, predictive modelling, and text analysis are considered exciting techniques by many data scientists. But it’s often the more foundational techniques that can be the most useful – and lead to the most impact – for social change organisations.
Welsh charity Llamau supports young homeless people and vulnerable women. They wanted to better understand who benefits from their services. Among the techniques used on this project was segmentation analysis. It showed the considerable variation between outcomes for different segments of Llamau’s beneficiaries, with less positive outcomes for people who were male, ex-youth offenders, or previously in the care system. When an individual fitted into two or more of these categories at the same time, the likelihood of a successful outcome dropped even further. These insights helped Llamau to improve the services they provide to individuals.
Using network analysis to uncover corporate corruption
Global Witness is an NGO that campaigns to end the environmental and human rights abuse that is driven by corruption and the exploitation of natural resources. DataKind UK partnered with Global Witness to better understand networks of corporate ownership by analysing a newly opened register of UK companies. This information has the potential to lift the lid on chains of corporate ownership and uncover webs of corruption that were previously far more difficult to investigate.
Together, the project team created a network graph that could be used to discover ‘red flag’ activities that might indicate nefarious behaviour. They found that thousands of UK companies are owned by other companies in tax havens – potentially unlawfully. Some of these tax-haven-owned companies are also in receipt of government contracts. Read more about the project here.
Get involved
If you’d like to get involved with DataKind UK, please get in touch!
- If you’re interested in offering your time and skills as a data scientist or expert, please read a bit more about what we’re looking for and submit your information here
- If you know of a social sector organisation that needs support, please share our information and application form here.
- We exist because of the donations of philanthropists and sponsorship from companies. If you’re interested in supporting our work as a donor or sponsor, please get in touch: [email protected]