Daily CSR
Daily CSR

Daily CSR
Daily news about corporate social responsibility, ethics and sustainability

Maximizing social impact using data science


Data science can be a daunting topic for those working in the philanthropic sector. Where do you begin? What information is critical? What role does it play in your community work?

Before you get too far into the weeds, take a step back to understand what data science is and what it can do for you.

Data science for social impact is about organising the unstructured in order to make data-driven decisions. It's not about inundating you with information you don't require. It's all about using data strategically to make a bigger difference.

In collaboration with Chantal Forster and the Technology Association of Grantmakers, we hosted a webinar. Three data scientists working in the social impact sector—Antonio Campello of the Wellcome Trust, Jonathan Hertel of the Impact Genome Project®, and Andrew Spott of Submittable—discussed how data science and artificial intelligence help them achieve their goals.

We've outlined nine ways data science can help your organisation do more, based on insights and examples from this event.

We've got some useful insights for everyone, whether you have a strong background in data science or you're just getting into data.

Some definitions of data science
Before we begin, let's go over some terminology. For those new to data science, the language itself can be intimidating, so here's a quick primer.
  • Data analytics is the process of analyzing raw data to uncover trends and provide answers.
  • Most businesses use data analytics in some capacity. You might wonder, for example, where our grantees live. Data analytics at work is tracking geographic data to determine the top states or countries to which you provide funding.
  • Machine learning is the process of leveraging algorithms to enable computers to automatically improve through experience and data.
  • Data science is the application of machine learning algorithms to various data sources (such as numbers, text, images, video, audio, and others) to create artificial intelligence systems capable of performing tasks that would normally require human intelligence.
  • Image recognition is a common application of machine learning; if your phone or computer has ever identified who is in a photo on its own, you've seen machine learning in action.    
  • Even if you've never worked in data science, you've almost certainly benefited from it. Data science is being used in every sector, both public and private, including healthcare, transportation, and public safety. For example, data science is now being used to assist in the analysis of medical imaging and the identification of malignancies.
  • Natural language processing is a branch of artificial intelligence that focuses on teaching computers to understand and respond to text and spoken words in the same way that humans do.
Consider Siri or Alexa.
  • Taxonomy is a system of classification or categorization used to group similar items.
  • In philanthropy, nonprofits may be classified according to their size, mission, program design, location, or another factor.
Get the guide before you dive in
Get this comprehensive guide to developing a data strategy for your social impact programs to learn more: A Guide for Social Impact Professionals on How to Build Your Data Strategy. Access online or save a printable version for later.

Gain a more detailed understanding of your portfolio.
Leveraging data and social impact analytics can help grantmakers gain a better understanding of who and what they fund, as well as the outcomes of that funding. Trends, overlaps, and gaps in resource distribution can be revealed.

What is the significance of this?
Having a more granular view of your programs allows you to identify and eliminate bias as well as uncover missed opportunities. Furthermore, it can provide you with the insights you require to expand your efforts and create effective new programmes.

You most likely already have a lot of the information you require. However, if that data is not organised and accessible, it is meaningless. A large amount of unstructured data makes it difficult to see patterns or draw conclusions.

For example, the Wellcome Trust, whose programs are primarily concerned with health, examined its portfolio of over 100,000 grants. They added 30,000 disease tags to their portfolio using natural language processing. This is a critical step in preparing data for analysis and segmentation. It enables grant managers to easily search previous grant programmes. If the organisation is thinking about funding research on a specific disease, they can now look back and see if they've previously funded that focus area and what the results were.

Even if you don't have a portfolio of 100,000 grants, having a more detailed understanding of your programs is critical.

Automatic tagging is one way to begin organizing your data. Submittable, for example, supports automatic tagging, allowing you to apply a tag to a grant application based on answers to specific questions. This gives you the ability to track, categorize, and analyse data that is important to your team, whether it is demographics, geography, mission focus, organisation size, or something else.

You can use those tags to sort your applicants and grantees to help you analyze how you distribute your resources and whether you need to change your strategies.

Establish more precise standards and taxonomies
In the world of social impact, we often talk about the same things but in different ways. This can make it difficult to see where our work overlaps with that of others and to compare similar programs.

For example, when grantmakers, nonprofits, and researchers all use different terminology to discuss the same issues, it's much more difficult to draw connections between their work.

This disconnect makes it difficult for everyone to develop effective solutions and make accurate predictions. Consider it the foundation of a house. If we don't have a strong foundation, everything built on top of it will be unstable.

Within Mission Measurement, the Impact Genome Project is working to systemize these standardizations. For example, they have developed an Impact RegistryTM, a searchable database containing impact and beneficiary information on over 2 million nonprofit programmes in the United States and Canada. This type of tool can help to strengthen that foundation, allowing the philanthropic sector to more effectively build towards data-driven solutions.

Clearly track outputs and outcomes
The goal of philanthropic work is to make a difference. Are the programs you run producing the desired results? If you don't know this, no matter how big your budget is, you won't know if your program is making a difference.

Data analytics can help you track the effects of your work more precisely. You can draw parallels between the resources you devote to a program, the short-term outputs it produces, and the long-term results achieved.

The Wellcome Trust, for example, funds a significant amount of research and clinical trials. Their team wanted to monitor the academic outcomes of their grants. In other words, they wanted to know how many academic articles were produced as a result of the programs they supported.

The Wellcome Trust team was able to create an automatic pipeline to aggregate research articles related to the programs they funded using data science techniques. They were then able to trace this work back to specific grants in order to determine which programs produced the most academic outputs.

Consider your programs and the outputs and outcomes that are most important to your organization. Are you getting clear, detailed information about how they are linked? Do you know which programmes consistently yield the best results? Leveraging data science for social good can help you see how everything fits together.

Determine funding gaps
Data analysis can assist grantmakers in identifying funding gaps. Are you underfunding certain groups? Are you overlooking any aspects of the work? Investigating the data can assist you in identifying opportunities to do more.

It's easy for funders to get caught up in the aspects of community work that feel more personal or emotional. On an individual level, it feels good to help someone in need. That empathetic response, however, can sometimes skew how organisations build their programmes. Strong infrastructure is required for social impact work. Supporting these less exciting but critical frameworks can often enable organizations to accomplish much more.

You can use data science to identify funding gaps in your portfolio or the philanthropic sector as a whole. Too often, prevention and infrastructure are given insufficient attention.

Wellcome Trust, for example, used data science to analyse their portfolio and determine how frequently they funded technology across their programmes. Using these insights, they shifted their priorities in order to fund more grants for software infrastructure, which is a critical tool for many researchers.

Make new contacts
It's easy to overlook the numerous ways in which subjects and identities overlap and intersect. However, these insights can have a significant impact on how you design your programs.

Grantmakers can use data science to identify opportunities to connect disparate topics or projects. Data visualization tools allow you to see where connections already exist and where connections should exist but do not.

If we don't understand these intersections, it will be difficult to develop effective solutions to large problems. Take, for example, climate change. In complex ways, the effects of climate change overlap and intersect with the effects of poverty and racial injustice. If your organisation attempts to address climate change without taking this into account, your programs will most likely have a limited impact.

Do you want to see how this looks? The Wellcome Trust collaborated with their neuroscience team on a topic modelling project to identify data-driven research topics. They created a map using data science to analyse text documents in order to visualise the connections and gaps between topics.

Place your program within the context of the larger picture
Many organizations and individuals are attempting to address the same set of issues. Frameworks such as the United Nations' Sustainable Development Goals can assist in providing a high-level understanding of how your work fits into larger efforts to effect change. However, it does not provide you with the necessary information.

You want a clear picture of how your programs are contributing to larger-scale progress. Are you assisting in moving the needle? Do your programmes duplicate other efforts, or do you play a distinct role?

Data science can help you find these answers and distribute your resources more strategically.

The Wellcome Trust, for example, used a combination of API searches, web scraping, and natural language processing to assess how their work fits into larger efforts to combat COVID. They measured how frequently the research was mentioned in scientific evidence from government agencies and acknowledged in general literature about COVID to assess the effectiveness of the research they funded. This gave them a clear sense of whether the programmes they supported were contributing to the larger effort.

Assess the program's effectiveness and cost.
Balancing program cost and effectiveness is an important aspect of grantmaking. You want to know that your resources are not only producing the desired results, but that they are doing so as efficiently as possible.

The stakes are extremely high. We're not talking about a profit margin or shareholder dividends here. Your program costs frequently dictate how many people or community organizations you can assist.

Data science enables grantmakers to gain a more in-depth understanding of how outputs and outcomes are related and how they compare to costs. It's impossible to know whether you're making wise investments if you don't have this information.

Impact Genome, for example, focuses on financial health. They use data science to assess program effectiveness and how program costs compare to program outcomes. They analyzed 990 tax records for all nonprofits in the United States using machine learning. They can then compare budget size to outcomes achieved.

Of course, there are numerous strategies and program designs for addressing financial health. For example, you could assist someone with emergency expenses or provide financial counselling. These are likely to have different costs and outcomes.

The Impact Genome team conducted a meta analysis to identify the outcomes that represent financial health and then outlined the associated costs.

Consider how useful this information could be to nonprofits involved in financial health. They could compare their expenses and outcomes to the national averages. Are they spending significantly more money without getting better results? Or do they compete with other non-profits? That is critical information for them to have as they plan their future programs.

Make your review process more automated.
Data science can assist grantmakers in reviewing a large volume of applications. Some organisations replace open-response questions with multiple choice questions to achieve clean data (data that has been processed and formatted to eliminate inaccuracies and inconsistencies). This may make it easier for people to review, but it takes away an applicant's ability to explain their work on their own terms.

You can keep those open-response questions and spend less time manually reviewing applications by leveraging machine learning. You can also achieve more consistent and less biassed results. How does this function?

The Accelerated Review feature of Submittable puts data science into action. Submittable uses machine learning to create a model based on the reviews performed by your team. In a nutshell, you're teaching the computer to review applications in the same way that your team does.

What does the procedure entail?
Your team evaluates approximately 200 applications. The model is based on learning how your reviewers score and rate applicants. Before applying the model in its entirety, it is tested for accuracy on a small set of applications. The model can be retrained if necessary.

By automating your review process in this manner, you will be able to review thousands, if not millions, of applications at incredible speed and accuracy. This reduces fatigue and human error significantly. Simply put, automated review allows your team to do more work with fewer people by leveraging machine learning to highlight important information.

Reduce the financial burden on grantees
With a growing awareness of the administrative burden often placed on grantees, data science can be an incredible tool for grantmakers to identify opportunities to lighten the load.

For example, the Technology Association of Grantmakers recently determined that a random grant form is 39% similar to any other grant form using machine learning and similarity analysis.

What is the significance of this?
This means that many organizations frequently redo the same work when applying for different grants. What if grantmakers banded together and standardized parts of their grant applications to ensure that everyone was requesting the same information in the same format? Consider how much time this would save grantees. 39%!

Putting ideas into action
Data science is the application of technology to extract insights and structure from data. However, in order to effect change, people must put their insights into action.

You may lack the resources and capacity to form your own in-house data science team. That's fine; Submittable has your back. We can provide all of the benefits of a data science team at a fraction of the cost as a social impact platform designed to assist organizations of all sizes. Find out more right now.