Preventing Bias in Nonprofit Data

Data & Studies

Before any fundraising campaign, it is prudent to conduct a feasibility study, which is a formal process where an organization conducts research, talks to stakeholders, and analyzes current and past data to determine if there will be a successful outcome.¹ The feasibility study is shared with management, fundraisers, and sometimes even major donors, all with the purpose of determining if a project is worth undertaking. Specifically, feasibility studies research future prospective donors, assess fundraising capacity, and, most importantly, work with stakeholders (beneficiaries and donors) to make sure the community is treated with respect and priority.² Therefore, it is essential these studies are conducted without bias.

In most major industries, data is thought of as the pinnacle of source information when making predictions. In fact, humanity generates 2.5 quintillion bytes of data each day, and with current technology, it is easier than ever to turn those numbers into qualitative answers.³ However, when people think of using data to make predictions, the majority of that work occurs in finance and computer fields. In fact, McKinsey & Company predict that companies that embrace data sharing for finance could see GDP gains between 1 to 5 percent by 2030.⁴ This piece of information brings up two important points: one, the nonprofit industry could greatly benefit by using data; and, two, what is this “data,” where is it coming from, and what is its trustworthiness?

In Our Sector

In the nonprofit sector, data typically refers to stakeholder responses, economic or health statistics, or a variety of other indicators that determine community well being. Typically, this data is stored in a CRM (Constituent Relationship Management), that sorts through all of this data, links it to donors, beneficiaries, and subgroups, and then delivers insights to fundraisers.⁵ But, how do we know this data is unbiased? How do we know we can trust these insights? This thought process leads to some interesting points:

Data that has been biased by collectors does not paint an accurate picture of the community being served.
AI programs that run through data and their algorithms are historically biased based on the information they have been trained on.⁶
Selection and confirmation bias by evaluators can consistently influence outcomes of surveys and who is selected to give data.⁷
People from marginalized communities, who may not speak the native language, have economic means, or are facing political persecution have a harder time reporting data and working with nonprofits, eliminating the most vulnerable from the pool.
With those who truly need help oftentimes unable to communicate their circumstances, what data are we getting, and does it paint the whole picture?

What Can We Do?

Ultimately, the answer to that last question is “no, we are not getting the full picture”; however, there are ways we can work towards a more equitable landscape in data collection. For starters, we must continue to promote primary accounts of people from marginalized communities, where information is direct and first hand. Focusing on primary stakeholders is the bedrock for any research in a campaign, but there is also much more to do, including:

Have somebody from the community being interviewed consult on the wording and phrasing of the study.
Create focus groups from the community prior to the campaign to get reliable, qualitative feedback.
If there are borders to cross, communicate with organizations who have long been in the country and work directly with people in need.
Ensure that the use of AI is only based on internal data, as the large scale data that most AI platforms are trained on are historically biased.
Have third party audits to ensure objectivity in evaluations.

There is significantly more that nonprofits can do to help with this issue, and it will be very interesting to observe the future of data privacy and bias as we continue to see AI make its way into our lives. But, at the end of the day, we must ensure that data is there to complement a diverse story and contains primary accounts from the community.

References

Donorly. (n.d.). Fundraising feasibility study. https://donorly.com/thedonorlyblog/fundraising-feasibility-study
Graham Pelton. (2024, May 18). Capital campaign feasibility study. https://grahampelton.com/capital-campaign-feasibility-study/
Harvard Business School Online. (2020, October 1). Data driven decision making. https://online.hbs.edu/blog/post/data-driven-decision-making
McKinsey & Company. (2021). Financial data unbound: The value of open data for individuals and institutions. https://www.mckinsey.com/industries/financial-services/our-insights/financial-data-unbound-the-value-of-open-data-for-individuals-and-institutions
Bloomerang. (2021, April 27). How nonprofits are using data to do more good. https://bloomerang.com/blog/how-nonprofits-are-using-data-to-do-more-good/
Independent Sector. (2023, August 8). Data privacy, misinformation, and algorithmic bias: AI challenges for nonprofit organizations. https://independentsector.org/blog/data-privacy-misinformation-and-algorithmic-bias-ai-challenges-for-nonprofit-organizations/
Charity Digital. (2021, November 10). A guide to eliminating data bias. https://charitydigital.org.uk/topics/a-guide-to-eliminating-data-bias-10267

CoffeyWritesImpact