Candid updates its taxonomy with community input 

Earlier this month, Candid released a new version of its taxonomy, the Philanthropy Classification System (PCS). Think of the PCS as a system of “tags” that Candid applies to its data to make it more searchable and usable. This update, developed with considerable input from our partners and larger community, is one of the most significant updates to our taxonomy since we first developed it in 2015 as a replacement for the National Taxonomy of Exempt Entities (NTEE). We believe these changes will provide nonprofits, funders, researchers, and other users with a deeper understanding of the sector and are excited it can now be put to use. 

What are the benefits of Candid’s updated taxonomy? 

This update to the taxonomy consists of over 200 changes from the previous version, including: updates to definitions of terms, such as Human services or Researchers (the most common type of change); changes to term labels; additions and, to a much lesser extent, removal of terms; and shifts in where terms are located within the taxonomy. More information on the changes can be found on our taxonomy site, so here I’d like to highlight how we think these updates will help the sector: 

• New terms capture important areas of work obscured or ignored in the previous version of the taxonomy. Some examples of this are our new Epidemics and pandemics and Artificial intelligence subject codes—work that would have been captured only as “Health” or “Computer science”—and new People living in rural areas population code, a group that we wouldn’t have been able to capture with our previous population codes. 

• Addition of child terms will allow for a more granular understanding of an issue, including where funding is going and which organizations are working on what. For example, Candid’s updated taxonomy includes new terms for Climate change mitigation and Climate change adaptation and resilience (under Climate change); Affordable housing and Eviction protection (under Housing development); and Police reform and Prison reform (under Criminal justice system rights). 

• Updated language reflects how organizations in the sector currently talk about their work and the populations they serve. These updates include removing or replacing language now considered outdated or offensive (e.g., Dropouts or Substance abusers) or that simply made it difficult to understand what a term was meant to capture. 

• New organization types will facilitate more nuanced analyses of funding and other trends. These additions to Candid’s updated taxonomy include Universities and colleges, Hospitals, and Fiscal sponsors.  

What does it take to update Candid’s taxonomy? 

Implementing these updates to the PCS is a complex process. Part of the challenge is that any update to the taxonomy requires an update to our autocoding algorithm, one of the major sources of coding of our data, especially transactions (such as grants). To minimize inconsistencies in the coding of our data, we also recode all of our data when we make changes to our taxonomy or autocoder. At this point, Candid has collected data on millions of organizations and roughly 30 million transactions. So, what goes into flipping the switch? 

First, we developed a crosswalk between the two versions of the taxonomy. While for the most part, this consisted of mapping to new term codes or term labels, for new terms we also had to develop search strategies that would allow us to identify grants and organizations that should receive these codes. For example, our search strategy for Gender-based violence included keywords such as GBV, femicide, FGM, violence against women, and more to try to capture all relevant grants in our database.  

Once we had these crosswalks, we applied them to our data. Our Data Science team now had what it needed to train a new version of the autocoder, one that could not only predict the new PCS terms but also correct known issues with the previous version (for this reason, users may notice shifts in the coding of our data). We ended up evaluating at least five models before landing on one we felt represented a significant improvement. With the new version of the autocoder in place, the final step was to apply it to previously autocoded data.  

Where will you see Candid’s updated taxonomy? 

As we’ve previously shared, Candid is working to improve our tools. This includes app.candid.org, which will bring together the data currently found in products like GuideStar and Foundation Directory. We have listened to feedback to expand access to the full picture of nonprofits, the grants they receive, and the funders that support them—all in one place. This latest version of our enhanced taxonomy will power Candid’s future. Data in GuideStar and Foundation Directory continues to be updated, but these will reflect the previous version of the PCS. 

As noted above, this is the most extensive update to the PCS to date—due largely to the more open process we used to seek feedback. While such a consultative approach has its challenges, we believe it leads to a better outcome. So, we’ll be considering ways to make this process even more open the next time we update the PCS, likely in three to four years. Many people inside Candid spent considerable effort making the transition. However, I’d be remiss to not once again thank the many people outside Candid who contributed to this update. We don’t consider the PCS only Candid’s taxonomy; we hope it will become the sector’s. So to all who shared their input on how we could make things better, thank you!