Collecting Identity Data Isn't Decoration

TL;DR: Identity data — the questions about race, gender, disability, language — gets quietly labeled “optional,” so we skip it. But when we don’t count people, we don’t protect them; we make them invisible. Data equity doesn’t start with better dashboards. It starts with who we choose to make visible, and how much care we bring to the asking.

Somewhere in almost every nonprofit database, there is a set of fields nobody wants to touch. Race. Gender. Disability. Sometimes income, immigration history, primary language. They sit there, half-filled or empty, quietly labeled optional. And every time an organization decides not to collect them, it tells itself a comforting story: we’re protecting people. We don’t want to make anyone uncomfortable. We’d rather not ask.

I understand the instinct. I’ve felt it myself. But I’ve come to believe that “we’d rather not ask” is one of the most consequential decisions a mission-driven organization makes — and we almost never treat it like a decision at all.

What do we actually mean by “identity data”?

Identity data is the information that describes who someone is, not just what they did. A gift amount is behavioral data. A program attendance record is behavioral data. But whether the person who gave that gift or showed up to that program is a first-generation immigrant, a wheelchair user, a Punjabi speaker, a trans woman — that is identity data. It’s the part of the picture that tells you whose experience your work is actually reaching.

Here’s the uncomfortable part. If you don’t collect it, you don’t become neutral. You become blind. Your data still describes a population — it just quietly defaults to whoever was easiest to count.

A small example of what invisibility costs

Picture a health outreach program that never asks which languages its participants speak. On paper, everything looks fine: attendance is steady, the numbers are up and to the right. But because the question was never asked, no one notices that an entire non-English-speaking community stopped showing up two years ago — not because they didn’t need the service, but because the flyers, the forms, and the follow-up calls were never in a language they could read.

The data didn’t lie. It just never contained the people it was quietly failing. That’s the thing about invisibility in a dataset — it doesn’t announce itself. It looks like success right up until someone thinks to ask a question the system was designed not to hold.

So why do we avoid collecting it?

Usually for reasons that sound like care. We worry about privacy. We worry about tokenizing people, or flattening a whole life into a checkbox. We worry about what we’d be obligated to do if the data revealed a gap we couldn’t yet close. And sometimes there is a real, specific fear — of a funder, a board, or a political climate that has made the word equity itself feel risky to write down.

These worries aren’t wrong. They’re just incomplete. Because the alternative to collecting identity data badly is not “not collecting it.” The alternative is making decisions about entire communities while quietly pretending they’re not in the room.

What does collecting identity data with care look like?

This is where I want to slow down, because “just collect more data” is exactly the wrong lesson to take from any of this. Care is the whole point. In practice, it tends to look like a handful of commitments:

• Explain why you’re asking, in plain language, before you ask.

• Make “prefer not to say” a real, respected option — not a data-quality failure you’ll nag people about later.

• Collect only what you will actually use to change a decision. If it won’t change anything, leave it out.

• Tell people what the data will and won’t do — and then keep that promise.

• Let communities see themselves in the results, rather than only extracting from them.

None of that is technical. All of it is relational. Identity data collected without trust isn’t equity — it’s surveillance with better branding.

But isn’t it safer to just not ask?

This is the tension I sit with most, and I don’t want to resolve it too cleanly. Yes — there are real risks to collecting sensitive data. It can be breached. It can be subpoenaed. It can be turned against the very people it was meant to protect. In some contexts, the most equitable choice genuinely is to collect less, or nothing at all.

But “safer for whom?” is the question underneath the question. Not asking is often safest for the institution and riskiest for the people who then go uncounted, underfunded, and unseen. Comfort and equity are not the same thing, and it’s worth being honest with ourselves about which one we’re actually optimizing for.

Where data equity actually begins

In the AI equity work I do, data equity is the first layer — and that ordering is deliberate. Every model, every dashboard, every “AI strategy” a board gets excited about is built on top of the data an organization chose to collect, and chose not to. If the foundation renders certain people invisible, no amount of sophistication on top will fix it. It will just automate the blind spot — faster, and with more confidence.

So before we talk about AI, before we talk about tools, I’d ask a smaller and harder question: who is fully visible in your data, and who is only there by accident? That answer — not your tech stack — is where equity starts.

Identity data isn’t decoration. It’s not a compliance checkbox or a diversity statement rendered in a spreadsheet. It’s the difference between a community you can actually see and one you’ve simply assumed. And that is worth the discomfort of asking.

---

Frequently asked questions

What is identity data?

Identity data is information describing who a person is — such as race, ethnicity, gender identity, disability, language, or immigration background — as opposed to behavioral data about what they did. For nonprofits, it’s what reveals whose experiences a program is actually reaching.

Is it risky for nonprofits to collect identity data?

There are real risks — privacy, security, and misuse among them. But choosing not to collect it carries its own risk: making decisions about communities while rendering them invisible. The goal isn’t more data; it’s collecting only what you’ll use, with consent and care.

How is data equity connected to AI?

AI systems inherit whatever the underlying data includes and excludes. If identity data is missing or biased, AI tools will scale that blind spot at speed. Data equity — collecting and stewarding data fairly — is the foundation any responsible AI depends on.