Decolonial Data Practice: What It Actually Looks Like in the Field

By Community Programs Team

A practical look at how Datum Africa's community programs approach representation, language, and context in African data documentation — and what we've learned.

Decolonial data practice has become a phrase that means many things to many people. In academic circles, it can refer to broad frameworks for thinking about knowledge production and epistemic justice. In policy discussions, it often appears as a commitment to community consent and data sovereignty. Both of these uses have value.

But we work in the field — in the daily practice of building and maintaining open data — and in that context, decolonial practice is not primarily about frameworks. It is about a set of daily choices that either preserve or erase context, that either center or sideline African knowledge systems, that either build or erode trust with the communities whose data is being documented.

Here is what those choices look like in practice.

When we document a dataset, we choose whose vocabulary to use in the metadata. If a dataset records crop varieties grown by smallholder farmers in Northern Nigeria, the standard approach is to map those varieties to international agricultural taxonomy codes. This is useful for interoperability. But it erases the local names, the local knowledge, and the relationships between varieties that exist in the practical knowledge of the farmers themselves. Decolonial practice means including both — the international taxonomy code and the local name, with a note explaining the relationship between them.

When we build data stewardship programs, we choose whose expertise is valued. Data stewardship has historically been framed as technical work requiring formal training. Our programs treat contextual knowledge — knowing the community, the history, the language, the ecological conditions — as equal in value to technical metadata skills. This changes who can participate and who produces the most useful contributions.

When we evaluate dataset quality, we choose what counts as complete documentation. Standard metadata quality frameworks measure completeness against a checklist of fields: title, description, license, format, date. A dataset that passes this checklist but is documented only in English, with no contextual notes about collection methodology or community context, scores as complete. We do not think this is adequate. Completeness, for us, includes contextual documentation that makes the data meaningful to the communities it concerns.

These choices are not dramatic. They do not require confronting anyone. They require attention, persistence, and the willingness to do more work than the standard frameworks require. That is what decolonial data practice actually looks like, from where we sit.