The Hidden Cost of Missing Metadata: Evidence from 500 African Datasets

By Datum Africa Research Unit

Our analysis of 500 community-contributed datasets found that missing or incomplete metadata reduces discoverability by an estimated 60% — with non-English datasets most affected.

In 2025, the Datum Africa research team analysed 500 community-contributed datasets from across sub-Saharan Africa, focusing specifically on metadata completeness and its relationship to discoverability and use.

The findings were stark. Datasets with complete metadata — defined as having title, description, license, format, date, and at least one keyword — were discoverable via search in 89% of cases. Datasets with incomplete metadata were discoverable in only 29% of cases. That is a discoverability gap of 60 percentage points, driven entirely by metadata quality, not dataset quality.

The gap was not random. It was systematically concentrated in datasets contributed by non-English-speaking communities, by smaller institutions, and by individual contributors without formal data documentation training. Datasets from national governments and large research institutions had significantly better metadata completeness — not because their data was better, but because they had staff whose job included documentation.

What is lost when a dataset is not discoverable? In the best case, it is simply underused — it sits in a repository, never found, never analysed, never contributing to the decisions it was meant to inform. In worse cases, the absence of community-contributed data from discovery means that decisions about those communities are made using only the data that large institutions have chosen to collect and document — which is never the full picture.

The hidden cost of missing metadata is not technical. It is epistemic and political. When African community datasets are undocumented and undiscoverable, the knowledge they contain is effectively invisible to researchers, policymakers, and platform users. The communities that contributed the data receive none of the benefit of having contributed it.

The interventions that worked: Our analysis of the 2025 Data Stewards cohort contributions showed that targeted metadata improvement improved discoverability scores by an average of 43% per dataset. The most impactful single intervention was adding a plain-language description of collection methodology. This alone improved discoverability scores by 28% on average.

These findings are driving Datum Africa's 2026 program priorities. We are expanding the Data Stewards cohort, developing a multilingual metadata documentation guide, and working with platform partners to surface incomplete datasets for community improvement. The data infrastructure problem is solvable. The question is whether we invest in solving it.