What IATI Data Tells Us About Aid Fragmentation — And What It Doesn’t

How many organizations work on governance in Uganda? According to IATI data, the answer is 85. That number is concrete, citable, and wrong — or at least, deeply misleading.

When we systematically decompose it, roughly one-fifth of the apparent fragmentation disappears into data artifacts, and the remainder describes not one coordination challenge but several distinct ones operating in different spaces.

This piece walks through what happens when you take IATI fragmentation numbers at face value, what emerges when you don’t, and what the data genuinely cannot tell us. The analysis covers 12 country-sector pairs across five East African countries (Uganda, Kenya, Rwanda, Tanzania, Ethiopia) and three sectors (governance, health, education), using data from the IATI Datastore API.

The seduction of the count

Fragmentation is a persistent concern in development effectiveness. The Paris Declaration, the Accra Agenda for Action, and the Busan Partnership all highlight the coordination costs imposed when too many actors operate in the same space. IATI data makes these costs appear measurable: query a country and sector, count the reporting organizations, and you have a fragmentation indicator.

The numbers are striking. Across the five East African countries in this analysis, governance sectors average 67 reporting organizations. Health averages 48. Even education, often considered less crowded, averages 43. These figures seem to confirm what practitioners have long suspected: there are too many actors chasing too few coordination mechanisms.

But the count is doing something subtle. It treats every organization equally regardless of scale, ignores whether activities are still running, accepts sector classifications at face value, and cannot distinguish coordinated action from uncoordinated duplication. Each of these creates a specific measurement distortion.

What decomposition reveals

We applied a multi-layer decomposition to each of the 12 country-sector pairs, examining four sources of distortion: organizational deduplication, activity status, sector miscoding, and thematic conflation.

Layer 1: The same organization, counted twice. IATI data frequently records the same organization under variant names. The UK’s development agency appears as both “UK - Foreign, Commonwealth Development Office (FCDO)” and “UK - Foreign, Commonwealth and Development Office” — a comma separating two records that describe one funder. After deduplication, the 12 pairs lose an average of 2% of their organization count.

Layer 2: History mistaken for presence. IATI is an archive as much as a registry. It retains records of completed, cancelled, and suspended activities alongside current ones. When we filter to active and pipeline activities only, Uganda’s education sector drops from 53 to 36 — a 35% reduction. Across all 12 pairs, filtering to active organizations reduces the count by an average of 19%.

Layer 3: Labeled governance, doing forestry. Sector miscoding is subtler and more consequential. IATI activities are tagged with DAC sector codes, but these sometimes reflect the implementing channel rather than the work content. CISU, a Danish umbrella organization for civil society groups, appears in the governance data of every East African country in our sample. But their activity titles reveal projects in forestry, HIV/AIDS, education, and agriculture. They are coded as “democratic participation and civil society” because civil society organizations carry out the work — not because the work concerns democratic governance.

Our keyword-based miscoding detector flags an average of 8.2% of activities across the 12 pairs as potentially miscoded. This creates a specific policy risk: if a governance coordination group convenes based on IATI data showing 85 organizations, it will spend time accounting for actors whose work is not governance at all.

Layer 4: One sector, three coordination challenges. The most consequential distortion is thematic conflation. “Governance” is not one sector — it encompasses at least three distinct domains with different actors, different government counterparts, and different budgets.

A civil society support program and a public financial management reform do not create coordination overhead for each other in any meaningful operational sense. Yet both contribute to the headline figure of “85 governance organizations,” implying a coordination burden that does not exist in that form.

What the data cannot show

Financial scale is invisible for most actors. Across all 12 pairs, an average of only 11% of organizations report disbursement data. This means fragmentation analysis treats a $50,000 pilot project and a $500 million sector program as equal contributors to coordination cost. Where financial data does exist, concentration is extreme: the top three organizations account for 85–99% of reported disbursements in every pair examined.

Coordination mechanisms exist but are structurally invisible. Every country-sector pair contains evidence of donor coordination: basket funds, sector-wide approaches (SWAps), joint programs, and pooled funding arrangements. But these appear only in free-text activity titles, not in structured IATI fields. There is no standard way for an organization to say “this activity is coordinated with activities X, Y, and Z.”

We found coordination signals in every pair: 20 in Uganda’s governance sector, 62 in Rwanda’s, 25 in Kenya’s. These are lower bounds — our keyword search catches only the most explicit references.

Major actors are missing entirely. USAID’s IATI reporting is limited and inconsistent. China and most Gulf state donors are absent. Any fragmentation analysis based on IATI data is necessarily a partial picture.

What this means for practice

None of this means fragmentation does not matter. Coordination costs are real, and the proliferation of small projects in crowded sectors imposes genuine burdens on recipient governments. But IATI data, used naively, gives a misleading picture of where those costs are highest and what kind of coordination response they need.

Fragmentation metrics need decomposition before they can inform policy. A raw organization count is as useful as a raw GDP figure without population — technically accurate, practically misleading. At minimum, sector-level fragmentation assessments should filter for active organizations, deduplicate names, disaggregate by thematic cluster, and note the financial reporting gap.

IATI data quality improvements would help. Better organizational identifiers would reduce deduplication problems. Structured fields for coordination mechanisms would make coordination visible. Consistent activity status maintenance would prevent historical records from inflating current counts.

Coordination assessment should be sector-specific and context-aware. The governance sector is not “fragmented” in the same way across its component domains. Government capacity building may need a different coordination response than human rights programming, and treating them as one problem produces solutions that fit neither.