The Measurement Layer

How fragmented is aid in Uganda? Is it getting worse? Are donors coordinating? How much goes to climate? Each question seems answerable with IATI data: query the datastore, count the results, get a number. Each answer has a measurement layer that shapes it as much as the underlying reality does.

Across five analyses of IATI data covering East African aid flows, we found that every one of these questions is mediated by methodological choices invisible to the end user. The number you get depends less on what is happening on the ground than on how you define, count, and filter your data.

This is not a criticism of IATI. The data is genuinely valuable for transparency — it lets anyone see who reports doing what, where, with what money. But transparency data and analytical data serve different purposes. When the first is used as if it were the second, the gap between them fills with systematic distortion.

Same data, different answers

The chart below shows the central finding across all five analyses: when you change how you measure, the answer changes — sometimes dramatically. Each row compares what the data appears to say with what closer examination reveals.

Five demonstrations

1. Counting organizations is not measuring fragmentation

Ask IATI how many organizations work on governance in Uganda and you get 85. Decompose that number and roughly one-fifth disappears: duplicate names, completed activities still in the database, and work miscoded to the wrong sector. The remaining organizations span at least three distinct domains that don’t create coordination costs for each other. Across 12 country-sector pairs, the average artifact rate was 21%. Full analysis →

2. Counting organizations over time is not measuring trends

Every one of ten country-sector pairs shows apparent fragmentation growth from 2010 to 2024 — an average of +155%, with no pair declining. But IATI’s publisher base grew 2,051% over the same period, 14 times faster. The data cannot tell us whether fragmentation is actually increasing or just becoming more visible. Full analysis →

3. Structured coordination fields cannot measure coordination

IATI has fields designed to capture coordination: multi-funder activity types, participating organization roles, collaboration types. Across 4,756 activities, the multi-funder field was used zero times. The “accountable organization” role was filled for 28% of activities. Meanwhile, 387 basket fund activities and 98 coordination references were detectable only through free-text keyword search. Full analysis →

4. Half of apparent fragmentation growth is reporting expansion

Cross-referencing fragmentation entrants with IATI publisher registry dates showed that roughly 50% of organizations entering each country-sector pair after 2015 had never published IATI data before. They didn’t start working in the sector — they started reporting. Publisher growth and fragmentation growth share the same timeline, the same acceleration phase, and the same plateau. Full analysis →

5. The measurement method determines which donors count

Three standard approaches to identifying climate finance — DAC sector codes, keyword searches, and Rio markers — find almost entirely different activities. In Kenya, only 7% of the combined set was found by all three methods. USAID, the largest environmental donor by activity count, has 0% Rio marker fill; the Green Climate Fund and Germany achieve 100%. An assessment using Rio markers systematically counts European contributions and misses American ones. Full analysis →

The common structure

These five findings are not independent problems. They are expressions of a single structural feature: IATI was designed for transparency, not for the analytical questions commonly posed of it.

Transparency data answers: “who reports doing what, where, with what money?” Analytical questions ask: “how fragmented, how coordinated, how much climate finance?” These are different questions. When the first is used to answer the second without decomposition, the gap fills with systematic distortion — artifact inflation, reporting confounders, empty fields, and method-dependent counts.

The measurement layer is not random noise. It has structure: it inflates fragmentation (by counting duplicates and historical records), obscures coordination (by leaving structured fields empty), and makes quantities method-dependent (by offering multiple inconsistent classification systems). Someone who trusts the output number is trusting a methodology they cannot see.

What would help

1. Organizational identifiers. Replacing free-text names with persistent IDs would eliminate the deduplication problem at source.
2. Mandatory activity status maintenance. Requiring publishers to update status — or auto-closing after an inactivity threshold — would prevent historical records from inflating current counts.
3. Structured coordination fields. Making coordination mechanism fields functional, with “not coordinated” as an explicit choice, would make coordination visible to structured queries.
4. Consistent marker completion. Requiring Rio markers on all activities, with “not targeted” as an explicit option, would distinguish “not tagged” from “does not apply.”
5. Cross-method validation as standard practice. Any quantitative claim derived from IATI data should be tested against at least one alternative measurement approach. If two methods produce substantially different numbers, report the disagreement rather than picking one.

Scope and limitations

This analysis covers five East African countries, three sectors, and one additional domain (climate finance). Whether the exact magnitudes hold elsewhere is an empirical question. But the structural features that produce the measurement layer — free-text organization names, archived activities alongside current ones, optional coordination fields, inconsistent marker completion — are inherent to IATI’s data model, not specific to this region. The measurement layer is present everywhere IATI data is used for analytical purposes. The question is not whether it exists, but how large it is.