How the country picker ranks need
One figure per country.
Never summed, never guessed.
Every humanitarian crisis has several published caseload figures — OCHA's, a donor's, an NGO's own count. Adding them together feels more thorough. It isn't; it mostly double-counts the same people. Here's what we do instead, and what it deliberately leaves out.
Three choices, made on purpose
One source, ranked, never added together
OCHA's Humanitarian Needs and Response Plan figure is already the inter-agency consolidated number — negotiated across clusters specifically to avoid double-counting. UNHCR refugee stock, IOM's displacement counts and UNRWA's caseload sit largely inside it already. So each country gets exactly one figure, from the highest-ranked source that has published one: OCHA GHO, then ECHO, then ACAPS, then IRC, then agency reporting — and the register says which one it used.
Ranked against the resident population, not everyone there
The headline ranking divides each caseload by the population minus the refugees that country hosts — refugees stay in the numerator, but leave the denominator. That's deliberate, and it isn't the same question as "what share of everyone here is in need." It measures the caseload as a multiple of the host society's own size. See what that reveals below.
Hosting countries stay, even with no caseload of their own
Ranking by caseload alone made the register blind to most of the places it was built to show. Venezuela's crisis produced a caseload figure; Colombia, absorbing 2.8 million people who left it, did not — and would have vanished from the table entirely on caseload alone. Countries that host at scale but report no caseload of their own stay in, for the hosting ranking, with the gap stated rather than filled with a zero.
What the resident-population choice actually reveals
Real numbers from the current table, not illustrative ones.
Lebanon's caseload as a share of its resident population — 4.1 million people in need against 4.3 million residents once the 1.5 million refugees it hosts are set aside. On total population the same crisis reads as 71%. Eleven points either way changes which country looks worse off; that's why the choice is stated rather than assumed.
Refugees hosted for every 100 residents in Lebanon — the highest hosting burden in the table, and a different question from caseload entirely. A country can carry a heavy hosting load with a modest caseload of its own, or the reverse. Ranking by one hides the other, so both are offered separately rather than folded into a single score.
How a tool ends up matched to a country
By hazard type, not by having been used there. Sudan's crisis is tagged conflict, displacement and food insecurity; any tool in the register whose description matches one of those hazards appears as a candidate for Sudan — including tools built for a different country entirely, if the hazard is the same kind of problem. A flood-detection model built for Pakistan shows up for any other flood-tagged crisis. That is the point of matching on hazard rather than location: it surfaces precedent that a location-only search would miss.
It is precedent, not a recommendation. A match means a tool was built for a comparable problem somewhere — never that it has been tested, deployed, or verified to work in the country shown. Whether any candidate actually suits a given response is a question for the people running it, not a claim this register makes on their behalf.
Five ways to rank the same table
There is no single correct ranking — each answers a different question, and the country picker lets you switch between them.
Caseload vs. resident population
The default. People in need as a share of the resident population, refugees removed from the denominator but not the numerator — a multiple of the host society's own size, which is why it can run past 100%.
People in need
The sector default elsewhere: absolute headcount. It favours large countries by construction — Sudan and Afghanistan lead on this measure largely because they are large countries, not necessarily because the crisis is worse per person.
Caseload vs. total population
The same ratio with refugees left in the denominator. Shown so the effect of excluding them is visible rather than assumed — compare this to the resident ranking for the same country to see the choice matter.
Tools a responder could obtain
Not a need measure at all — from the register itself. Ranked least-served first: which crises have the fewest tools that are actually obtainable, not just built.
Refugees hosted per 100 residents
Hosting burden on its own terms, independent of caseload. The measure that keeps Colombia, Turkiye and Jordan visible at all.
Where this is weaker than the rest of the register
Hand-transcribed, not pulled live
These figures were typed in by hand from published reporting, not fetched from an API. They mix reporting years for different countries and are refreshed periodically, not continuously — treat this table as provisional between refreshes, the same caution the file itself states.
Some refugee figures aren't adjusted
Where a country's refugee count is marked "not recorded" rather than zero, the resident-population denominator has NOT been reduced for it — so that country's ratio is understated relative to ones where the subtraction is evidenced. The register flags this per country rather than silently treating a gap as a zero.
Selection, not a census
23 countries are in this table: the largest caseloads, plus the largest refugee-hosting countries. Real need exists well outside this list — a country's absence here is a limit of this table, never a claim that the need doesn't exist.