Disaster and Humanitarian Data Diplomacy: Negotiating the Numbers Behind Communities in Need

*In disaster and humanitarian data diplomacy, the numbers are the easy part. Deciding what they are allowed to mean is often the harder, and more consequential, work.*

Some of the most consequential decisions I have made in a decade of humanitarian data work were never about data at all. They were about a sentence. Whether a needs figure should be published this week or held. Whether a map should show a settlement at village resolution or stop at the district line. Whether a dataset disaggregated by group would help target assistance or quietly hand someone a targeting list of a different kind. These are not technical questions. They are diplomatic ones, and they are decided in rooms where the spreadsheet is the least important thing present.

We talk about humanitarian data as though it were a thermometer. You take the reading, you report the number, the number is the truth. In reality, every figure that leaves a crisis has passed through a series of negotiations. With the host government, over what the data implies about its competence and control. With affected communities, over whether being counted will protect them or expose them. And with the wider world, over how much attention the numbers should attract and at what cost. I have come to think of this as disaster data diplomacy, and it is the part of the job that almost never appears in a methodology note.

The diplomacy starts before a single record is entered, with a question that sounds academic and is anything but. What are we going to call this, and what are we going to count?

In the aftermath of a sudden emergency, multiple actors arrive with multiple datasets and multiple definitions. One agency's "affected population" is another's "people in need." A government's official figure and a cluster's assessment can differ by an order of magnitude, and neither is simply lying. They are measuring different things, for different purposes, with different thresholds. My job, repeatedly, has been to sit in the middle of that and negotiate a shared picture that everyone can live with, knowing that whichever numbers we settle on will travel far beyond the room and do work none of us fully control.

This is why the humanitarian principles of neutrality and impartiality are so much harder in practice than on paper. The moment you decide which data sources are credible, you have made a political choice about whose account of the disaster counts. In one operation, the most complete records sat with the national authority but carried the authority's framing. In another, the more independent picture came from open-source satellite analysis but lacked the ground truth only local responders had. Choosing between them, or more often, reconciling them, is not a neutral act of data cleaning. It is a negotiation over whose version of events becomes the official one.

If there is a single number that concentrates all of these tensions, it is the People in Need figure, the count at the centre of every humanitarian appeal and intersectoral analysis. I have argued elsewhere that what gets counted decides who gets funded. The People in Need figure is where that logic becomes a live negotiation.

On its face it is a technical product, built through severity scoring and intersectoral frameworks like the JIAF. In practice it is one of the most politically loaded numbers in the entire response. A high figure can be read as an indictment of the government in place, evidence that it has failed to protect or provide for its own people. A low figure can starve a response of the resources real communities urgently need. I have sat in the analysis sessions where those two pressures meet, and the honest truth is that the methodology, however rigorous, never fully escapes them.

What I learned is that you cannot resolve this by pretending the politics away. You manage it by being scrupulous about method and transparent about uncertainty, so that the number can withstand scrutiny from any direction. When a government challenges a needs figure as exaggerated, your defence is not indignation. It is a defensible methodology, clear assumptions, and disaggregation that shows your work. The credibility of the number is what protects it, and protecting the number is, in the end, how you protect the people it represents. Get the method right and you can hold the line in the negotiation. Get it wrong and the figure becomes just another contested claim, easy to dismiss and easy to ignore.

Here is the part that is hardest to convey, and the part I think about most carefully. Effective humanitarian data work in someone else's country requires you to communicate need without turning that communication into a verdict on the administration in place at the time.

This is not about flattering anyone. It is about a practical reality. Response happens with the consent and cooperation of the host authorities, and a government that experiences every dataset as a political attack will, sooner or later, restrict the very data flows the response depends on. I have worked in contexts where the line between "documenting a humanitarian situation" and "embarrassing the state" was thin, contested, and watched closely. Cross that line carelessly and you do not win an argument. You lose access, and the people who pay for that loss are the affected communities, not the analysts.

So a great deal of the diplomacy is about framing without distortion. The same flood losses can be presented as "the scale of the disaster overwhelmed local capacity," which is true and protective of the partnership, or as "the government failed to prepare," which may also be arguable but ends the conversation. In Afghanistan, part of why we produced disaster and needs products bilingually, in Dari and Pashto, and shared open data directly with the national disaster authority, was precisely this. Data offered to a counterpart as a shared instrument it can use is data that strengthens cooperation, while the same data deployed as an external judgement invites the shutters to come down. The aim was never to soften the truth. It was to keep the channel open through which the truth could keep flowing.

There is a real ethical tension here, and I do not want to flatten it. There are moments when the data does implicate those in power, and concealment would itself be a harm. The skill, and it is a skill I am still refining, is knowing the difference between protective framing that keeps a response alive and self-censorship that betrays the people you serve. That judgement cannot be outsourced to a guideline. It is the diplomacy.

The second negotiation is with the vulnerable people in the data themselves, and here the stakes are not reputational. They are physical.

Do no harm is one of the oldest commitments in humanitarian work, and in the data age it has become, quietly, a technical discipline. Every decision about how granular to make a dataset is a decision about exposure. Disaggregating by ethnicity, religion, displacement status, or gender can make assistance far better targeted, and can also, in the wrong hands, become a map of where a persecuted group lives. The same categories that let us hear the people our systems were built to silence can, mishandled, expose them. Geolocating a settlement to help deliver aid can also help someone find it who means harm. The same precision that makes data useful makes it dangerous, and the line between the two depends entirely on context.

I have made these trade-offs in practice. In one displacement response, the political constraints on data sharing were such that the protective move was to lean on open-source remote sensing rather than collect and hold sensitive personal records that could not be adequately secured. In others, the protective move was the opposite: collect carefully, but aggregate before release, publishing at a resolution coarse enough to protect individuals while still steering the response. There is a well-understood "mosaic effect" in this work, where several innocuous datasets combine to re-identify the very people each was anonymised to protect, and guarding against it is now part of the basic craft. The field has matured here, with the ICRC's Handbook on Data Protection in Humanitarian Action and OCHA's data responsibility guidance giving practitioners real standards to work to. But standards inform judgement. They do not replace it. No guideline can tell you, for this group, in this place, this week, exactly how much to show.

The hardest version of this is the conflict between visibility and protection. Vivid, specific, human data drives attention and funding. It also concentrates risk on identifiable people. Every time I have chosen to blur, aggregate, or withhold, I have been trading some measure of advocacy power for some measure of safety, and I have not always found that trade comfortable. But the principle has to hold. The people in the dataset never consented to become evidence, and their safety cannot be spent to strengthen an argument, however good the cause.

The third negotiation is with the world, and it is the one we are least honest about.

Humanitarian communication runs on attention, and attention runs on emotion. The starkest number, the most affecting image, the single devastating statistic. These are what cut through a saturated news cycle and move donors. There is nothing inherently wrong with that. People who need help deserve to be seen, and dry, hedged, perfectly responsible data has never once unlocked an emergency appeal on its own. Part of doing this work well is knowing how to make suffering legible to people far away who have the power to respond.

But the attention economy has its own gravity, and it pulls against both of the negotiations above. The framing that generates the most global concern is often the one most likely to antagonise the host government or to expose the most vulnerable. The story that funds the response can be the story that complicates it. So the practitioner sits at a three-way junction, balancing impactful communication that draws the world's attention and resources, against the diplomatic relationships that keep the operation running, against the safety of the people whose situation is being communicated. There is rarely a clean answer. There is only a defensible one, arrived at deliberately rather than by accident or reflex.

What I have tried to hold onto is that all three obligations are real, and none can be allowed to silently win. Communicate too cautiously and you fail the people who need the world's attention to survive. Communicate too aggressively and you may lose the access, or endanger the very people, the attention was meant to help. The discipline is to keep all three in view at once and to make the trade-off consciously, knowing what you are trading and why.

None of this appears in a data science curriculum. You learn it in the room, usually by getting the balance slightly wrong and watching what happens. A number that travelled further than intended, a map that showed a little too much, a framing that closed a door you needed open. Over time it stops feeling like a series of compromises to the data and starts feeling like the actual work, the part that determines whether all the careful collection and analysis ever does any good.

I have come to believe that disaster data diplomacy is a core humanitarian competency, as important as the statistics and the systems, and far less taught. We train people to build the dataset. We rarely train them to negotiate what it is allowed to mean, to whom, and at what risk. Yet that negotiation is where data either serves vulnerable people or quietly fails them.

The numbers, in the end, are the easy part. Anyone with the right tools can count. The real skill is knowing what the count is for, whose hands it will pass through, and what it might do when it gets there. A disaster figure is never just a fact. It is a fact with consequences, and somebody has to take responsibility for them before it is ever released. For a long time now, in a lot of difficult rooms, that somebody has been me. The work was never really about the data. It was about the diplomacy the data demanded.