How these scores are built, and what they cannot tell you.

Every tracked problem carries five scores from 0 to 100. Each is a calibrated judgement against fixed anchors, not a raw figure — a death toll of ten million and a death toll of ten thousand cannot share a linear axis and still leave the middle of the list legible.

Loss of life · default weight 30

Direct and attributable mortality, normalised across annual and cumulative tolls.

Lives negatively impacted · default weight 25

People displaced, injured, food-insecure, exposed or otherwise materially harmed.

Financial cost · default weight 20

Direct damage, lost output and health-system spend attributable to the event.

Environmental impact · default weight 15

Emissions, habitat and biodiversity loss, contamination and irreversibility.

Trajectory · default weight 10

Whether the situation is escalating, and how much worse it is plausibly getting.

The headline severity score is the weighted mean of those five factors, using the house weighting of 30 / 25 / 20 / 15 / 10. Loss of life counts most, trajectory least. That choice is editorial, not scientific.

Because it is a choice, the sliders on the index let you replace it. Move loss of life to 50 and everything else to zero and you get a pure mortality ranking; weight the environment heavily and slow, diffuse damage rises past acute emergencies. The URL carries your weighting, so any ranking you produce is a link you can send to someone who thinks you have it wrong.

0–34 limited, 35–49 serious, 50–64 severe, 65–79 critical, 80+ catastrophic. Bands are colour-coded consistently across the site.

High — figures come from institutional reporting with an established methodology, such as UN agencies, WHO, IPCC or major statistical bodies, and the range between credible estimates is narrow.

Medium — credible figures exist but estimates diverge materially between sources, or the count is partial (for example, direct deaths counted but excess mortality unknown).

Low — figures are contested, extrapolated, actively suppressed, or drawn from early reporting that will be revised. Most news-derived entries begin here.

Because credible sources routinely disagree, often by a wide margin, and usually for a good reason rather than a careless one. Two published death tolls for the same war can differ several-fold depending on whether they count direct deaths or excess mortality, where the counting window starts, whether they use verified incident reports or statistical modelling, and whether the counting body has access to the territory at all. The same is true of “lives impacted”, which depends entirely on where the threshold for impact is drawn, and of economic cost, which depends on whether indirect and future losses are included.

This index does not average those estimates into a false consensus, and it does not claim to have found the one true number. It picks a defensible published estimate, notes its basis, and labels the entry’s confidence so you can see how firm the ground is. So a different figure elsewhere usually means a different definition, not a mistake here.

What is worth reporting on an entry: a source link that no longer works, a figure that has been superseded by a newer published one, or an entry filed under the wrong category or region. Reports are read in batches and are not answered individually.

The index started as roughly a hundred researched entries covering the standing global problems — wars, disease burdens, famines, environmental collapse, structural economic harm. Those are labelled Curated and carry baseline figures drawn from institutional sources.

A scheduled sweep then searches world news, clusters headlines into events, and either matches them onto an existing entry or creates a new one. Anything created that way is labelled News-derived. The distinction matters: news-derived figures reflect what reporting supports this week, not a settled count.

When a sweep touches a curated entry, the researched baseline is treated as a floor. A news figure replaces it only when it is higher — reporting revises tolls upward far more often than downward, and we would rather understate movement than let a partial headline count erase a researched total.

After each sweep, every tracked problem's rank under the house weighting is stored. The ▲ / ▼ marks on the index compare today's rank against the most recent snapshot at least seven days old. Movement reflects changes in the underlying scores and the arrival of new entries above or below a problem, not a change in how bad it feels.

The “what changed” digest is generated from that same snapshot diff plus the record of which figures were revised during the sweep. It is assembled from the data, not written by hand or embellished.

The news pipeline can publish routine movement on its own, but not the claims that would most damage this index if they were wrong. A brand new entry is held back from the public list — invisible until a person clears it — if it lands at severity 45 or above, claims a thousand or more deaths, claims a hundred thousand or more displaced, or falls in the conflict category, where attribution is contested by design.

Changes to existing figures are held on the same principle. A revision has to clear both a proportional and an absolute line before it is treated as a claim rather than noise: deaths must move by at least 25% and at least 500; lives impacted by 25% and 100,000; displaced by 25% and 50,000; economic cost by 30% and one billion dollars. Below those lines a figure updates straight away and is listed in the change digest. Above them it waits in a review queue, and the published figure does not move until it is approved. A first-ever value for a previously empty field is always held, because it is a new claim rather than a revision.

Entries confirmed by a person carry an editor-checked date. Every entry, checked or not, carries a report control — if a figure looks wrong, it goes to the same queue.

The scorer researches a problem that is not yet ranked and estimates where it would land. Those results are explicitly provisional: they are produced from a single research pass, they are not stored in the index, and they should be read as an indication of magnitude rather than a placement.

Deaths are counted inconsistently worldwide; excess-mortality estimates and confirmed counts can differ by an order of magnitude. Financial costs mix direct damage with lost output. Chronic, diffuse problems are systematically better documented in rich countries than poor ones, which biases any global comparison. Slow catastrophes are undercounted relative to sudden ones because news coverage drives what gets revised.

This is an index of reported impact. It is a tool for arguing about proportion, not an authoritative ledger of suffering.