In short. DESK does not collect data itself. Every figure comes from a public source or interface named with it and on the Sources page. This page lists everything DESK calculates from those figures, each with its formula. Where a rule contains a value that DESK chose (rather than one that comes from the data), it is marked as a setting.
amount in US$ = amount in currency X ÷ (units of X per US$)
units of X per US$ (year Y) = mean over all trading days of Y of the ECB reference rates (X per €) ÷ (US$ per €) (full years since 2015)
otherwise the World Bank official annual rate (WDI PA.NUS.FCRF, since 1960) of the currency's country
Taiwan dollar (not in the World Bank data): the Federal Reserve annual rate (G.5, since 1983)
euro before 1999 = the country's old currency's World Bank rate ÷ the irrevocable conversion rate (e.g. DM 1.95583 per €); from 1999 a euro amount is converted at the euro rate, even for a country that joined later
redenominated currency = for an amount from before the new units, the daily rate × the factor (Syrian pound 2026: × 100)
Every company figure of a year is converted at that year's rate — Wikidata, SEC, ESEF, OpenDART (Korea), EDINET (Japan) and German company reports alike, so that one company's series never mixes two rate bases. The daily rate (ExchangeRate-API) is used only for the current and the previous year; an amount that cannot be converted at the rate of its own year is not shown. Growth is therefore measured in US dollars: companies from different countries are compared on one currency, and growth in a high-inflation currency does not count as growth.
TTM = last fiscal year + Σ quarters after its end − Σ the same quarters one year earlier
Only consecutive quarters count, and only if the first one follows the fiscal year directly. A TTM that ends with a fourth quarter carries a whole year as its label; if a filed fiscal year already has that label (a fiscal year to June), the filed fiscal year counts and the TTM is left out.
m/m (or q/q, w/w) = value of the period ÷ value of the immediately preceding period − 1
change = value of the period − value of the immediately preceding period
Both values come from the same release day. If the preceding period was never published (October 2025 in the US government shutdown), there is no change for the period, and DESK shows none rather than a change over two periods.
reference month of a BLS release = the release month − 1 (employment, CPI, PPI)
JOLTS = the latest month that ended at least 28 days before the release
(2 June → April, 30 June → May)
nowcast = the GDPNow estimate of the Federal Reserve Bank of Atlanta for the quarter of the release, as it stood
the day BEFORE the release (upcoming releases: the latest estimate); taken over unchanged from the Bank's
workbook (GDPTrackingModelDataAndForecasts.xlsx), real GDP growth at an annual rate
A model estimate, not an official figure, not a forecast by DESK and not an analysts' consensus; the second and third GDP estimates carry none. GDPNow is a product of the Federal Reserve Bank of Atlanta and available at GDPNow - Federal Reserve Bank of Atlanta.
a country stays in the comparison if its latest value is no more than 400 days older than the latest value of the
other countries (the US and the regulated Gulf tariffs are not counted in that "latest")
Setting: 400 days.
No check on company figures uses a factor chosen by DESK. Every limit is measured, and mostly on the same reference: the mandatory filings with the U.S. SEC (XBRL, us-gaap frames since 2009 and IFRS company facts) — audited statements in machine-readable form with a clear unit. What practically never happens there is an error in a Wikidata series. The limits are measured again with every build (scripts/zaun-grenzen-bauen.js).
change over D years = |ln(value_t ÷ value_(t−D))| for every pair of years of every company
limit L(D) = 99th percentile of that change, within the company's BEA industry (at least 100 pairs),
otherwise its sector, otherwise all (measured, all companies: D=1 ×2.85, D=2 ×4.53, D=3 ×7.04)
largest jump J(D) = the largest change of revenue over D years in any filing (D=1: ×229)
Tukey fence = Q1 − 3 × IQR and Q3 + 3 × IQR of a distribution, IQR = Q3 − Q1 (Tukey 1977, "far out")
ESEF revenue: the total revenue of the report; if a report has none, the sum of the revenue types it lists (goods, services, construction contracts, royalties) — never a single part of it.
Net income and equity mean the share of the owners of the parent company in the filings DESK reads as data (SEC NetIncomeLoss and StockholdersEquity, ESEF, OpenDART, EDINET); only where a report gives no such figure does the total including minority interests count. The German reports read from PDF follow the same rule: next to the net income and equity line, DESK looks for the owners' line on the same page (up to ten lines above, eight below — "attributable to shareholders of …", "davon Aktionäre …", "Equity attributable to owners of the parent") and reads it in the same columns. It counts only with the same sign, at least 30 % of the headline and at most a quarter above it (it can be above with negative minority interests). "Shareholders' equity" counts only above total equity with a minority line in between, because some reports use it for the total. Lines for ordinary shareholders only (without hybrid capital) do not count. Without such a line the headline stays — for companies without minority interests both are the same.
Mandatory filings (SEC, ESEF, OpenDART, EDINET, the companies' own annual and interim reports) take precedence over Wikidata for the same year. A Wikidata value next to a filed value is dropped when it differs by more than L(D), where D is the distance to the nearest filed year (revenue, equity, total assets; not for years after the last filed year + 1, and not for distances that never occur in the filings). On the company page, the years from filings and the years from Wikidata are drawn as separate lines, so the seam is visible.
Which year a fiscal year belongs to: the SEC frames assign it to the calendar year it overlaps most (FedEx's year to May 2022 is 2021); Wikidata, ESEF and EDINET date it by its end (2022); the SEC IFRS company facts are read by their frame, like the us-gaap frames. Before the two are merged, the years are aligned:
Wikidata shift d ∈ {−1, 0, +1} = the d for which the most Wikidata values equal a filed value of year + d exactly
(d ≠ 0 only if it matches more values than d = 0)
fiscal year label = the calendar year of its end, or the year before if it ends January to May
(the rule of the SEC frames; used for ESEF and EDINET, and for Wikidata when no value matches exactly)
Entries with the same SEC CIK, LEI or ISIN are one company. An unlisted subsidiary counts under its parent in DESK (which reports at least as much revenue) if the parent consolidates it: GLEIF lists the parent in the chain of consolidating parents, or both have the same ultimate consolidating parent; otherwise Wikidata states the parent's share (more than half = consolidated, up to half = not); without either, investment firms (BEA 523/525) count as not consolidating. A parent and subsidiary with the same name or website are one company. A corporate group that is not itself a company (keiretsu, chaebol, umbrella brand) does not count; its members count. Of several entries for one company, the one that stands for it is, in this order: never a subsidiary in place of its parent, the entry with figures, the entry closest to the SEC filer's name, the newest year, in the same year the listed entry, then the higher revenue.
The same rule applies everywhere companies are counted or added up: sector rankings, the company map, company counts and revenue sums per country, supply chains, peers and crisis lists leave out the entry that counts under another company. Search still finds it.
The table lists every company once, with its latest figures. With a year chosen, it shows the figures of that year as reported (revenue, net profit, total assets, employees, market value) and the ratios of that year by the same rules as the history behind each figure: revenue growth against the year before, net margin = net profit ÷ revenue (at most 100 %), operating margin = operating income ÷ revenue, equity ratio = equity ÷ total assets. The DESK score exists only for the latest year. “Only major countries & companies” keeps the major economies (the same ones as “only major economies” on the country side) and, of their companies, those among the 500 largest by revenue in the year shown, the same cut as the Fortune Global 500. A country chosen in the country menu narrows the table to the companies headquartered there. Table, company list and country menu follow the same filter.
A site (subsidiary, second seat, facility) carries the country Wikidata gives for it. If its coordinate lies outside that country on the DESK world map, OpenStreetMap (Nominatim) is asked which country the coordinate is in; the simplified map does not know small territories such as Gibraltar or Singapore. The site stays if OpenStreetMap confirms its country or one of that country's overseas territories; at sea (platforms, offshore wind farms) it stays if it lies at most 600 km off its country. If the coordinate lies in another country that Wikidata also lists for the site, it takes that country. Otherwise no source confirms the site, and it is left out.
growth = (revenue_t ÷ revenue_(t−k))^(1/k) − 1 k = years back to the oldest value 3 to 5 years before t (t = latest value, including the trailing twelve months)
net margin = net income_t ÷ revenue_t (latest period with both, including the trailing twelve months; above 100 % is not counted)
EBIT margin = operating income_t ÷ revenue_t (same rule)
equity ratio = equity_t ÷ total assets_t (balance at the end of the fiscal year; above 100 % is not counted)
size = log10(revenue)
productivity = log10(revenue ÷ employees) (years at most 2 apart)
track record = current year − founding year
sites = log10(1 + sites + branches); countries = number of countries with sites; listings = number of exchanges
valuation = − market value ÷ net income (only with positive net income; years at most 1 apart)
Score, company page and DESK Radar show the same figures from one calculation: growth, net margin, EBIT margin and equity ratio are computed once for the score and taken over by the company page and the Radar, with the period they refer to. Without a value 3 to 5 years back there is no growth. The company page shows the same annual series as the score (for SEC filers the checked SEC series, before that Wikidata).
P/E = market value ÷ net income (same year; only with positive net income)
P/B = market value ÷ equity (same year)
P/S = market value ÷ revenue (same year)
dividend yield = dividend per share × shares ÷ market value (same year, SEC filers)
free cash flow = operating cash flow − capital expenditure (property, plant and equipment) (same fiscal year, SEC filers)
The market value comes from Wikidata and has its own year; each ratio shows the year it refers to, which can be older than the latest market value. DESK AI names the same ratios. Where a company reports its fiscal fourth quarter only in the annual report, DESK AI derives that quarter as the year minus the three reported quarters (revenue, net income, operating income, R&D; not EPS).
rank = share of all comparable entries with a lower value × 100 (equal values share the mean rank)
Companies are ranked within their BEA industry if it has at least 8 companies with a value, otherwise within their sector (at least 8), otherwise among all. Setting: 8.
figures = Σ wᵢ × rankᵢ (missing = 50)
weights: growth 1.2 · net margin 1.2 · EBIT margin 1 · equity ratio 1 · size 0.6 · productivity 0.6 ·
track record 0.4 · sites 0.5 · countries 0.5 · listings 0.3 · valuation 0.5 (divided by their sum)
chain = 50 + 45 × tanh(S_chain ÷ 2)
reports = 50 + 45 × tanh(S_reports ÷ 2)
DESK score (company) = 0.40 × figures + 0.20 × home country + 0.15 × chain + 0.25 × reports (missing part = 50)
A company needs at least two of the six core figures (growth, net margin, EBIT margin, equity ratio, size, productivity). The score carries the year of its most recent figure. Settings: the standard weights, the minimum of two.
Your own weights. Every account can replace all weights in this section (Settings → DESK Score); DESK then calculates every score, ranking and history with them, for that account only. The parts of the company and country score are scaled to 100 %; the key-figure weights are relative to each other. "Back to standard" restores the weights shown here.
weight = kind × breadth × rumour × severity × (1/3)^(days ÷ 45) × (1 − 0.5 × rank in list)
kind: filing, conflict, sanction 1 · news, commodity 0.8 · business cycle 0.7 · route, interest rate, macro 0.6 · company 0.5
breadth: min(1, log10(1 + sources) ÷ log10(11)) for news and company reports, otherwise 1
rumour: 0.35 if every headline is a rumour, otherwise 1
severity: for conflicts the rank of the annualised deaths among all countries with conflict deaths in the
last complete year (0 to 1), otherwise 1; conflicts are sorted by ln(1 + surprise) × severity
(rank in list: among all reports of the kind, not only those shown)
S_reports = Σ weight × direction (strain −1, growth/relief +1)
S_chain = Σ over strained commodities: weight × direction × (cents per dollar × share of the commodity) ÷ the largest
such value of any pair of BEA industry and commodity (measured: 4.25 cents)
Settings: all factors, the 45-day decay.
figure = share of all values of this indicator — all countries, the ten years up to the reference year
(today 2017–2026; in the history each year its own ten years; monthly series: all months) — that are
worse, 0 to 100 (ties count half); the yardstick moves only when new years arrive, and then for all
countries alike — a country's score does not move because another country's figures change
chapter = weighted mean of its figures (they must weigh at least 2 together; most figures weigh 1)
level = Σ chapter × weight ÷ Σ weight of the chapters present (only values at most 4 years older than the newest)
weights: economy 0.28 · security 0.15 · labour 0.14 · public finances 0.11 · financial system (incl.
money) 0.07 · trade 0.07 · resources 0.06 · conflict 0.05 · society 0.03 · technology 0.02 ·
politics 0.02
what counts (+ higher is better, − lower is better):
economy + GDP growth (quarterly yoy, else yearly) · + GDP per head (PPP) · − inflation (monthly,
else yearly) · + industrial production · + retail sales · + FDI inflows · + investment
· + human capital · + productivity level (USA = 1) · + industry share of GDP
· + leading indicator · + business and consumer confidence
labour − unemployment (monthly, else yearly) · − youth unemployment · + participation
security − homicide rate (double weight) · − prison population · − unsentenced detainees
public fin. − debt · + budget balance · − interest on debt · − seven tax rates (together as much
as debt and budget balance together)
financial − bad loans · − lending rate · − spread · − risk premium · + bank capital · + bank
liquidity · − IMF credit · + stock market (yoy) · + import cover of reserves
· − 10-year bond yield
trade + current account · + exports (% of GDP) · + logistics performance
resources + resource rents · + oil production per head · − energy imports (net) · − food imports (net, % of GDP)
· + electricity access · − electricity and gas prices for households and for industry
(Europe and the US; electricity also for Bahrain; elsewhere
they do not count)
Gulf regulated block tariffs, priced for the middle of the Eurostat bands: households
3,750 kWh a year incl. VAT, industry 1,250 MWh a year excl. VAT; price = bill ÷ amount;
the general tariff (not the subsidised one for citizens); the UAE and Saudi Arabia
are paused until their data portal permits reuse; without meter fees and fuel surcharges;
a tariff stays valid up to four years (statistics lag behind)
conflict − conflict deaths · − armed conflicts · − internally displaced
society + life expectancy · − Gini · + wealth per head · + median income · + health
coverage · + hospital beds · + water · + sanitation · − age dependency
technology + internet users · + R&D spending · + high-tech exports · + patents per head
politics + stability · + voice & accountability · + rule of law · + control of corruption
· + government effectiveness · + regulatory quality
not in the score: CO2 and the renewable shares, population growth, real interest rate, terms of
trade, and every figure that only measures size
a country gets a score only with figures in the economy, labour and public-finance chapters
(conflict figures count the last complete year; the score history uses, for every year, the
figures reported up to that year, with the same 4-year limit)
companies = 0.6 × revenue-weighted mean of company scores (without home) + 0.4 × (company revenue ÷ GDP) on the
same ten-year yardstick (its values for all countries in the ten years up to today)
(its weight grows with the number of companies: × min(1, n ÷ 10), none below 3 companies)
reports = 50 + (50 + 45 × tanh(S ÷ 2) − 50) × 0.5
S = Σ weight × direction of the country's reports + 0.15 × those of its companies
(a commodity under pressure does not count again for its strained suppliers: their own
conflict, violence or sanctions report already counts once)
DESK score (country) = 0.6875 × level + 0.1875 × companies + 0.125 × reports
(the ratio 55 : 15 : 10; the change over five years no longer counts — a decline
already shows in today's figures, and counting it again lifted countries coming
from far below)
(always measured against all countries; "only major economies" shows the same
scores and only renumbers the places)
(missing parts are left out and the rest scaled to 100 %; without reports the part counts as 50)
Settings: the weights of the parts and chapters, 0.15 in the reports; terms of trade are left out (an index with its own base per country); sums over countries (total GDP, reserves) are left out as a measure of size.
Price levels and working hours do not count in the score: purchasing power is already measured by GDP per capita and median income, both in purchasing-power parities (income divided by prices); long working weeks can mean low hourly pay, short ones part-time work. Both stay on the country pages.
articles a day = Σ articles ÷ files × 96
reported if the last 7 days are above every 7-day stretch before and
P(X ≥ k | λ = (b + 1) ÷ F_b × F_w) < 1 ÷ (countries × 3) (Poisson, Bonferroni)
order = (articles a day − usual) ÷ world articles a day × 100
headlines = GDELT themes of the article match the group and the headline itself names the country (or a place in it)
and a word of the group (attack, sanctions, protest …)
evidence = the articles of the counted events and the headline search, each only if its headline names a word of
the group and is not entertainment, sport or everyday life; a report without one such headline is not shown
stays = a report that met the rule stays for 24 hours while its week is above its own mean (it does not flicker
in and out at the edge of the rule); it only ever starts through the full rule
since = the time a report first appeared; "Most recent" is ordered by it. A report missing for more than
24 hours counts as new when it returns
most recent = the reports that pass the filters above the list (level, effect, countries, size, listed), and between
them the most-reported stories of the news stream from the last 24 hours (at least two sites) for the
chosen level -- a story already shown as evidence of a report is left out; with an effect chosen only
reports appear, because a news item has no effect. On top the five newest red, yellow or green reports
of the last 48 hours; below them everything from the last 24 hours by time -- older reports leave this
list but stay in the tiles below (countries, commodities, companies) while they hold
read = every GDELT file (one every 15 minutes), each run catches up the last hour. DESK checks every
minute whether GDELT has published a new file; once it is read, open pages show the new reports
without reloading ("Live"). A report's time is when GDELT published it -- GDELT stamps its files
with the end of their 15-minute window, about ten minutes ahead
a news item = the headline names a commodity, a market or crypto, or reports a supply event (below); or it
names a country AND the HEADLINE itself is about war and violence, sanctions and blockades,
unrest, state and foreign policy, or the economy and finance
left out = entertainment, sport, everyday life; for country headlines also police reports, courts,
municipalities and federal states, unless the headline is about war, sanctions or the economy
at large
one story = headlines of the same country within 24 h whose key words overlap (at least 60 %, or 40 % and
three words when a country is named); once three reports belong to it, its key words are the
ones at least a third of them share. The headline shown is the one whose words occur in most
of the reports; reactions, questions and explainers come last
most reported = stories with a report in the last 24 hours, ordered by the number of different sites that
carry them under a headline of their own (counted, not rated). The same headline on several
sites — agency copy, the stations of one network — counts once; otherwise a feature that a
hundred local stations share stands above a war. "Newest first" orders the same stories by
their latest report
Setting: the word lists that decide what a headline is about; 60 % / 40 % overlap; 24 hours.
Paused since 11 October 2026: the IMF has not yet answered our request for commercial use of PortWatch, so DESK currently shows no PortWatch figures anywhere. The rules below apply again once the permission is there.
week = ships in the last 7 days
normal = mean of the same week in all earlier years, ± 1 and 2 weeks
reported if the week is below (above) every one of the last 52 weeks, or below (above) every week of normal
change = week ÷ normal − 1
pace = deaths so far × 12 ÷ months covered
reported if pace > last year and P(X ≥ deaths so far | λ = (last year + 1) × months ÷ 12) < 1 ÷ countries
The current year comes from UCDP's monthly candidate releases. UCDP re-issues some events in later months with corrected numbers; each event counts with its latest release.
deaths = UCDP's best estimate; the low and high estimates are shown as a range, never as the toll
dots = only events whose place UCDP knows to within about 25 km (location precision 1–2), on a quarter-degree
grid. An event known only by district, province or country carries that area's centre as its coordinate;
it counts in the country and conflict figures but is not drawn. Under the map: deaths at the dots ÷
all deaths recorded for the year
first day = the earliest event of a conflict whose day UCDP knows exactly (date precision 1); an event dated only
to a week, month or year does not set a start day
entry = one event or a summary of several (event clarity 2); "events" counts entries
what UCDP = an event needs at least one death and an organised actor; a pair of opponents enters only once it
records reaches 25 deaths in a calendar year. The three kinds of violence (state-based, non-state, one-sided)
are mutually exclusive and are added, never counted twice
recession = two negative quarters in a row (GDP, quarter on quarter); contraction = one; recovery = growth after a negative quarter;
slowdown = two slower quarters in a row; otherwise expansion
rate turn = the first move of a central bank rate in the other direction within the last 6 months
sanctions = listed entries now against the count one month earlier (same counting method only)
strained = a country with a DESK Radar report of an escalating conflict (UCDP, above), new sanctions (entries
added in the last month), or violence in a country with a state armed conflict; news of blockades do not
count (boycott reports are not supply)
share = Σ world export shares of the strained countries
route = Σ over disrupted chokepoints: (1 − week ÷ normal) × share of the ship type that carries the commodity
order = (Σ world export share × severity + route) × 100 severity: for a conflict its rank (above), otherwise 1
The route term is an approximation: PortWatch counts the ships through a chokepoint by type, not the share of a commodity's world trade that passes it.
Only countries with at least 0.1 % of world exports are listed; the share counts all.
Since 9 October 2026 the public page shows no ship numbers: the transit data of IMF PortWatch come from Kpler, and their commercial use needs the IMF’s permission, which we have asked for. Until it arrives, the page shows the oil and LNG flowing through the strait (U.S. EIA, since 15 October 2026), the share of exports with no other sea route, who depends on the Gulf, oil prices and the start of the war; the ship formulas below apply once the IMF has agreed.
ships a day = mean of the last 28 reported days (IMF PortWatch); next to it the mean of the SAME calendar days of
each of the three years before. "Fewer than in any of those years" is stated only when it is
true of every one of them (no threshold)
change = last 28 days ÷ mean of those three years − 1
month by month = mean per calendar month since the first of those years; the current month with its days
start of war = the day of the first event UCDP records with an exact date for the conflict that began this year on the coasts of
the Gulf (the one with the most deaths). No such conflict: no line, and prices are compared
with a year earlier instead
share of world = exports of a good by Iraq, Kuwait, Bahrain and Qatar (their whole coast lies behind the strait)
exports ÷ world exports of that good, latest trade year (CEPII BACI); beside it the same for all seven
Gulf states. The four goods with the highest value from the four states
who depends = a country's imports of crude oil and of natural gas from the seven Gulf states ÷ its total imports
of that good; shown for the five largest economies (World Bank, GDP in US dollars)
prices = monthly average (World Bank Pink Sheet): the month before the start of the war, highest month since,
latest; change = latest ÷ month before − 1
oil and LNG = U.S. EIA estimates from tanker tracking (Vortexa), million barrels a day (oil: crude and condensate plus
petroleum products) and billion cubic feet a day (LNG), mean per quarter and per year, taken over
unchanged with the date of the EIA release. Quarters the EIA no longer lists stay from earlier releases
change = latest quarter ÷ the SAME quarter a year earlier − 1 (no season in between)
share of world = oil through the strait ÷ world total oil supply of the same quarter (EIA, Short-Term Energy Outlook)
oil supply
table = the two latest full years and the four latest quarters
The page states no cause: a change over time is not the effect of the strait alone.
report = a GDELT headline that names a country and one of four kinds of event:
facility a supply object (pipeline, refinery, terminal, port, strait, canal, oil or gas field, power
plant, grid, mine, smelter, dam) together with what happens to it
exports a country stops, bans, curbs or resumes the export of a commodity DESK carries
output a decision to cut or raise the output of such a commodity
ship an attack on, or seizure of, a tanker or cargo ship
effect = restart, resumption, reopening, restoration, lifted ban → easing; shutdown, halt, attack, damage,
fire, outage, disruption, seizure, export stop → under strain (a headline with both counts as the restart)
announced = a headline that only offers, plans, threatens or calls for it ("ready to reopen", "threatens export
ban", "could halt"); headlines that deny it, comment on it or look back on earlier years are left out
one event = same country, same effect, same kind, within 3 days of its FIRST report, and the headlines share
their key words (at least half of the shorter one, or three words; the country's name does not
count; "halts", "suspends", "blocks", "bans" count as one word). A later, similar headline is a new event
order = the time of the first report, newest on top — nothing is pinned
dependent = for a commodity: the importers of that commodity from the country, ranked by the share of their own
imports that comes from it (CEPII BACI, latest year); not shown for straits and canals
Kept for 14 days. The oil pattern for the News page and this list also knows oil pipelines, refineries, the Strait of Hormuz and fuel prices.
story = headlines of the same country within 24 h whose key words overlap by at least 60 % (the country's name
does not count), counted by the number of different sites
gold = the story is among the most-reported 2 % of all stories in the news stream (at least 2 sites)
AND a headline of it reports an event: a supply event (above), or a ceasefire, truce or peace agreement
in the supply panel additionally: the event has happened (not merely announced) AND it reaches across a border
(a commodity other countries import from there, or a shipping route)
how long = 72 hours after the first report; after that the line keeps its place in the order of time and
loses the colour
red = a major event (see below) -- gold never replaces it
In the evidence lists a gold headline stands first for those 72 hours; the supply panel is ordered by time only. Setting: 2 %, 72 hours.
company report = at least 2 sources or a matching SEC filing within 30 days of a headline; order = sources + 3 × filings
headline match = the headline names the company; if several companies share the first word, the second too;
a different legal form after the name is a different company
SEC filing = order = 10 × severe filings (3 × for companies without a DESK page) + filings + log10(revenue) ÷ 3
major event = more sources in 24 h than on any of the 30 days before, at least 0.5 % of all articles, at least 7 days to compare;
a war only if one side is the USA, Russia, China or a European state (EU, UK, Norway, Switzerland,
Iceland), or both sides are nuclear powers; attacker = the side named before the attack word more often
than after it in the headlines
Important Alert = a new major event
ongoing war = a UCDP conflict between states, still active, a war by UCDP's definition (at least 1,000 battle deaths in a year:
coded as war in the last complete year, or over 1,000 so far this year), at least one side from the G20 or the
nuclear powers; shown as a note for as long as it lasts, without an alert and outside the score
What a filing means is read from its text: an agreement (item 1.01) counts as expansion only if it is a business agreement (acquisition, purchase, joint venture, licence, supply, partnership), not a credit facility, bond or underwriting agreement, share buyback, share issue, preferred units or supplier finance, and not a rights plan, shareholder cooperation, settlement or employment agreement (whichever the text names first decides); the end of an agreement (1.02) that replaces a financing is not a burden, and a 1.02 that only refers to another item is judged by that item (a note redemption, an executive change or a rights plan is not a burden); an event accelerating a debt obligation (2.04) that is a voluntary redemption is not a burden; a delisting notice (3.01) that is a change of exchange or the end of listing after a merger is not a burden.
The 0.5 % is measured, not chosen: the share of all articles on four days with a known major event (24 Feb 2022, 11 Mar 2020, 22 Jun 2025, 10 Mar 2023) against ordinary days (scripts/radar-gross-eichen.js). Setting: 7 days to compare.
Everything refers to one event only: the countries where the fighting takes place (UCDP: deaths per country, among the parties) and its start year (the first year UCDP codes it as a war, otherwise the conflict start or the first GDELT report). The comparison year is the year before.
sea routes = straits on the coasts of those countries (geography); ships per day since the start year against the
average of the years before, the last 28 days against the same days of those years (IMF PortWatch);
"disrupted" = since the start lower than in EVERY year before (no threshold)
behind a strait = only countries whose whole coast lies behind a disrupted strait (Ukraine, Romania, Bulgaria, Georgia
behind the Bosporus; Finland, the Baltic states and Poland behind the Øresund; Iraq, Kuwait, Bahrain,
Qatar behind Hormuz; Sudan, Eritrea, Jordan behind Bab el-Mandeb) -- geography, not a choice
commodities = share of world exports from the countries at war in the pre-war year (CEPII BACI), plus the share of
the countries behind a disrupted strait; ordered by the sum
world price = World Bank Pink Sheet: highest month since the start and latest month, each / average of the pre-war year − 1
countries = imports from the countries at war (plus via disrupted straits) / all commodity imports of the country
energy (Europe) = imports by country of origin, natural gas / crude oil / hard coal (Eurostat nrg_ti_gas, nrg_ti_oil,
nrg_ti_sff); used because the trade table misses gas origin (Germany reports no gas by origin).
"Latest" for crude oil = the last 12 months (nrg_ti_oilm, by origin like the annual figures); for gas
and coal the latest year -- monthly gas figures (nrg_ti_gasm) count the border crossing, not the
origin (Hungary's Russian gas appears there as Serbian)
consumer prices = highest and latest value since the start / average of the pre-war year − 1 (EU Weekly Oil Bulletin,
Eurostat, EIA, HICP)
restrictions = OECD inventory: measures of the countries at war in force since the start year; one row per country,
measure and four-digit HS group, with the OECD title of the goods, the product named in the law,
the stated purpose and every period
companies = largest based there; foreign companies with known sites there (Wikidata, GLEIF): sites there / all
known sites; companies whose industry needs the most of the three most exposed commodities (BEA)
extra cost = cents of the raw-material industry per dollar of output of the company's industry (BEA) x share of the
commodity in that raw-material industry x world price change since the pre-war year; share of profit
= extra cost / the company's net margin
A change since the pre-war year is not the effect of the event alone. Taxes, exchange rates and other events are in it too; the data cannot separate them, and DESK does not estimate the share of the war.
Start day (events DESK detects itself, from GDELT headlines): the day the event turned red counts, not the year.
daily prices = not used at present: the EIA's daily spot prices come from a third party (Refinitiv/LSEG) and
need its permission, which has been requested
monthly prices = World Bank: the month before the start month against every month from the start month
shipping = PortWatch transits since the start day against the SAME calendar days of every earlier year with
complete data; "disrupted" = below every one of those years (no threshold). PortWatch reports with
a delay of a few days; until it reaches the start day, the card says so
consumer prices = the last value before the start day against every value since
where fighting = the parties named AFTER the war word in the headlines ("strikes on Iran") count as targets, the ones
is before it as attackers; the parties more often target than attacker are where the fighting is (a
sign, no threshold); if none, all parties
Wars that UCDP records keep the start year – unless the conflict began in that very year: then the day of its first recorded event with an exact date counts as the start day (since 1 October 2026; an event dated only to a month or year no longer sets it, since 2 October 2026). Order by impact on the world, measured, without own weights:
becomes scarce = value of the world exports of a good that come from the fighting parties (and from countries behind
a disrupted strait), US dollars, pre-war year (CEPII BACI) -- the list is ordered by it
more expensive = world exports of the good x its price rise at the peak: what the same volumes cost more -- ordered
by it
importers = the countries that imported the most of the good from there (value), with its share of their own
imports; rows that round to 0 % are left out
The companies with extra cost describe their industry, not their own purchases: the U.S. input-output tables book crude oil and natural gas together ("oil and gas extraction"), so a power utility can appear under crude oil although it mostly uses gas.
The page shows compact tiles (one sentence, the chain of effects, the key numbers, a delayed TradingView chart where a commodity or a share has one); "Details" opens the full card. The overview shows the first three tiles of each kind. Companies: quarterly results of the companies of the Companies map in the last 14 days (SEC 8-K, item 2.02). The sentences are quoted word for word from the press release filed with it (EX-99.1): those that name revenue, earnings, margins or the outlook with a number, at most five, without legal notices or table rows. The figures are the latest quarter filed in XBRL (10-Q/10-K, SEC calendar quarters) against the same quarter a year earlier; when the quarterly report for the new results is not filed yet, they are the quarter before, and the card says so. Ordered by revenue, no judgement of the results.
One card per trigger. Wars: the sheet above, condensed (highest price changes, largest shares of world exports, who imports those goods from there, consumer prices, companies with the largest extra cost). Price moves: the World Bank commodities with the largest change of the latest month against the same month a year earlier, both directions (an order, no threshold); their largest importers (CEPII BACI, latest year) and the companies with the largest extra cost at that change (same rule as above). Central banks (Fed, ECB, Bank of Japan) and rate turns (the central banks the Radar reports):
situation = the sign of the change over three months: balance sheet (FRED WALCL, ECBASSETSW, JPNASSETS) or
policy rate (Fed funds, ECB deposit rate; BoJ and all other countries: BIS policy rates) -- no threshold
turn = the first month in which the sign of the three-month change flips (the situation the Radar reports)
what followed = for every past month with the same situation: the change over the next 12 months of the share price
index and the long-term yield of that country (OECD, Financial market; ECB: the euro area, EA20), gold and crude oil (World
Bank Pink Sheet, since 1960); shown as the share of months in which it rose and the median change,
next to the same over ALL months of the period (the comparison)
minimum = at least 6 comparable months; fewer: not shown
vs. usual = share of months with a rise after months like now minus the same after all months, in percentage
points; only the sign is interpreted, no threshold
last times = runs of consecutive months in the same situation; for the three most recent
runs that began at least 12 months ago, the change over the 12 months after their first month.
Names such as "Financial crisis" are labels for the period, not an explanation
Currency swings: the three currencies with the largest change of the nominal effective exchange rate over three months (BIS, broad basket, monthly; only series that reach the latest month) -- an order, no threshold; situation = the sign of that change, targets = share prices, inflation and long-term yield of that country. Recession: the countries the Radar reports as two quarters in a row with falling GDP (quarterly national accounts); situation = every month of a past quarter that was the second (or later) falling quarter in a row. Targets: share prices (OECD; where the OECD has no series, the World Bank GEM stock index), long-term yield (OECD), industrial production and inflation (World Bank GEM). Industrial production and the GEM stock index are published as annual rates: their value 12 months later is itself the change over those 12 months, so it is used as it stands. GEM exports are left out because the series changes its unit within the history (Kuwait 1991, Saudi Arabia 2009 and 2015). Export restrictions: all measures of the OECD inventory in force today (no end date or an end date in the future), one row per country, HS heading and kind of measure; “new” = start in the inventory’s latest year. “Only major goods” keeps the rows whose HS heading matches a commodity DESK follows (Commodities tab) -- from every country, so the “Only major economies” switch does not apply here. The inventory covers industrial raw materials in the countries it surveys; agricultural goods are not in it. Earthquakes: the significant earthquakes of the last 30 days as selected by the USGS; company sites (DESK company map) and seaports (World Port Index) within 100 km of the epicentre; if none, the nearest known site and port with their distance (great-circle). The enlarged card shows the USGS products of the event unchanged: PAGER population and value exposed per intensity (Modified Mercalli), the probability of each range of deaths and economic losses, cities by shaking, the USGS assessment text and earlier earthquakes nearby; ShakeMap for the strongest shaking; “Did you feel it?” for the number of reports. “People at the strongest level” = the highest intensity with any people exposed. These are models of the USGS, not reported damage. Next date: the first future entry of the economic calendar (U.S. agencies, FOMC, EIA) that can change the card. My portfolio: positions and watchlist entries matched to companies via ISIN and ticker; a card concerns them if one of these companies is listed on it or has its headquarters in the card’s country (Fed: United States, ECB: euro area, BoJ: Japan).
The windows overlap (one per month), so the cases are not independent; the past does not repeat reliably. The page says what happened, not what will happen, and is not investment advice.
10-year yield = OECD via FRED; where that series is missing or older, the OECD series directly (same author)
stock market yoy = World Bank GEM; where it is missing or older, the change of the OECD share price index against the
same month a year earlier
share prices = OECD share price index (2015 = 100), monthly since 1960; 3-month rate = OECD interbank rate
energy by origin = Eurostat nrg_ti: the largest origins in the latest year and five years earlier (crude oil also the
last 12 months); European countries only
company: events = sites in the countries at war / all known sites; extra cost = the industry's direct purchases of the
commodity per dollar of output (BEA detail table) x share of the commodity x price change (war peak, or
latest month vs a year earlier); share of profit = extra cost / net margin
company: crises = home market: the share price index of the headquarters country, lowest month of the crisis / average
of the year before - 1 -- the market, not the company's own share
From the company's Wikidata industry and kind entries (the industry named most often). Without a recognised entry, the official meaning of its BEA industry code decides (for example 3361MV motor vehicles → automotive, 521CI → banks); if a code covers several industries (325 chemicals and pharmaceuticals, 334 hardware and semiconductors), the company's own name and product entries decide, otherwise the industry the BEA names the code after. Companies without either stand as "Not classified". A "real estate investment trust" is real estate, not asset management.
Sector › industry on the companies map. There every company has exactly one path: its sector and, under it, its industry. An industry that normally belongs to another sector stands under the company's own sector (Walmart: consumer staples › retail; Daimler Truck: industrials › automotive), so the industries of a sector add up to the sector. An industry appears in the tree (menu, matrix, "Sectors by importance") from three companies; the rest of the sector stands as "other companies of the sector". Industry dot in the matrix = the same averages as a sector dot, over the industry's companies with a DESK score (at least three); the lines stay at the average of all companies. Competitors of a company = the companies with the same path.
cents per dollar = cents of a raw-material industry needed per dollar of output, directly and through all earlier
stages (BEA Total Requirements); a single commodity: × its share of the BEA commodity line
moat = in every comparable year (at least 3, the latest at most 3 years ago) in the top quarter of the net
margins of its BEA industry (a year is comparable with at least 4 companies)
key position = named producer of a commodity whose three largest exporters hold more than the median of all commodities
world share = exports (or production) of a country ÷ world total
sum of world shares = Σ over the commodities of a group of the world shares of one country; a commodity only one
country reports (USGS lists e.g. crushed stone, sand and gravel, noble gases only for the United States)
has no world share: it is left out here and not offered on the world map, but stays on the production tiles
historical = a FAO series whose period in the name ends before the year before last (e.g. particle board and OSB,
1961-1994) is marked "(historical)"; the FAO no longer publishes it
colour scale = the rank of the value, lowest to highest
concentration = exports (or production) of the three largest countries ÷ world total
(shares and concentration: all countries of the latest year of the source, the same year for all)
strained share = Σ world-export shares of the strained countries — the same countries and the same number as in
DESK Radar ("Commodity under pressure")
raw-material filter (company map) = companies whose BEA industry needs more of the raw material per dollar of output
than the average of all BEA industries (the raw-material industry itself excluded)
dependence, industries = cents of the resource per dollar of the industry's output, directly and through every
earlier stage (BEA Total Requirements); the raw-material industry itself excluded
dependence, countries = Σ over the goods of the resource of max(0, imports − exports) (CEPII BACI, latest year of
each good) ÷ all merchandise imports of the country in the same year (World Bank) × 100;
a net exporter has 0; countries whose net imports would exceed all their imports are left out (the sources contradict each other there);
the history behind each figure: the same sum for every year, all goods in that year
dependence, companies = dependence of its industry (¢ per $) × (1 + q) / 2, q = (imports − exports) ÷ (imports +
exports) of the resource in its home country: 0 = the country only exports it, 1 = it only
imports it; no revenue; producers of the resource, companies without an industry code or
without trade data for their home country are left out
dependence over the years = the same per year: ¢ per $ of the industry in year J (BEA) × (1 + q of year J) / 2
BEA industries of a sector = revenue of the sector's companies in that BEA industry ÷ revenue of all its companies with
a BEA code (latest revenue of each company); over the years: both revenues chained from year to
year over the companies reporting both years (at least two, at least a quarter of the revenue),
anchored at today's revenue in the same year — the last point is the value in the table
world production, all resources = median over the resources of (world production in the year ÷ world production
in 2015 × 100); a resource counts in a year if at least 80 % as many countries report it as in its
strongest year; a year counts if at least 80 % as many resources have a value as in the strongest
year. The world index of the production tab (mining, agriculture, forestry) is the same index.
Setting: base year 2015.
current = at least one statement without an end date (P582); a second term next to an ended one counts
several = of the current heads of state or government, the one with the latest start (all, if they started together)
military site = still a military site by at least one current type statement, not dissolved, closed or out of service
quantity = the value as entered when it is already in the target unit (km, m³/s, km², m); otherwise converted
country list = rankings and percentiles count countries only (the World Bank list plus Taiwan) — regional aggregates of a source are left out
codes = one code per country as at the World Bank; where a source uses its own (Kosovo: XKS, KOS, KSV, UVK, OWID_KOS;
Palestine: WBG; euro area: XM, EA20, EA, EURO; EU: EU27_2020, EU, EUQ), the value is filed under the DESK code,
never over a value the source reports under the DESK code itself; euro area and EU have their own pages but
never count in rankings or world sums
regions = each administrative unit takes its Wikidata ID from Natural Earth; the same unit listed twice (same border)
appears once; if units with different borders carry the same ID, the ID is ambiguous and none of them shows
Wikidata figures (population etc.) -- no figures rather than another unit's
This applies to office holders and their party, company leadership, operators and countries of military sites, and subsidiaries. A statement that was ended once does not remove a person or an operator who holds the role again.
price level (US = 100) = PPP conversion factor ÷ market exchange rate × 100 (World Bank, same year)
IMF credit (% of GDP) = IMF credit outstanding ÷ GDP in current US$ × 100 (IDS and World Bank, same year)
oil production per 1,000 = oil production ÷ population × 1,000 (same year)
patents per million = patent applications ÷ population × 1,000,000 (same year)
foreign-born share (refugee flow map) = international migrant stock ÷ population × 100 (UN DESA via World Bank,
same year; several countries: both sums; the latest year up to the one shown)
net food imports (% of GDP) = (food share of merchandise imports × merchandise imports − food share of merchandise
exports × merchandise exports) ÷ GDP × 100 (World Bank, same year; negative = net exporter)
index = Σ weight × percentile of each part ÷ Σ weight (0 to 10; every part required)
percentile of a part = position among all countries with a value, lowest 0 to highest 1 (ties share the middle);
turned round where lower is better
crime index = homicide rate 0.6 · political stability and absence of violence (lower is worse) 0.4
self-sufficiency = energy imports (% of energy use) 0.25 · net food imports (% of GDP) 0.2 · imports (% of GDP)
0.25 · current account (% of GDP) 0.2 · IMF credit (% of GDP) 0.1 (no IMF credit = 0)
countries = the countries of the country list only; regional aggregates of a source do not count
Settings: the parts and weights of both indices.
year = the latest year (or month) in which at least 90 % as many countries reported as in the 5 years
(12 months) before; every country counts in the same year
value = sum over the countries for quantities that add up (output, exports, seizures)
= median country for index values
= unweighted mean (each country once) for everything else
countries = the World Bank country list without its aggregates, plus Taiwan (not in the World Bank list;
its figures come from the other sources, e.g. the IMF or USGS)
world share of a country = its value ÷ the world value of the same year; the year is the country's latest
year in which at least 90 % as many countries reported as in the 5 years before, and it is shown
next to the share
The World chart, the World country page and the panel under the world map all use this rule. Where the World Bank or another source publishes its own World row (ICP price levels, UN median age, Maddison, hospital beds, and ratios of World Bank world rows), the World page and the World chart show that row and the rule's value beneath it. The panel under the map follows the "only major economies" box; the World page always uses every country.
With two to four indicators combined, each country gets its percentile rank per indicator (0 = lowest, 100 = highest among the countries that have all of them), and the colour is the mean of these ranks. The direction of an indicator does not count here. The time line runs over the years that all chosen indicators have in common; in a chosen year the ranks are formed from the values of that year only.
gold value = total reserves including gold (World Bank, FI.RES.TOTL.CD, gold at London market price)
- total reserves excluding gold (World Bank, FI.RES.XGLD.CD), year-end, in US dollars
gold price = World Bank gold price (Pink Sheet), December of the same year, US dollars per fine troy ounce
tonnes = gold value / gold price / 32,150.7466 (fine troy ounces per tonne), one decimal
An estimate: the World Bank values the gold at the market price on the reporting date, DESK divides by the December average, so the volume can differ by a few per cent from what the central bank reports (around 1-2 % above the IMF tonnes for most countries). Until the IMF permits commercial reuse of its data, DESK does not show the reported volumes (IMF, International Reserves and Foreign Currency Liquidity). These are official holdings, not gold in the ground (that is the USGS reserves on the resources map).
general government = central, state and local government and social security funds (sector S13), % of GDP
EU, EFTA, candidates = Eurostat: debt as defined for the excessive deficit procedure (gov_10dd_edpt1, GD), net
lending/borrowing (B9); total revenue (gov_10a_main, TR), interest payable (D41PAY)
other OECD members = OECD National Accounts at a Glance, chapter 6: gross debt (LES13_FD4), net lending/borrowing
(B9S13), total revenue (OTRS13)
all other countries = World Bank Fiscal Space Database (Kose, Kurlat, Ohnsorge, Sugawara): general government gross
debt (ggdy) and fiscal balance (fby), % of GDP; debt and balance only (CC BY 4.0)
interest, OECD = interest paid (D41) / total revenue (OTR), both in national currency (OECD Table 12)
x total revenue in % of GDP (OTRS13), two decimals
both available = Eurostat (the official EU notification)
central government = World Bank (GC.DOD.TOTL.GD.ZS, GC.NLD.TOTL.GD.ZS, GC.REV.XGRT.GD.ZS, GC.XPN.INTP.RV.ZS),
only for countries without a general government figure
General government and central government are separate indicators with their own names and are never ranked together: central government leaves out states, municipalities and social security, so the two figures differ by definition. OECD partner countries outside the 38 members are not taken from the OECD: their coverage there is incomplete and not comparable. For them, and for every other country without a Eurostat or OECD figure, debt and budget balance come from the World Bank Fiscal Space Database; revenue and interest stay central government.
money-flow map = OECD FDI positions by counterpart area (BMD4): directional principle (inward DI, outward DO),
immediate counterpart, all resident units incl. special purpose entities, million USD;
reporters = the OECD members, the 15 largest partners per direction
Bulgaria, Croatia, Cyprus, Malta, Romania: Eurostat bop_fdi6_geo (stocks, directional principle),
million euros × ECB end-of-year USD per EUR; Malta keeps its partner figures confidential
since 2017, so it reports up to 2016
monthly trade = EU-27: Eurostat Comext DS-045409, product TOTAL, value in euros × ECB monthly average USD per EUR
USA: US Census Bureau, exports (total value) and general imports by country
United Kingdom: HMRC Overseas Trade Statistics (EU and non-EU), pounds
Canada: Statistics Canada table 12-10-0176, all industries, Canadian dollars
Australia: Australian Bureau of Statistics MERCH_EXP / MERCH_IMP, Australian dollars
Norway: Statistics Norway table 08803, kroner
other currencies = value × (ECB USD per EUR ÷ ECB currency per EUR), monthly averages
world total = sum over all partner countries; an EU month counts only once its intra-EU partners are reported
These replace the IMF's Direct Investment Positions and International Trade in Goods while the IMF has not granted commercial use. Countries that report neither appear only as partners.
source per country = EU member: European Commission (AMECO, spring/autumn forecast)
otherwise: OECD Economic Outlook, then AMECO, then (growth only) World Bank Global Economic Prospects
one source per country and indicator, never mixed within a series
government debt = AMECO only, for EU members (debt as defined for the excessive deficit procedure, the same basis
as the Eurostat history); the OECD Outlook uses a different debt measure (all liabilities)
budget balance = AMECO (EU) or OECD net lending/borrowing of general government (NLGQ), as the history
growth (AMECO) = real GDP (OVGD) in year t ÷ real GDP in year t−1 − 1, in %
inflation (AMECO) = harmonised consumer price index (ZCPIH, else national ZCPIN) in t ÷ t−1 − 1, in %
long term = after the last year of that source: OECD Long-Term Baseline (scenario BAU1, business as usual),
up to 2060 -- growth (real GDP GDPV in t ÷ t−1 − 1) and budget balance (NLGQ) only; it has no
consumer prices, unemployment or comparable debt, so those end with the short-term source
forecast years = every year after the last actual year of DESK's own series for that country
no own series = no forecast is shown
The figures are projections by the European Commission, the OECD or the World Bank, not by DESK; each year names its source. They replace the IMF World Economic Outlook while the IMF has not granted commercial use.
agglomeration of a city = the latest estimate of the United Nations World Urbanization Prospects (population of the
built-up city area at mid-year), where the UN lists the city; otherwise Natural Earth
history behind the number = the UN estimates from 1975 to the year of the revision (projections after it are left out)
matching = same country, and the same name (either name of the UN entry, or one containing the other)
within 60 km, or at most 8 km to the UN city's population-weighted centre; each UN city goes
to one place only: the largest one with a matching name
sum of several places = sum of the listed places that the UN series covers, per year
The list per country shows the city proper where Natural Earth has it; its line is still the agglomeration, and the chart says so.
domestic credit (US$) = net domestic credit in local currency (World Bank FM.AST.DOMS.CN)
÷ official exchange rate of the same year (World Bank PA.NUS.FCRF, local currency per US$,
annual average); a year without an exchange rate is left out; both unrounded, only the
result is stored to six significant digits
renewable electricity = Ember "share of generation"; a value below 0 % or above 100 % is a source error and is left out
(Costa Rica 2025: 109.9 %)
refresh = each week the last five years and 24 months per country; the whole series every 90 days
new edition = each run also reads the IMF world totals (all years and months, one small request per country); if a year
before the latest two or a month before the latest 24 differs from DESK's copy (by more than 1 in a million),
the IMF has published a new edition, and the whole series of that country is fetched at once
reserves of a country = USGS Mineral Commodity Summaries, "Reserves" of the latest edition (tonnes, carats ...);
where a mineral has several reserve figures, the one with the most countries;
several lines of one country in the same year are added up
share of world reserves = reserves of the country ÷ the USGS "World total" (else the sum of the countries)
price = World Bank Pink Sheet (latest month vs. the same month a year before);
where it has no series: the first USGS price series of the mineral, annual,
with the USGS wording and unit, never chained across different definitions
exports under restriction = exports of the countries with at least one measure in force in the latest
year of the OECD inventory for the HS codes of the resource ÷ world exports (BACI)
change over 5 years = (sum in year L ÷ sum in year L−5 − 1) × 100, over the countries that report
in BOTH years; L = latest year of the series. Minerals with a USGS world total:
(world total L ÷ world total L−5 − 1) × 100, as on the world chart
world production = sum of the countries (World Bank countries and Taiwan, no territories or former
states); minerals: the USGS world total, as on the world chart
top 3 share = three largest producing countries ÷ world production, year L
reserves ÷ production = world reserves ÷ world production in year L (years) — a ratio, not a forecast
reserves over the years = the "Reserves" column of each earlier USGS edition; edition J states reserves
at the end of year J − 1; only editions in the same unit as today's figure
A resource the OECD inventory never records (it covers 65 industrial raw materials, not energy or farm goods) reads "not in the OECD inventory", not "no restriction". In the trend table a series that ends before the latest year of its own source is shown grey at the end, with its last year.
Every figure of the trend table opens its history, computed per year with the same rules as the table:
world production / exports = sum of the countries in that year (World Bank countries and Taiwan); minerals:
the USGS world total, as on the world chart
5-year change = production (exports) of year Y ÷ that of year Y-5 - 1, over the countries that report
both years only; minerals with a USGS world total: from the world total
top-3 share = the three largest producers of the year ÷ the sum of the year
imports = sum of reported imports of all countries in that year (CEPII BACI, importers' view)
imports, 5 years = imports of year Y ÷ imports of year Y-5 - 1, over the countries that report both years
reserves = latest reserves estimate (USGS), world total or the country's own
companies = companies that depend on the resource, counted as in the Tree: producers, named users,
the same industry and the companies of the five industries with the highest input share
industries = the end uses the USGS summary lists for the mineral; else the industries that buy exactly
this good in the detailed BEA input-output table; else the industries of its producers and
named users
exports under restrictions = exports of the countries with a restriction in the inventory year ÷ world exports
country row = the same for one country: its production and exports, their 5-year change, its share
of world production, its reserves ÷ its production, and its measures in force
end uses = USGS Mineral Commodity Summaries, section "Domestic Production and Use", in its words;
shares in % where the report states them (world or United States, as the sentence says),
otherwise only the order; a list is kept only if its shares add up to 60–105 %
named users = companies whose own Wikidata entries (industry, products, description) name a product made
from the resource ("lithium-ion battery", "lithium grease") or one of its USGS end uses that
means a product of its own (batteries, superalloys, solar cells, cemented carbides, tinplate,
solder, galvanizing, brake linings, flame retardants, ammunition, jewelry, catalytic
converters, drilling fluids) — never collective items such as "building construction"
unit value = USGS Data Series 140: value of US production or consumption ÷ quantity, per year since 1900,
nominal and in 1998 dollars — not an exchange price
The 60–105 % check only tests whether the list was read completely from the two-column PDF; it says nothing about the resource. End uses without a German entry in DESK's table stay in the USGS wording.
mineral production of a country = USGS Mineral Commodity Summaries where they list the country and year;
otherwise the USGS country report (Minerals Yearbook vol. III, table 1);
otherwise World Mining Data (Federal Ministry of Finance, Vienna, table 6.4)
rule check = for every country and year in BOTH: country report ÷ Summaries;
a rule is used only if the median lies within 0.9–1.1 and at least 70 %
of the pairs within 0.8–1.25 (41 of 55 rules pass; 5 pairs at least)
World Mining Data check = the same test per mineral: World Mining Data x unit factor ÷ Summaries over all
country-years both report; unit factor 1, 1/1000 (kg) or 1000, and oxide to
element (Li2O, Nb2O5, Ta2O5) by molar mass; World Mining Data reports metal
content where the USGS often reports concentrate, so a mineral fills gaps only
if it passes (27 of 52; e.g. chromium Cr2O3 and phosphate P2O5 do not)
world total = Summaries only — filled years never enter a world sum or "All minerals"
world line of a mineral = the USGS "World total (rounded)" of every edition, same kind of production as the
countries, later edition wins; used where it reaches the latest country year and
the listed countries carry at least half of it (median over the years); a year in
which the listed countries sum to more than 105 % of it is left out; otherwise the
sum of the countries
alcohol chain = FAO production (beer, wine, hops) + FAO Supply Utilization Accounts
(malt, must, spirits, cider, vermouth, other beers; since 2010), tonnes
occurrence (Trade Chain map A) = minerals: USGS reserves; oil, gas, coal: proved reserves (CIA World Factbook);
crops: area harvested (FAO); meat, milk, eggs, wool, hides: stock of the animal (FAO);
wood: forest area (FAO land use) — latest year per country, share = country ÷ sum of countries
All energy production = Σ over oil, natural gas, coal of (country ÷ sum of all countries), latest year each (EIA)
The two limits are measured against the Summaries themselves, not assumed: the rules that pass have a median of 1.000, the ones left out either measure something else (titanium, zirconium: 1.2 to 1.7) or match in fewer than 70 % of the pairs (salt, graphite, peat). A column in the Overview table that holds no value for any country is not shown.
Every figure DESK shows is checked against its source. Each night, independent readers fetch the sources a second time through another interface and compare every stored value; a cross-check compares every place where the same number appears; and every tile of the site has its own check that recalculates its figure from the raw data. Tiles with a member’s own data are checked on invented test data, without reading anyone’s entries. Findings are reported the next morning and fixed.
Rules added by these checks. A Wikidata amount and a filed amount count as equal when they differ by less than one dollar (the filed one carries decimals); only then does the direction of a fiscal-year shift follow from matching values. Equity above total assets, a negative balance-sheet total without a filing, a year before 1800 or after next year, an ISIN or LEI whose check digit fails, and a company listed as its own owner are not shown. A negative liabilities line in a filing is replaced by total assets minus equity. The market-cap rank of a coin is its place by market value. On a food, protein, carbohydrates, fat and salt together cannot exceed 100 g per 100 g after the tolerances the EU allows for nutrition labels (Commission guidance, December 2012: below 10 g ±2 g, 10–40 g ±20 %, above 40 g ±8 g; salt below 1.25 g ±0.375 g, otherwise ±20 %); if they do, the values of a pack or portion stand in the per-100-g fields and none of them is used. A part larger than its whole beyond the same tolerances (sugars above carbohydrates, saturated fat above fat) removes both. Long chart series are thinned to real daily values at their real dates, never to averages.
A filed amount repeated to the dollar in two consecutive years (for example equity of 5,778,500,000 $ for both 2011 and 2012) is a duplication in the SEC frames: the earlier year is dropped; round amounts (whole millions) can repeat and stay.
How fresh the data are. Every source is rebuilt on its own schedule: daily for policy rates, bond yields, exchange rates, the business cycle, sanctions and shipping; every few days for consumer prices and commodity prices; weekly or monthly for the rest. Where a source publishes new editions or table versions (OECD data flows, the UNODC World Drug Report, ICP rounds of the World Bank), the build looks up the newest one itself instead of a fixed number. Each ICP round is taken in the World Bank’s latest revision (the 2011 round as recomputed with ICP 2017); a country missing from that revision keeps the earlier one. Drug seizures that a country reports through two channels with exactly the same quantity count once. When a source revises figures it had already published (the World Bank’s Governance Indicators, for example, rework every past year with each new edition), DESK notices that a freshly fetched value differs from the stored history and reloads that history in full, together with the other indicators of the same edition; indices built from such figures are recomputed as soon as one of their parts is newer than they are. Where the same figure is available sooner elsewhere, DESK takes the sooner one: forest, agricultural and arable land come directly from FAOSTAT (Land Use) rather than via the World Bank, and effective exchange rates from the BIS (monthly) for the 64 economies it covers; for the other countries they remain those of the World Bank’s Global Economic Monitor, which ends in 2024. Likewise the renewable share of electricity comes from Ember, the Corruption Perceptions Index from Transparency International and real house prices from the BIS (rebased to 2015 = 100 like the OECD series they replace); a country keeps one source for its whole series. Series that originate with SIPRI, IISS, WIPO, UNODC or ICAO are taken via the World Bank, which licenses them for reuse (CC BY 4.0); the originals are more recent but not licensed for a paid service. Each figure shows its year or month.
CME & Coinbase futures (CFTC Commitments of Traders, Traders in Financial Futures, futures only)
amount in coins = contracts × contract size (CME Bitcoin 5 BTC, Micro 0.1, Coinbase Nano 0.01;
CME Ether 50 ETH, Micro 0.1, Coinbase Nano 0.1), summed over the markets of one coin
net position = long − short of a trader group, without spread positions
change = this week − previous week (status Tuesday, published Friday)
Crypto on company balance sheets (SEC XBRL)
holding = CryptoAssetFairValue; if only the parts are reported, Current + Noncurrent
(market value of all crypto holdings on the reporting date)
share of assets = holding ÷ Assets on the same reporting date
group = ETFs and trusts: SEC industry code 6221, "ETF" in the name or a crypto trust;
treasury companies: share of assets ≥ 50 %; all others
quarter = sum of the last reporting date of each company in the calendar quarter, shown once
90 days have passed after the quarter end
Bitcoin and liquidity (FRED)
net liquidity = Fed balance sheet (WALCL) − Treasury General Account (WTREGEN) − overnight reverse repo
(RRPONTSYD, the value of that Wednesday or the last before it), every Wednesday
M2 in a week = the latest published month (M2SL)
Bitcoin in a week = daily close of that Wednesday (mempool.space)
Mining (mempool.space)
pool share = blocks of the pool ÷ all blocks in the period
top 3 = share of the three largest pools
revenue per day = Σ (reward per block in satoshi ÷ 10⁸ × blocks in the section × price) ÷ days covered
hashprice = revenue per day ÷ (hashrate in H/s ÷ 10¹⁵) → dollars per PH/s per day
block fullness = block weight ÷ 4,000,000 (the limit of BIP 141)
Settings: 50 % for treasury companies; 90 days until a quarter is shown.
total return = price gain + realised gain + dividends − transaction costs − taxes
price gain = value today − cost basis + buying fees still in the cost basis (positions with a price only)
realised gain = sale proceeds − cost basis of the shares sold, both before fees
dividends = paid out before withholding tax
(every fee and every tax is deducted exactly once, in its own row)
cost basis = price × quantity + fee, moving average; a sale takes its share along
time-weighted return = Π over the days (value of the day ÷ (value of the day before + money flow of the day)) − 1;
if the portfolio starts on the first day of the series, that day counts from the purchase price
money flow of a day = purchases (price × quantity + fee) − sale proceeds (price × quantity − fee); without a sale
price the cost basis of the shares sold
internal rate of return = the rate at which all payments and today's value net to zero; payments of open
positions without a price are left out, because their value is not in today's value either
return in % = total return ÷ cost basis of the positions with a price
dividend growth (CAGR) = (dividends of the last full year ÷ those of the first full year)^(1 ÷ years) − 1;
only years in which the portfolio existed on 1 January
answer = Claude (Anthropic) with DESK's own lookup tools, no internet access; every figure comes from
the same routes as the pages (rankings, country pages, reserves, DESK Impact, DESK Score)
credits used = input tokens × input weight + output tokens × output weight
+ tokens written to the cache × cache-write weight + tokens read from the cache × cache-read weight
(the weights follow Anthropic's published token prices; a question can take several rounds, each is counted)
shared block = DESK AI's tools and instructions are the same for everyone and kept in the cache for an hour;
when they have to be written anew, your credits count them at the cache-read weight, the rest
is borne by DESK -- so a question costs the same whether or not someone asked just before
monthly credits = 1,500 per account, shared by DESK AI, the portfolio analyst and the Radar explanations;
counted from the 1st of the month, 00:00 Berlin time
ranking = countries ranked by their latest value from the comparison year on; comparison year = the latest
year in which at least 80 % as many countries report as in the best-reported year; older values
are listed separately with their year and not ranked
net margin = net income / revenue of exactly the periods shown
The free calculators at deskalmanac.com/tools each show their formula on the page. In short:
compound interest = P · (1 + i)^n + C · ((1 + i)^n − 1) / i; i = annual rate / 12, n = months, contribution at month end
withdrawal = month by month: withdrawal at the start of the month, then the rest grows by return / 12;
constant withdrawal for n months = K · i / ((1 − (1 + i)^−n) · (1 + i))
real return = (1 + return) / (1 + inflation) − 1
position size = account × risk % / (|entry − stop| × value per 1.0 price move), rounded down to whole units
loan (by term) = K · i / (1 − (1 + i)^−n); by initial repayment: K · (rate + repayment) / 12
purchasing power = amount × index(today) / index(back then); back then = annual average, today = latest month
(US: BLS CPI-U; Europe: Eurostat HICP, 2025 = 100)
currency = amount / rate(A) × rate(B), ECB reference rates (1 EUR = x), for information only
central bank rates = Fed, ECB and Bank of England directly from the central bank; SNB and Bank of Japan as the BIS
monthly value (both require permission for commercial use of their own data)
market hours = local time of each exchange from the browser's time zone database, regular sessions only
(no holidays, shortened days or pre-/after-market)
Median age. DESK shows the median age as the UN publishes it (World Population Prospects, MedianAgePop, 1 July). The population pyramid next to it comes from the World Bank, which starts from national population counts and can differ from the UN by more than a million people (Germany 2023: 83.3 against 84.5 million); the median of the pyramid shown can therefore differ from the UN figure by up to about a year. Only where the UN gives no figure is the median read from the pyramid’s five-year bands.
Mine production before 2017 (USGS). From 2017 on, production figures come from the USGS data release. For earlier years DESK reads the annual Mineral Commodity Summaries. Each edition gives two years: the final figure for the year before last and a preliminary estimate of the agency for the last year. The final figure is taken wherever it could be read from the following edition; where that edition does not carry the row, the year shows the USGS’s own preliminary estimate. This affects about 1,200 country values up to 2016. DESK does not replace them with its own numbers.