Skip to content
Aondeir

Methodology

How we calculate the personal ranking, handle missing data and use exchange rates. The calculation is deterministic and uses no AI.

Versions in use: algorithm rank-2026.09.3 · indicators ind-2026.09.3 · FX fx-2026.09.1. Each saved ranking stores the versions used.

Steps

  1. Must-have conditions are checked first: met, not met or unknown. A city that does not qualify leaves the ranking.
  2. Each active criterion looks up the city's latest local data, in the source's classification.
  3. A common set is formed: criteria with data for every candidate city and in the same comparability group.
  4. Each criterion scores 0 to 100 with fixed parameters; the final score is the weighted average over the common set.
  5. Coverage = common-set weight ÷ active weight. Below 60%, the result is provisional with no position.

score = sum(weight × criterion score) ÷ sum of common-set weights

Normalisation (0–100)

Parameters do not depend on the cities being compared: adding or removing a city does not change the others' criterion scores.

CriterionIndicatorRule
Safetytaxa de homicídios por 100 mil hab.Linear: 0 or less = 100; 40 or more = 0.
Jobstaxa de desocupação (%)Linear: 3 or less = 100; 15 or more = 0.
Climatetemperatura média anual (normal climatológica, °C)100 inside the preferred range (hot 22–28 °C, mild 15–22 °C, cold 5–15 °C); loses 10 points per °C outside.
Health serviceshospitais por 100 mil hab.Linear: 5 or more = 100; 0 = 0.

Budget, schools, hospital proximity, accessibility, mobility and sports have no comparable source yet: they appear as “no data”.

Safety

  • We use the municipality or metropolitan area; district/neighbourhood only when data are robust.
  • Each source keeps its classification (e.g. homicídio doloso in SINESP, homicide in StatCan). Typologies from different countries are never in the same common set.
  • National figures appear only as labelled context and never replace local data.
  • Counts below 20 incidents are flagged as unstable.
  • Recorded incidents and perceived safety are separate indicators; phase 1 uses records only.
  • We store count, denominator (population), period and area for each rate.

Missing data

  • Missing data never becomes zero.
  • Weights are never silently redistributed: criteria outside the common set are listed with the reason.
  • Incomplete results are provisional or have no position.
  • Proximity does not prove quality, availability, right of access or accessibility.

Exchange rates

  • European Central Bank reference rates (EUR base); cross rates are calculated by us.
  • No publication on weekends and holidays: we use the latest available rate and show its date.
  • Rates older than 7 days are flagged; older than 30 days are not used.
  • Original amounts are preserved; half-even rounding only to the currency's decimal places.
  • A reference rate is not a guaranteed exchange or transfer price.

Visas

  • Government sources only, with verification date and links.
  • The checklist compares what you declared: met, not met or unknown.
  • Information with expired verification is not shown as a current rule.
  • We never guarantee eligibility or estimate approval likelihood.

Advertising

Ads (when enabled) appear only on public pages and never change scores. Profiles and preferences are not sent to Google or advertisers.