Why a personalised relocation atlas?
Choosing where to live is one of the higher-stakes decisions a person makes, yet the public tools available to support it are unusually weak. This page sets out the gap in the existing literature and tooling that motivated the project, before the following tab documents the visual and interaction choices that follow from it.
Relocation is a multi-criteria problem treated as a single-number one.
A relocation decision is, in formal terms, a multi-criteria decision-analysis (MCDA) problem: a set of alternatives (cities) evaluated against several, often conflicting criteria (income, cost, climate, safety, social fit). The dominant public artefacts — Mercer’s Quality of Living index, the Economist’s Liveability ranking, Numbeo’s indices, “best places to live” listicles — collapse this structure into a single ordinal ranking. A scalar output is convenient for headlines but discards the information the decision actually depends on: which criteria the reader weighs, and how heavily.
City-level averages describe a citizen who does not exist.
Headline figures — mean salary, median rent, “cost of living” — aggregate across occupations, household types and neighbourhoods whose distributions are heavily skewed. The mean of a long-tailed distribution is a poor estimator of any individual outcome. For a prospective mover, the relevant quantities are conditional: salary given their occupation and experience, rent given their household size, purchasing power given both. Public tools almost never report conditional figures, so the user is left to do the conditioning informally and badly.
The criteria that decide the move are rarely the ones that get measured.
Migration research (Florida, 2002; Glaeser, 2011) and survey work on subjective well-being consistently identify climate, walkability, daylight, social tolerance, language friction and perceived safety as strong predictors of post-move satisfaction. These factors are harder to quantify than rent or GDP per capita and are therefore systematically under-weighted in indices that optimise for data availability. The result is a measurement bias: tools become more confident about the variables that matter least.
The relevant data exists, but not in one place.
Occupational salary distributions, rental indices, climate normals, crime statistics, broadband penetration, walkability scores and visa frameworks are all publicly available, but they are maintained by different institutions in different schemas. Joining them at the city level is non-trivial — names, administrative boundaries and reporting years rarely align — and the cost of doing so falls entirely on the end user. Most therefore stop at whichever single source they encountered first.
A reader-conditioned atlas, not another global ranking.
Hidden Gems treats the four issues above as design constraints rather than acknowledgements. It conditions every figure on the reader’s declared occupation and home city; it exposes the weights of the underlying criteria as direct inputs rather than burying them in a methodology footnote; it gives qualitative factors the same visual weight as monetary ones; and it joins six datasets at build time so the reader inherits the integration for free. The output is not a definitive ranking but a structured space the reader can interrogate.
What the project does not claim.
The atlas does not predict individual outcomes. Salary conditioning is occupation-level, not person-level; soft factors rely on proxies (e.g. daylight hours for “climate fit”) whose validity varies by reader. The contribution is methodological — making the conditioning explicit and the weighting visible — rather than predictive. The intended use is to narrow a search space from hundreds of cities to a tractable shortlist, on terms the reader has set themselves.
Continue to Design rationale →