What the numbers mean—and what they do not.

Compare Living Cost uses local-currency source evidence, explicit normalization, dated FX conversion and visible uncertainty. The current numeric model is CLC-v5.9-global-detail-1; this editorial/data-trust build is V7.3. Core city values were last reviewed in the September 2026 data cycle; source metadata and evidence taxonomy were re-audited in the September 2026 release cycle.

1. The published household model

The headline model is one adult renter living alone in a long-term, unfurnished one-bedroom home of roughly 35–55 m² (or the local equivalent) in a non-prime but transit-connected district, with a moderate non-luxury lifestyle, no car and no children.

2. What a metric represents

The six headline categories are monthly planning ranges. Rent is a one-bedroom non-prime renter baseline. Utilities/connectivity groups recurring household services. Groceries represents a moderate at-home food basket. Eating out is a moderate restaurant/takeaway allowance. Transit represents public-transport use consistent with the published persona. Household/personal covers basic recurring household and personal essentials.

The detailed layer uses an up-to-52-field schema. A field can represent an official average, median, published tariff, listed asking price, market index, or a documented derived value. The metric registry and derivation ledgers identify which kind of evidence is being used. Values from different source definitions are not silently treated as identical.

3. Missing data are missing—not zero

If a city does not have a detailed source-backed field, that field is omitted. It is never silently inserted as zero and is not counted as evidence. Side-by-side detailed comparisons show only fields that are source-backed for both selected cities. The six core categories remain available for all 47 current cities; deeper coverage varies materially by city.

4. Sources and normalization

1. Government statistical/administrative data and regulated public data.
2. Public utilities and transport agencies.
3. Transparent transaction or large-listing indexes with methodology.
4. Recognized local market research and major-retailer/local price observations.
5. Independent aggregators with transparent sourcing.
6. Crowdsourced cost-of-living databases only as reasonableness checks, not production inputs.

Raw sources often measure different household sizes, geographies, dwelling types or periods. A core component’s mode can therefore be a documented research-normalization anchor rather than a copied source number. Low/high values widen when geography, definition, volatility or source independence is weaker. Supplemental CPI or survey-status sources may be shown for context; they do not drive a numeric anchor unless that source is explicitly listed in the component’s source references.

4A. Core evidence tiers and the search-index gate

Every one of the six headline components now carries an explicit evidence tier. “Direct component evidence” means at least one captured numeric observation measures that category or a closely matched subcomponent. “Broad calibration evidence” means a real numeric household or per-person benchmark supports the normalization but does not directly measure that category. “Source-linked model anchor” means the source and definition have been identified, but the numeric observation has not yet been captured into the core ledger. None of these labels means the displayed planning midpoint is itself a raw observation.

Search indexing follows the evidence state. A city with zero captured numeric calibration components (direct or broad) and no deeper published evidence profile remains browsable but is set to noindex,follow and omitted from sitemap.xml. The page can return to the search sitemap after numeric core evidence is captured or a deeper verified profile is published.

Static city-pair pages show the evidence state for both sides of every category. Model-only rows remain visible for planning context, but matching model anchors are not described as measured equality. The full six-category total is always labeled as a planning model output.

5. Outliers and conflicting evidence

The model does not automatically average every observed number. A value that conflicts with the intended persona, unit, geography or period is treated as a definition problem first. Listing evidence is used as market context rather than as a precise city mean when the sample is heterogeneous. Large movements or contradictions require manual review before a new anchor is published; unresolved conflicts remain visible through wider ranges, lower confidence or an unpublished field.

6. Ranges, midpoint and uncertainty

Each core component has low/mode/high anchors. From those anchors, the total is simulated with triangular component distributions under correlation sensitivity scenarios (ρ = 0.00, 0.35 and 0.70). The public range is the outer P15–P85 envelope and the displayed model midpoint is P50. It is a planning estimate, not a claim that a real person will spend that exact amount.

Medium means the city has useful, reasonably current evidence but still relies on normalization or limited independent evidence families. Low / PROVISIONAL means the page is useful only as a wider planning range because a key source is old, geographically broad, definition-mismatched, listing-heavy or otherwise below the preferred evidence standard. A High label is deliberately withheld when the core model lacks enough independent direct evidence.

7. Currency and FX

Local-currency values are canonical. FX changes display only. This build uses Rates By Exchange Rate API; the bundled snapshot is 25 Sep 2026 UTC. The release process calls for a fresh snapshot before deployment, a stale warning after seven days, and a stronger stop after fourteen days. Historical screenshots can differ after FX refresh even when the local city record has not changed.

Taxes are included only when a particular source/metric definition says they are. The site does not assume every market listing, tariff or retail observation has identical tax treatment; the source/basis text is authoritative for that metric.

8. Salary equivalency

The optional salary output is deliberately simple: entered annual salary × (destination one-person core-cost midpoint ÷ source one-person core-cost midpoint), with the entered salary and city midpoints converted through the same FX snapshot for the calculation. It includes housing because rent is part of the core midpoint.

This ratio is intentionally a quick screening tool, not the recommended way to set a personal relocation salary target. For a real move, use the salary-planning guide to build from after-tax monthly needs, excluded costs, savings, benefits and household circumstances.

9. Childcare, income and tax detail

Childcare is not included in the one-person headline. Where a city has a separate source-backed childcare field, it is shown only as detail. Income/earnings and tax-context rows are also separate detailed evidence and are not silently injected into the core spending model. Cities without those fields show them as unavailable rather than estimated.

10. Precision

Core planning totals and salary-ratio outputs are rounded to whole local-currency units. Detailed official tariffs or rates can retain decimals when the source itself supports that precision. Integer observations are no longer formatted with artificial “.00” solely because the amount is small. Converted values can show currency decimals where appropriate, but conversion precision does not make the underlying cost estimate more certain.

11. Publication and refresh gate

Every public metric requires a valid source state. Source-backed fields are published; unresolved QA fields remain out of the interface. Volatile rent, food and utilities evidence should be reviewed more often than slower household surveys. A city can remain indexable with limited prose when its structured data, sources, dates, currency, uncertainty and comparison utility are meaningful; the quality gate is based on data value, not word count.

Source registry · Public metric registry · City data · Core derivations · Traceability gaps · Detail derivations · FX metadata