merge: Score v6 — World Bank global economic data for non-EU countries

Non-EU countries (AR, MX, AE, AU, etc.) previously got NULL for
median_income_pps and pli_construction, falling back to EU-calibrated
defaults (15K PPS, PLI=100) that produced wrong scores.

New World Bank WDI extractor fetches GNI per capita PPP and price level
ratio for 215 countries. dim_countries uses Germany as calibration anchor
to scale WB values into the Eurostat range (dynamic ratio, self-corrects
as both sources update). EU countries keep exact Eurostat values.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

CHANGELOG.md

@@ -7,6 +7,10 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 ## [Unreleased]
 
 ### Changed
+- **Score v6: Global economic data** — `dim_countries.median_income_pps` and `pli_construction` now cover all target markets, not just EU. World Bank WDI indicators (GNI per capita PPP + price level ratio) fill gaps for non-EU countries (AR, MX, AE, AU, etc.) with values calibrated to the Eurostat scale using Germany as anchor. EU countries keep exact Eurostat values. New extractor (`worldbank.py`), staging model (`stg_worldbank_income`), and `dim_countries` fallback CTEs. No changes to scoring formulas — the fix is upstream in the data layer.
+- **Market Score v3 → v4** — fixes Spain averaging 54 (should be 65-80). Four calibration changes: count gate threshold lowered from 5 → 3 venues (3 establishes a market pattern), density ceiling lowered from LN(21) → LN(11) (10/100k is reachable for mature markets), demand evidence fallback raised from 0.4 → 0.65 multiplier with 0.3 floor (existence of venues IS evidence of demand), economic context ceiling changed from income/200 → income/25000 (actual discrimination instead of free 10 pts for everyone).
+- **Opportunity Score v4 → v5** — fixes structural flaws: supply gap (30pts) + catchment gap (15pts) merged into single supply deficit (35pts, GREATEST of density gap and distance gap) eliminating ~80% correlated double-count. New sports culture signal (10pts) using tennis court density as racquet-sport adoption proxy. New construction affordability signal (5pts) using income relative to PLI construction costs from `dim_countries`. Economic power reduced from 20 → 15pts. New dependency on `foundation.dim_countries` for `pli_construction`.
+
 - **Unified `location_profiles` serving model** — merged `city_market_profile` and `location_opportunity_profile` into a single `serving.location_profiles` table at `(country_code, geoname_id)` grain. Both Marktreife-Score (Market Score) and Marktpotenzial-Score (Opportunity Score) are now computed per location. City data enriched via LEFT JOIN `dim_cities` on `geoname_id`. Downstream models (`planner_defaults`, `pseo_city_costs_de`, `pseo_city_pricing`) updated to query `location_profiles` directly. `city_padel_venue_count` (exact from dim_cities) distinguished from `padel_venue_count` (spatial 5km from dim_locations).
 - **Both scores on all map tooltips** — country map shows avg Market Score + avg Opportunity Score; city map shows Market Score + Opportunity Score per city; opportunity map shows Opportunity Score + Market Score per location. All score labels use the trademarked "Padelnomics Market Score" / "Padelnomics Opportunity Score" names.
 - **API endpoints** — `/api/markets/countries.json` adds `avg_opportunity_score`; `/api/markets/<country>/cities.json` adds `opportunity_score`; `/api/opportunity/<country>.json` adds `market_score`.

extract/padelnomics_extract/pyproject.toml

@@ -22,6 +22,7 @@ extract-census-usa-income = "padelnomics_extract.census_usa_income:main"
 extract-ons-uk = "padelnomics_extract.ons_uk:main"
 extract-geonames = "padelnomics_extract.geonames:main"
 extract-gisco = "padelnomics_extract.gisco:main"
+extract-worldbank = "padelnomics_extract.worldbank:main"
 
 [build-system]
 requires = ["hatchling"]

extract/padelnomics_extract/src/padelnomics_extract/all.py

@@ -7,7 +7,7 @@ A graphlib.TopologicalSorter schedules them: tasks with no unmet dependencies
 run immediately in parallel; each completion may unlock new tasks.
 
 Current dependency graph:
-  - All 9 non-availability extractors have no dependencies (run in parallel)
+  - All 10 non-availability extractors have no dependencies (run in parallel)
   - playtomic_availability depends on playtomic_tenants (starts as soon as
     tenants finishes, even if other extractors are still running)
 """
@@ -38,6 +38,8 @@ from .playtomic_availability import EXTRACTOR_NAME as AVAILABILITY_NAME
 from .playtomic_availability import extract as extract_availability
 from .playtomic_tenants import EXTRACTOR_NAME as TENANTS_NAME
 from .playtomic_tenants import extract as extract_tenants
+from .worldbank import EXTRACTOR_NAME as WORLDBANK_NAME
+from .worldbank import extract as extract_worldbank
 
 logger = setup_logging("padelnomics.extract")
 
@@ -54,6 +56,7 @@ EXTRACTORS: dict[str, tuple] = {
     GEONAMES_NAME:             (extract_geonames, []),
     GISCO_NAME:                (extract_gisco, []),
     TENANTS_NAME:              (extract_tenants, []),
+    WORLDBANK_NAME:            (extract_worldbank, []),
     AVAILABILITY_NAME:         (extract_availability, [TENANTS_NAME]),
 }
 

extract/padelnomics_extract/src/padelnomics_extract/worldbank.py

@@ -0,0 +1,153 @@
+"""World Bank WDI extractor — GNI per capita PPP and price level ratio.
+
+Fetches two indicators (one API call each, no key required):
+  - NY.GNP.PCAP.PP.CD  — GNI per capita, PPP (international $)
+  - PA.NUS.PPPC.RF      — Price level ratio (PPP conversion factor / exchange rate)
+
+These provide global fallbacks behind Eurostat for dim_countries.median_income_pps
+and dim_countries.pli_construction (see dim_countries.sql for calibration logic).
+
+API: World Bank API v2 — https://api.worldbank.org/v2/
+No API key required. No env vars.
+
+Landing: {LANDING_DIR}/worldbank/{year}/{month}/wdi_indicators.json.gz
+Output:  {"rows": [{"country_code": "DE", "indicator": "NY.GNP.PCAP.PP.CD",
+                    "ref_year": 2023, "value": 74200.0}, ...], "count": N}
+"""
+
+import json
+import sqlite3
+from pathlib import Path
+
+import niquests
+
+from ._shared import HTTP_TIMEOUT_SECONDS, run_extractor, setup_logging
+from .utils import get_last_cursor, landing_path, write_gzip_atomic
+
+logger = setup_logging("padelnomics.extract.worldbank")
+
+EXTRACTOR_NAME = "worldbank"
+
+INDICATORS = ["NY.GNP.PCAP.PP.CD", "PA.NUS.PPPC.RF"]
+# 6 years of data — we take the latest non-null per country in staging
+DATE_RANGE = "2019:2025"
+MAX_PER_PAGE = 5000
+MAX_PAGES = 3
+
+WDI_BASE_URL = "https://api.worldbank.org/v2/country/all/indicator"
+
+# WB aggregate codes that look like real 2-letter country codes.
+# These are regional/income-group aggregates, not actual countries.
+_WB_AGGREGATE_CODES = frozenset({
+    "EU", "OE",
+    "XC", "XD", "XE", "XF", "XG", "XH", "XI", "XJ", "XL", "XM",
+    "XN", "XO", "XP", "XQ", "XR", "XS", "XT", "XU", "XV", "XY",
+    "ZF", "ZG", "ZH", "ZI", "ZJ", "ZQ", "ZT",
+    "V1", "V2", "V3", "V4",
+})
+
+
+def _normalize_country_code(wb_code: str) -> str | None:
+    """Normalize WB country code to ISO alpha-2. Returns None for aggregates."""
+    code = wb_code.strip().upper()
+    if len(code) != 2:
+        return None
+    # Reject codes starting with a digit (e.g. "1W" for World)
+    if code[0].isdigit():
+        return None
+    if code in _WB_AGGREGATE_CODES:
+        return None
+    return code
+
+
+def _fetch_indicator(
+    session: niquests.Session,
+    indicator: str,
+) -> list[dict]:
+    """Fetch all records for one indicator. Returns list of row dicts."""
+    rows: list[dict] = []
+    page = 1
+
+    while page <= MAX_PAGES:
+        url = (
+            f"{WDI_BASE_URL}/{indicator}"
+            f"?format=json&date={DATE_RANGE}&per_page={MAX_PER_PAGE}&page={page}"
+        )
+        logger.info("GET %s page %d", indicator, page)
+        resp = session.get(url, timeout=HTTP_TIMEOUT_SECONDS * 2)
+        resp.raise_for_status()
+
+        data = resp.json()
+        assert isinstance(data, list) and len(data) == 2, (
+            f"unexpected WB response shape for {indicator}: {type(data)}, len={len(data)}"
+        )
+        meta, records = data
+        total_pages = meta.get("pages", 1)
+
+        if records is None:
+            logger.warning("WB returned null data for %s page %d", indicator, page)
+            break
+
+        for record in records:
+            value = record.get("value")
+            if value is None:
+                continue
+            country_code = _normalize_country_code(record["country"]["id"])
+            if country_code is None:
+                continue
+            rows.append({
+                "country_code": country_code,
+                "indicator": indicator,
+                "ref_year": int(record["date"]),
+                "value": float(value),
+            })
+
+        if page >= total_pages:
+            break
+        page += 1
+
+    return rows
+
+
+def extract(
+    landing_dir: Path,
+    year_month: str,
+    conn: sqlite3.Connection,
+    session: niquests.Session,
+) -> dict:
+    """Fetch WDI indicators. Skips if already run this month."""
+    last_cursor = get_last_cursor(conn, EXTRACTOR_NAME)
+    if last_cursor == year_month:
+        logger.info("already have data for %s — skipping", year_month)
+        return {"files_written": 0, "files_skipped": 1, "bytes_written": 0}
+
+    rows: list[dict] = []
+    for indicator in INDICATORS:
+        indicator_rows = _fetch_indicator(session, indicator)
+        logger.info("%s: %d records", indicator, len(indicator_rows))
+        rows.extend(indicator_rows)
+
+    assert len(rows) >= 200, f"expected ≥200 WB records, got {len(rows)} — API may have changed"
+    logger.info("total: %d WDI records", len(rows))
+
+    year, month = year_month.split("/")
+    dest_dir = landing_path(landing_dir, "worldbank", year, month)
+    dest = dest_dir / "wdi_indicators.json.gz"
+    payload = json.dumps({"rows": rows, "count": len(rows)}).encode()
+    bytes_written = write_gzip_atomic(dest, payload)
+    logger.info("written %s bytes compressed", f"{bytes_written:,}")
+
+    return {
+        "files_written": 1,
+        "files_skipped": 0,
+        "bytes_written": bytes_written,
+        "cursor_value": year_month,
+    }
+
+
+def main() -> None:
+    run_extractor(EXTRACTOR_NAME, extract)
+
+
+if __name__ == "__main__":
+    main()

infra/supervisor/workflows.toml

@@ -72,3 +72,8 @@ description = "UK local authority population estimates from ONS"
 module = "padelnomics_extract.gisco"
 schedule = "0 0 1 1 *"
 description = "EU geographic boundaries (NUTS2 polygons) from Eurostat GISCO"
+
+[worldbank]
+module = "padelnomics_extract.worldbank"
+schedule = "monthly"
+description = "GNI per capita PPP + price level ratio from World Bank WDI"

transform/sqlmesh_padelnomics/models/foundation/dim_countries.sql

@@ -2,10 +2,14 @@
 --
 -- Consolidates data previously duplicated across dim_cities and dim_locations:
 --   - country_name_en / country_slug  (was: ~50-line CASE blocks in both models)
---   - median_income_pps               (was: country_income CTE in both models)
+--   - median_income_pps               (Eurostat PPS preferred, World Bank GNI PPP fallback)
---   - energy prices, labour costs, PLI indices  (new — from Eurostat datasets)
+--   - energy prices, labour costs, PLI indices  (Eurostat, WB price level ratio fallback)
 --   - cost override columns for the financial calculator
 --
+-- World Bank fallback: for non-EU countries (AR, MX, AE, AU, etc.), income and PLI
+-- are derived from WB WDI indicators calibrated to the Eurostat scale using Germany
+-- as anchor. See de_calibration CTE. EU countries keep exact Eurostat values.
+--
 -- Used by: dim_cities, dim_locations, pseo_city_costs_de, planner_defaults.
 -- Grain: country_code (one row per ISO 3166-1 alpha-2 country code).
 -- Kind: FULL — small table (~40 rows), full refresh daily.
@@ -82,6 +86,26 @@ de_elec AS (
 de_gas AS (
   SELECT gas_eur_gj FROM latest_gas WHERE country_code = 'DE'
 ),
+-- Latest World Bank WDI per country (GNI PPP + price level ratio)
+latest_wb AS (
+  SELECT country_code, gni_ppp, price_level_ratio, ref_year AS wb_year
+  FROM staging.stg_worldbank_income
+  WHERE gni_ppp IS NOT NULL OR price_level_ratio IS NOT NULL
+  QUALIFY ROW_NUMBER() OVER (PARTITION BY country_code ORDER BY ref_year DESC) = 1
+),
+-- Germany calibration anchor: Eurostat PPS + WB GNI PPP + WB price ratio + Eurostat PLI construction.
+-- Used to scale World Bank values into Eurostat-comparable ranges.
+-- Single row; if DE is missing from any source, that ratio produces NULL (safe fallthrough).
+de_calibration AS (
+  SELECT
+    i.median_income_pps   AS de_eurostat_pps,
+    wb.gni_ppp            AS de_gni_ppp,
+    wb.price_level_ratio  AS de_price_level_ratio,
+    p.construction        AS de_pli_construction
+  FROM (SELECT median_income_pps FROM latest_income WHERE country_code = 'DE') i
+  CROSS JOIN (SELECT gni_ppp, price_level_ratio FROM latest_wb WHERE country_code = 'DE') wb
+  CROSS JOIN (SELECT construction FROM pli_pivoted WHERE country_code = 'DE') p
+),
 -- All distinct country codes from any source
 all_countries AS (
   SELECT country_code FROM latest_income
@@ -93,6 +117,8 @@ all_countries AS (
   SELECT country_code FROM latest_labour
   UNION
   SELECT country_code FROM pli_pivoted
+  UNION
+  SELECT country_code FROM latest_wb
   -- Ensure known padel markets appear even if Eurostat doesn't cover them yet
   UNION ALL
   SELECT unnest(['DE','ES','GB','FR','IT','PT','AT','CH','NL','BE','SE','NO','DK','FI',
@@ -149,15 +175,21 @@ SELECT
       ELSE ac.country_code
     END, '[^a-zA-Z0-9]+', '-'
   ))                                                             AS country_slug,
-  -- Income data
+  -- Income: Eurostat PPS preferred, World Bank GNI PPP scaled to PPS as fallback
-  i.median_income_pps,
+  COALESCE(
-  i.income_year,
+    i.median_income_pps,
+    ROUND(wb.gni_ppp * (de_cal.de_eurostat_pps / NULLIF(de_cal.de_gni_ppp, 0)), 0)
+  )                                                            AS median_income_pps,
+  COALESCE(i.income_year, wb.wb_year)                          AS income_year,
   -- Raw energy and labour data (for reference / future staffed-scenario use)
   e.electricity_eur_kwh,
   g.gas_eur_gj,
   la.labour_cost_eur_hour,
-  -- PLI indices per category (EU27=100)
+  -- PLI construction: Eurostat preferred, World Bank price level ratio scaled to PLI as fallback
-  p.construction                                                 AS pli_construction,
+  COALESCE(
+    p.construction,
+    ROUND(wb.price_level_ratio / NULLIF(de_cal.de_price_level_ratio, 0) * de_cal.de_pli_construction, 1)
+  )                                                              AS pli_construction,
   p.housing                                                      AS pli_housing,
   p.services                                                     AS pli_services,
   p.misc                                                         AS pli_misc,
@@ -278,8 +310,10 @@ LEFT JOIN latest_electricity e  ON ac.country_code = e.country_code
 LEFT JOIN latest_gas g          ON ac.country_code = g.country_code
 LEFT JOIN latest_labour la      ON ac.country_code = la.country_code
 LEFT JOIN pli_pivoted p         ON ac.country_code = p.country_code
+LEFT JOIN latest_wb wb         ON ac.country_code = wb.country_code
 CROSS JOIN de_pli de_p
 CROSS JOIN de_elec de_e
 CROSS JOIN de_gas de_g
+CROSS JOIN de_calibration de_cal
 -- Enforce grain
 QUALIFY ROW_NUMBER() OVER (PARTITION BY ac.country_code ORDER BY ac.country_code) = 1

transform/sqlmesh_padelnomics/models/serving/location_profiles.sql

@@ -5,30 +5,36 @@
 --
 -- Two scores per location:
 --
--- Padelnomics Market Score (Marktreife-Score v3, 0–100):
+-- Padelnomics Market Score (Marktreife-Score v4, 0–100):
 --   "How mature/established is this padel market?"
 --   Only meaningful for locations matched to a dim_cities row (city_slug IS NOT NULL)
 --   with padel venues. 0 for all other locations.
 --
---   40 pts  supply development — log-scaled density (LN ceiling 20/100k) × count gate
+--   v4 changes: lower count gate (5→3), lower density ceiling (LN(21)→LN(11)),
---   25 pts  demand evidence   — occupancy when available; 40% density proxy otherwise
+--   better demand fallback (0.4→0.65 with 0.3 floor), economic context discrimination (200→25K).
+--
+--   40 pts  supply development — log-scaled density (LN ceiling 10/100k) × count gate (3)
+--   25 pts  demand evidence   — occupancy when available; 65% density proxy + 0.3 floor otherwise
 --   15 pts  addressable market — log-scaled population, ceiling 1M
---   10 pts  economic context  — income PPS normalised to 200 ceiling
+--   10 pts  economic context  — income PPS normalised to 25,000 ceiling
 --   10 pts  data quality      — completeness discount
 --
--- Padelnomics Opportunity Score (Marktpotenzial-Score v4, 0–100):
+-- Padelnomics Opportunity Score (Marktpotenzial-Score v5, 0–100):
 --   "Where should I build a padel court?"
---   Computed for ALL locations — zero-court locations score highest on supply gap.
+--   Computed for ALL locations — zero-court locations score highest on supply deficit.
---   H3 catchment methodology: addressable market and supply gap use a regional
+--   H3 catchment methodology: addressable market and supply deficit use a regional
 --   H3 catchment (res-5 cell + 6 neighbours, ~24km radius).
 --
---   25 pts  addressable market — log-scaled catchment population, ceiling 500K
+--   v5 changes: merge supply gap + catchment gap → single supply deficit (35 pts),
---   20 pts  economic power     — income PPS, normalised to 35,000
+--   add sports culture proxy (10 pts, tennis density), add construction affordability (5 pts),
---   30 pts  supply gap         — inverted catchment venue density; 0 courts = full marks
+--   reduce economic power from 20 → 15 pts.
---   15 pts  catchment gap      — distance to nearest padel court
+--
---   10 pts  market validation  — country-level avg market maturity (from market_scored CTE).
+--   25 pts  addressable market        — log-scaled catchment population, ceiling 500K
---                               Replaces sports culture proxy (v3: tennis data was all zeros).
+--   15 pts  economic power            — income PPS, normalised to 35,000
---                               ES (~60/100) → ~6 pts, SE (~35/100) → ~3.5 pts, unknown → 5 pts.
+--   35 pts  supply deficit            — max(density gap, distance gap); eliminates double-count
+--   10 pts  sports culture            — tennis court density as racquet-sport adoption proxy
+--    5 pts  construction affordability — income relative to construction costs (PLI)
+--   10 pts  market validation         — country-level avg market maturity (from market_scored CTE)
 --
 -- Consumers query directly with WHERE filters:
 --   cities API:       WHERE country_slug = ? AND city_slug IS NOT NULL
@@ -107,7 +113,7 @@ city_match AS (
     ORDER BY c.padel_venue_count DESC
   ) = 1
 ),
--- Pricing / occupancy from Playtomic (via city_slug) + H3 catchment
+-- Pricing / occupancy from Playtomic (via city_slug) + H3 catchment + country PLI
 with_pricing AS (
   SELECT
     b.*,
@@ -120,6 +126,7 @@ with_pricing AS (
     vpb.median_occupancy_rate,
     vpb.median_daily_revenue_per_venue,
     vpb.price_currency,
+    dc.pli_construction,
     COALESCE(ct.catchment_population, b.population)::BIGINT           AS catchment_population,
     COALESCE(ct.catchment_padel_courts, b.padel_venue_count)::INTEGER AS catchment_padel_courts
   FROM base b
@@ -131,6 +138,8 @@ with_pricing AS (
     AND cm.city_slug = vpb.city_slug
   LEFT JOIN catchment ct
     ON b.geoname_id = ct.geoname_id
+  LEFT JOIN foundation.dim_countries dc
+    ON b.country_code = dc.country_code
 ),
 -- Step 1: market score only — needed first so we can aggregate country averages.
 market_scored AS (
@@ -146,34 +155,38 @@ market_scored AS (
       WHEN population > 0 OR  COALESCE(city_padel_venue_count, 0) > 0 THEN 0.5
       ELSE 0.0
     END AS data_confidence,
-    -- ── Market Score (Marktreife-Score v3) ──────────────────────────────────
+    -- ── Market Score (Marktreife-Score v4) ──────────────────────────────────
     -- 0 when no city match or no venues (city_padel_venue_count NULL or 0)
     CASE WHEN COALESCE(city_padel_venue_count, 0) > 0 THEN
       ROUND(
         -- Supply development (40 pts)
+        -- density ceiling 10/100k (LN(11)), count gate 3 venues
         40.0 * LEAST(1.0, LN(
             COALESCE(
               CASE WHEN population > 0
                 THEN COALESCE(city_padel_venue_count, 0)::DOUBLE / population * 100000
                 ELSE 0 END
-            , 0) + 1) / LN(21))
+            , 0) + 1) / LN(11))
-             * LEAST(1.0, COALESCE(city_padel_venue_count, 0) / 5.0)
+             * LEAST(1.0, COALESCE(city_padel_venue_count, 0) / 3.0)
         -- Demand evidence (25 pts)
+        -- with occupancy: scale to 65% target. Without: 65% of supply proxy + 0.3 floor
+        -- (existence of venues IS evidence of demand)
         + 25.0 * CASE
             WHEN median_occupancy_rate IS NOT NULL
               THEN LEAST(1.0, median_occupancy_rate / 0.65)
-            ELSE 0.4 * LEAST(1.0, LN(
+            ELSE GREATEST(0.3, 0.65 * LEAST(1.0, LN(
                 COALESCE(
                   CASE WHEN population > 0
                     THEN COALESCE(city_padel_venue_count, 0)::DOUBLE / population * 100000
                     ELSE 0 END
-                , 0) + 1) / LN(21))
+                , 0) + 1) / LN(11))
-                     * LEAST(1.0, COALESCE(city_padel_venue_count, 0) / 5.0)
+                     * LEAST(1.0, COALESCE(city_padel_venue_count, 0) / 3.0))
           END
         -- Addressable market (15 pts)
         + 15.0 * LEAST(1.0, LN(GREATEST(population, 1)) / LN(1000000))
         -- Economic context (10 pts)
-        + 10.0 * LEAST(1.0, COALESCE(median_income_pps, 100) / 200.0)
+        -- ceiling 25,000 PPS discriminates between wealthy and poorer markets
+        + 10.0 * LEAST(1.0, COALESCE(median_income_pps, 15000) / 25000.0)
         -- Data quality (10 pts)
         + 10.0 * CASE
             WHEN population > 0 AND COALESCE(city_padel_venue_count, 0) > 0 THEN 1.0
@@ -199,23 +212,35 @@ country_market AS (
 -- Step 3: add opportunity_score using country market validation signal.
 scored AS (
   SELECT ms.*,
-    -- ── Opportunity Score (Marktpotenzial-Score v4, H3 catchment) ──────────
+    -- ── Opportunity Score (Marktpotenzial-Score v5, H3 catchment) ──────────
     ROUND(
       -- Addressable market (25 pts): log-scaled catchment population, ceiling 500K
       25.0 * LEAST(1.0, LN(GREATEST(catchment_population, 1)) / LN(500000))
-      -- Economic power (20 pts): income PPS normalised to 35,000
+      -- Economic power (15 pts): income PPS normalised to 35,000
-      + 20.0 * LEAST(1.0, COALESCE(median_income_pps, 15000) / 35000.0)
+      + 15.0 * LEAST(1.0, COALESCE(median_income_pps, 15000) / 35000.0)
-      -- Supply gap (30 pts): inverted catchment venue density
+      -- Supply deficit (35 pts): max of density gap and distance gap.
-      + 30.0 * GREATEST(0.0, 1.0 - COALESCE(
+      -- Merges old supply gap (30) + catchment gap (15) which were ~80% correlated.
-          CASE WHEN catchment_population > 0
+      + 35.0 * GREATEST(
-            THEN GREATEST(catchment_padel_courts, COALESCE(city_padel_venue_count, 0))::DOUBLE / catchment_population * 100000
+          -- density-based gap (H3 catchment): 0 courts = 1.0, 8/100k = 0.0
-            ELSE 0.0
+          GREATEST(0.0, 1.0 - COALESCE(
-          END, 0.0) / 8.0)
+            CASE WHEN catchment_population > 0
-      -- Catchment gap (15 pts): distance to nearest court
+              THEN GREATEST(catchment_padel_courts, COALESCE(city_padel_venue_count, 0))::DOUBLE / catchment_population * 100000
-      + 15.0 * COALESCE(LEAST(1.0, nearest_padel_court_km / 30.0), 0.5)
+              ELSE 0.0
+            END, 0.0) / 8.0),
+          -- distance-based gap: 30km+ = 1.0, 0km = 0.0; NULL = 0.5
+          COALESCE(LEAST(1.0, nearest_padel_court_km / 30.0), 0.5)
+        )
+      -- Sports culture (10 pts): tennis density as racquet-sport adoption proxy.
+      -- Ceiling 50 courts within 25km. Harmless when tennis data is zero (contributes 0).
+      + 10.0 * LEAST(1.0, COALESCE(tennis_courts_within_25km, 0) / 50.0)
+      -- Construction affordability (5 pts): income purchasing power relative to build costs.
+      -- PLI construction is EU27=100 index. High income + low construction cost = high score.
+      + 5.0 * LEAST(1.0,
+          COALESCE(median_income_pps, 15000) / 35000.0
+          / GREATEST(0.5, COALESCE(pli_construction, 100.0) / 100.0)
+        )
       -- Market validation (10 pts): country-level avg market maturity.
-      -- Replaces sports culture (v3 tennis data was all zeros = dead code).
+      -- ES (~70/100): proven demand → ~7 pts. SE (~35/100): emerging → ~3.5 pts.
-      -- ES (~60/100): proven demand → ~6 pts. SE (~35/100): struggling → ~3.5 pts.
       -- NULL (no courts in country yet): 0.5 neutral → 5 pts (untested, not penalised).
       + 10.0 * COALESCE(cm.country_avg_market_score / 100.0, 0.5)
     , 1) AS opportunity_score

transform/sqlmesh_padelnomics/models/staging/stg_worldbank_income.sql

@@ -0,0 +1,41 @@
+-- World Bank WDI indicators: GNI per capita PPP and price level ratio.
+-- Pivoted to one row per (country_code, ref_year) with both indicators as columns.
+--
+-- Source: data/landing/worldbank/{year}/{month}/wdi_indicators.json.gz
+-- Extracted by: worldbank.py
+-- Used by: dim_countries (fallback behind Eurostat for non-EU countries)
+
+MODEL (
+  name staging.stg_worldbank_income,
+  kind FULL,
+  cron '@daily',
+  grain (country_code, ref_year)
+);
+
+WITH parsed AS (
+  SELECT
+    row ->> 'country_code'                          AS country_code,
+    TRY_CAST(row ->> 'ref_year' AS INTEGER)         AS ref_year,
+    row ->> 'indicator'                             AS indicator,
+    TRY_CAST(row ->> 'value' AS DOUBLE)             AS value,
+    CURRENT_DATE                                    AS extracted_date
+  FROM (
+    SELECT UNNEST(rows) AS row
+    FROM read_json(
+      @LANDING_DIR || '/worldbank/*/*/wdi_indicators.json.gz',
+      auto_detect = true
+    )
+  )
+  WHERE (row ->> 'country_code') IS NOT NULL
+)
+SELECT
+  country_code,
+  ref_year,
+  MAX(value) FILTER (WHERE indicator = 'NY.GNP.PCAP.PP.CD') AS gni_ppp,
+  MAX(value) FILTER (WHERE indicator = 'PA.NUS.PPPC.RF')     AS price_level_ratio,
+  MAX(extracted_date)                                         AS extracted_date
+FROM parsed
+WHERE value IS NOT NULL
+  AND value > 0
+  AND LENGTH(country_code) = 2
+GROUP BY country_code, ref_year

web/src/padelnomics/admin/pipeline_routes.py

@@ -111,7 +111,7 @@ _DAG: dict[str, list[str]] = {
     "fct_daily_availability": ["fct_availability_slot", "dim_venue_capacity"],
     # Serving
     "venue_pricing_benchmarks": ["fct_daily_availability"],
-    "location_profiles": ["dim_locations", "dim_cities", "venue_pricing_benchmarks"],
+    "location_profiles": ["dim_locations", "dim_cities", "dim_countries", "venue_pricing_benchmarks"],
     "planner_defaults": ["venue_pricing_benchmarks", "location_profiles"],
     "pseo_city_costs_de": [
         "location_profiles", "planner_defaults",