Small fixes

pipeline/transform/transform_poi.py

@@ -1579,8 +1579,11 @@ def transform_gias_schools(
 # Store". GEOLYTIX is authoritative for its chains, so an OSM grocery row that
 # sits on top of a GEOLYTIX point AND carries that point's brand name is the
 # same physical store and is dropped. Independent corner shops never carry a
-# chain brand, so they are kept.
-GROCERY_DEDUP_RADIUS_M = 50.0
+# chain brand, so they are kept. 100m (not 50m) because OSM and GEOLYTIX
+# routinely place the same branded store 50-100m apart (entrance vs centroid vs
+# car park); at 50m ~166 same-brand duplicates (Iceland, Morrisons Daily, Tesco
+# Express, Sainsbury's Local, ...) survived just outside the radius.
+GROCERY_DEDUP_RADIUS_M = 100.0
 
 # Brand-token aliases so an OSM name spelt differently from the GEOLYTIX brand
 # still matches. GEOLYTIX's "Co-op" tokenises to "coop", but OSM frequently
@@ -1591,15 +1594,34 @@ _GROCERY_TOKEN_ALIASES = {
     "cooperatives": "coop",
 }
 
+# Brand tokens shorter than the 3-char floor that must still match. "M&S" strips
+# to "ms" (len 2) and would otherwise tokenise to set(), so M&S never deduped and
+# OSM M&S grocery rows survived as phantom duplicates of the GEOLYTIX M&S chain.
+# "M&S" is the only GEOLYTIX brand whose alnum form is <3 chars, so this
+# allowlist is safe (no other brand collides).
+_GROCERY_SHORT_BRAND_TOKENS = {"ms"}
+
+# OSM tags Iceland/Farmfoods stores shop/frozen_food, which CATEGORY_MAP routes
+# to "Deli & Specialty". The grocery dedup drops the ones colocated with a
+# GEOLYTIX twin, but genuine stores GEOLYTIX lacks would otherwise survive
+# mislabelled as deli, shadowing the brand pill and falling outside the headline
+# grocery metric. Relabel the survivors (matched by name) to the brand so they
+# line up with the GEOLYTIX rows.
+OSM_BRAND_RELABEL = {
+    "Iceland": "Iceland",
+    "Farmfoods": "Farmfoods",
+}
+
 
 def _significant_tokens(name: str | None) -> set[str]:
-    """Lower-case alphanumeric tokens of length >= 3 from a POI name (aliased)."""
+    """Lower-case alphanumeric tokens of length >= 3 from a POI name (aliased),
+    plus short brand tokens in _GROCERY_SHORT_BRAND_TOKENS (e.g. "ms" for M&S)."""
     if not name:
         return set()
     tokens: set[str] = set()
     for raw in str(name).lower().split():
         token = "".join(ch for ch in raw if ch.isalnum())
-        if len(token) >= 3:
+        if len(token) >= 3 or token in _GROCERY_SHORT_BRAND_TOKENS:
             tokens.add(_GROCERY_TOKEN_ALIASES.get(token, token))
     return tokens
 
@@ -1608,7 +1630,14 @@ def _significant_tokens(name: str | None) -> set[str]:
 # gaps. Where NaPTAN already has a stop within this radius the area is covered,
 # so the colocated OSM platform is dropped to avoid double-counting; OSM
 # platforms with no nearby NaPTAN stop (the gaps) are kept.
-BUS_STOP_DEDUP_RADIUS_M = 50.0
+#
+# 100 m, not 50 m: at 50 m ~24.6k OSM platforms 50-100 m from a NaPTAN stop
+# survived and double-counted (16.7k at 50-75 m, 7.8k at 75-100 m). The NaPTAN
+# nearest-neighbour spacing (opposite-side-of-road pairs) has a 166.7 m median
+# (p25=98.7 m), so 100 m clears the 50-100 m duplicate spike while leaving a
+# genuine opposite-direction stop (~166 m away) in place; 125-150 m would start
+# dropping those real opposite-side stops.
+BUS_STOP_DEDUP_RADIUS_M = 100.0
 
 
 def osm_stops_near_naptan(
@@ -1652,9 +1681,9 @@ def osm_groceries_colocated_with_geolytix(
 
     An OSM Groceries row is a duplicate when a GEOLYTIX point lies within
     ``radius_m`` metres AND that point's brand tokens (its ``category``, e.g.
-    "Tesco", "Co-op") are all present in the OSM row's name — i.e. the same
-    physical branded store. Brands with no token >= 3 chars (e.g. "M&S") never
-    match, so they are conservatively kept rather than risk a false drop.
+    "Tesco", "Co-op", "M&S") are all present in the OSM row's name — i.e. the
+    same physical branded store. Short brands like "M&S" match via
+    _GROCERY_SHORT_BRAND_TOKENS; brands that still tokenise to set() are kept.
 
     ``osm_groceries`` needs columns ``id``, ``name``, ``lat``, ``lng``;
     ``geolytix`` needs ``category`` (the brand), ``lat``, ``lng``.
@@ -1774,6 +1803,31 @@ def transform(
     # pre-deduplicated.
     lf = lf.unique(subset=["id", "category"], keep="first", maintain_order=True)
 
+    # Relabel OSM Iceland/Farmfoods rows that CATEGORY_MAP put in "Deli &
+    # Specialty" to the brand, so the genuine stores GEOLYTIX lacks (the grocery
+    # dedup drops the colocated twins) line up with the GEOLYTIX brand pills and
+    # the headline grocery metric instead of shadowing them as deli.
+    for _brand_name, _brand in OSM_BRAND_RELABEL.items():
+        _is_brand = (
+            (pl.col("group") == "Groceries")
+            & (pl.col("category") == "Deli & Specialty")
+            & (pl.col("name") == _brand_name)
+        )
+        lf = lf.with_columns(
+            pl.when(_is_brand)
+            .then(pl.lit(_brand))
+            .otherwise(pl.col("category"))
+            .alias("category"),
+            pl.when(_is_brand)
+            .then(pl.lit(_brand))
+            .otherwise(pl.col("icon_category"))
+            .alias("icon_category"),
+            pl.when(_is_brand)
+            .then(pl.lit("\U0001f6d2"))
+            .otherwise(pl.col("emoji"))
+            .alias("emoji"),
+        )
+
     naptan_df = pl.scan_parquet(naptan_path).collect()
     mask = in_england_mask(
         boundary_path,