diff --git a/Makefile.data b/Makefile.data
index bfcc286..660b9e1 100644
--- a/Makefile.data
+++ b/Makefile.data
@@ -29,10 +29,11 @@ PROPERTIES_PQ   := $(DATA_DIR)/properties.parquet
 MERGE_STAMP     := $(DATA_DIR)/.merge_done
 PRICE_INDEX     := $(DATA_DIR)/price_index.parquet
 PRICES_STAMP    := $(DATA_DIR)/.prices_done
-EPC             := $(MANUAL_DATA)/certificates.csv
+EPC             := $(MANUAL_DATA)/domestic-csv.zip
 ETHNICITY       := $(DATA_DIR)/ethnicity_by_la.parquet
 CRIME_DIR       := $(MANUAL_DATA)/crime
 CRIME           := $(DATA_DIR)/crime_by_lsoa.parquet
+CRIME_STAMP     := $(CRIME_DIR)/.downloaded
 NOISE           := $(DATA_DIR)/road_noise.parquet
 OFSTED          := $(DATA_DIR)/ofsted.parquet
 NAPTAN          := $(DATA_DIR)/naptan.parquet
@@ -65,7 +66,7 @@ PMTILES_VERSION := 1.22.3
 .PHONY: prepare merge tiles \
         download-arcgis download-price-paid download-deprivation download-ethnicity \
         download-naptan download-pois download-grocery-retail-points download-ofsted download-broadband download-rental-prices \
-        download-postcodes download-noise download-inspire \
+        download-postcodes download-noise download-inspire download-crime \
         download-oa-boundaries download-uprn-lookup download-transit-network download-greenspace download-os-greenspace download-pbf download-places download-lsoa-population download-median-age download-england-boundary download-rightmove-outcodes \
         transform-pois transform-epc-pp transform-crime transform-poi-proximity \
         transform-school-proximity transform-postcode-boundaries \
@@ -78,6 +79,7 @@ download-arcgis:              $(ARCGIS)
 download-price-paid:          $(PRICE_PAID)
 download-deprivation:         $(IOD)
 download-ethnicity:           $(ETHNICITY)
+download-crime:               $(CRIME_STAMP)
 download-naptan:              $(NAPTAN)
 download-pois:                $(POIS_RAW)
 download-grocery-retail-points: $(GROCERY_RETAIL_POINTS)
@@ -121,10 +123,10 @@ $(TILES):
 $(EPC):
 	@echo ""
 	@echo "=== EPC dataset not found ==="
-	@echo "The EPC certificates file is required: $@"
+	@echo "The EPC certificates archive is required: $@"
 	@echo ""
-	@echo "To obtain it, register at https://epc.opendatacommunities.org/login"
-	@echo "and place certificates.csv in manual-data/"
+	@echo "To obtain it, register at https://get-energy-performance-data.communities.gov.uk/filter-properties?property_type=domestic"
+	@echo "and place domestic-csv.zip in manual-data/"
 	@echo ""
 	@exit 1
 
@@ -140,6 +142,10 @@ $(IOD):
 $(ETHNICITY):
 	uv run python -m pipeline.download.ethnicity --output $@
 
+$(CRIME_STAMP):
+	uv run python -m pipeline.download.crime --output $(CRIME_DIR)
+	@touch $@
+
 $(NAPTAN):
 	uv run python -m pipeline.download.naptan --output $@
 
@@ -216,15 +222,7 @@ $(POIS_FILTERED): $(POIS_RAW) $(NAPTAN) $(GROCERY_RETAIL_POINTS) $(ENGLAND_BOUND
 $(EPC_PP): $(PRICE_PAID) $(EPC)
 	uv run python -m pipeline.transform.join_epc_pp --epc $(EPC) --price-paid $(PRICE_PAID) --output $@
 
-$(CRIME):
-	@if [ ! -d "$(CRIME_DIR)" ]; then \
-		echo ""; \
-		echo "=== Crime dataset not found ==="; \
-		echo "Place police.uk crime CSVs in $(CRIME_DIR)/"; \
-		echo "Download from https://data.police.uk/data/"; \
-		echo ""; \
-		exit 1; \
-	fi
+$(CRIME): $(CRIME_STAMP)
 	uv run python -m pipeline.transform.crime --input $(CRIME_DIR) --output $@
 
 $(POI_PROXIMITY): $(ARCGIS) $(POIS_FILTERED) $(OS_GREENSPACE)
diff --git a/manual-data/fixed_broadband_coverage.zip b/manual-data/fixed_broadband_coverage.zip
deleted file mode 100644
index b42cf58..0000000
Binary files a/manual-data/fixed_broadband_coverage.zip and /dev/null differ
diff --git a/manual-data/journey_times_bank.parquet b/manual-data/journey_times_bank.parquet
deleted file mode 100644
index fcdf3b6..0000000
Binary files a/manual-data/journey_times_bank.parquet and /dev/null differ
diff --git a/manual-data/journey_times_fitzrovia.parquet b/manual-data/journey_times_fitzrovia.parquet
deleted file mode 100644
index 825614f..0000000
Binary files a/manual-data/journey_times_fitzrovia.parquet and /dev/null differ
diff --git a/pipeline/download/crime.py b/pipeline/download/crime.py
new file mode 100644
index 0000000..adab8e5
--- /dev/null
+++ b/pipeline/download/crime.py
@@ -0,0 +1,393 @@
+"""Download police.uk crime archive ZIPs.
+
+The archive page lists rolling monthly snapshots. Newer snapshots overlap older
+ones, so extraction keeps files already written by newer archives.
+"""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import html
+import json
+import re
+import shutil
+import sys
+import zipfile
+from dataclasses import asdict, dataclass
+from datetime import UTC, datetime
+from pathlib import Path, PurePosixPath
+from urllib.parse import urljoin
+
+import httpx
+from tqdm import tqdm
+
+ARCHIVE_URL = "https://data.police.uk/data/archive/"
+ARCHIVE_LINK_RE = re.compile(
+    r'<div class="download">\s*.*?'
+    r'<a href="(?P<href>/data/archive/(?P<month>\d{4}-\d{2})\.zip)">'
+    r"(?P<label>[^<]+)</a>\s*\((?P<size>[^)]+)\)</span>\s*"
+    r'<p class="contained-range">\s*(?P<contained_range>.*?)\s*</p>\s*'
+    r'<p class="md5sum">(?P<md5>.*?)</p>',
+    re.DOTALL,
+)
+VALID_MD5_RE = re.compile(r"^[0-9a-fA-F]{32}$")
+MONTH_RE = re.compile(r"^\d{4}-\d{2}$")
+
+
+@dataclass(frozen=True)
+class CrimeArchive:
+    month: str
+    label: str
+    url: str
+    filename: str
+    size: str
+    contained_range: str
+    md5: str | None
+    raw_md5: str
+
+
+def _clean_text(value: str) -> str:
+    text = re.sub(r"<[^>]+>", " ", value)
+    return re.sub(r"\s+", " ", html.unescape(text)).strip()
+
+
+def parse_archives(page_html: str, base_url: str = ARCHIVE_URL) -> list[CrimeArchive]:
+    """Parse monthly crime archive links from the police.uk archive page."""
+    archives: list[CrimeArchive] = []
+    for match in ARCHIVE_LINK_RE.finditer(page_html):
+        raw_md5 = _clean_text(match.group("md5")).lower()
+        md5 = raw_md5 if VALID_MD5_RE.fullmatch(raw_md5) else None
+        href = html.unescape(match.group("href"))
+        archives.append(
+            CrimeArchive(
+                month=match.group("month"),
+                label=_clean_text(match.group("label")),
+                url=urljoin(base_url, href),
+                filename=Path(href).name,
+                size=_clean_text(match.group("size")),
+                contained_range=_clean_text(match.group("contained_range")),
+                md5=md5,
+                raw_md5=raw_md5,
+            )
+        )
+
+    return archives
+
+
+def fetch_archives(archive_url: str = ARCHIVE_URL) -> list[CrimeArchive]:
+    """Fetch and parse the archive index."""
+    with httpx.Client(
+        follow_redirects=True,
+        timeout=httpx.Timeout(30.0, read=60.0),
+        headers={"User-Agent": "perfect-postcode-data-pipeline/1.0"},
+    ) as client:
+        response = client.get(archive_url)
+        response.raise_for_status()
+
+    archives = parse_archives(response.text, archive_url)
+    if not archives:
+        raise RuntimeError(f"No monthly archive ZIPs found at {archive_url}")
+    return archives
+
+
+def filter_archives(
+    archives: list[CrimeArchive],
+    *,
+    from_month: str | None = None,
+    to_month: str | None = None,
+    limit: int | None = None,
+) -> list[CrimeArchive]:
+    """Filter archives by inclusive YYYY-MM bounds while preserving page order."""
+    filtered = [
+        archive
+        for archive in archives
+        if (from_month is None or archive.month >= from_month)
+        and (to_month is None or archive.month <= to_month)
+    ]
+    if limit is not None:
+        filtered = filtered[:limit]
+    return filtered
+
+
+def file_md5(path: Path) -> str:
+    digest = hashlib.md5()
+    with path.open("rb") as file:
+        for chunk in iter(lambda: file.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def download_archive(
+    archive: CrimeArchive,
+    archive_dir: Path,
+    *,
+    verify: bool,
+    force: bool,
+    timeout: float,
+) -> Path:
+    """Download one archive ZIP, resuming an existing .part file when possible."""
+    dest = archive_dir / archive.filename
+    partial = dest.with_suffix(dest.suffix + ".part")
+
+    if force:
+        dest.unlink(missing_ok=True)
+        partial.unlink(missing_ok=True)
+
+    if dest.exists():
+        if verify and archive.md5 is not None:
+            actual_md5 = file_md5(dest)
+            if actual_md5 == archive.md5:
+                print(f"{archive.filename}: already downloaded")
+                return dest
+            print(
+                f"{archive.filename}: checksum mismatch, downloading again",
+                file=sys.stderr,
+            )
+            dest.unlink()
+            partial.unlink(missing_ok=True)
+        else:
+            print(f"{archive.filename}: already downloaded")
+            return dest
+
+    resume_from = partial.stat().st_size if partial.exists() else 0
+    headers = {"Range": f"bytes={resume_from}-"} if resume_from else {}
+
+    with httpx.stream(
+        "GET",
+        archive.url,
+        headers=headers,
+        follow_redirects=True,
+        timeout=httpx.Timeout(30.0, read=timeout),
+    ) as response:
+        if response.status_code == 206 and resume_from:
+            mode = "ab"
+            initial = resume_from
+        else:
+            response.raise_for_status()
+            mode = "wb"
+            initial = 0
+
+        total_header = int(response.headers.get("content-length", 0))
+        total = initial + total_header if total_header else None
+        with (
+            partial.open(mode) as output,
+            tqdm(
+                total=total,
+                initial=initial,
+                unit="B",
+                unit_scale=True,
+                unit_divisor=1024,
+                desc=archive.filename,
+            ) as progress,
+        ):
+            for chunk in response.iter_bytes(chunk_size=1024 * 1024):
+                output.write(chunk)
+                progress.update(len(chunk))
+
+    partial.replace(dest)
+
+    if verify and archive.md5 is not None:
+        actual_md5 = file_md5(dest)
+        if actual_md5 != archive.md5:
+            dest.unlink(missing_ok=True)
+            raise RuntimeError(
+                f"{archive.filename}: MD5 mismatch: expected {archive.md5}, got {actual_md5}"
+            )
+
+    return dest
+
+
+def _safe_csv_members(
+    archive: zipfile.ZipFile,
+) -> list[tuple[zipfile.ZipInfo, PurePosixPath]]:
+    members: list[tuple[zipfile.ZipInfo, PurePosixPath]] = []
+    for info in archive.infolist():
+        rel_path = PurePosixPath(info.filename)
+        if (
+            info.is_dir()
+            or rel_path.is_absolute()
+            or ".." in rel_path.parts
+            or rel_path.suffix.lower() != ".csv"
+        ):
+            continue
+        members.append((info, rel_path))
+    return members
+
+
+def extract_csvs(
+    zip_path: Path,
+    output_dir: Path,
+    *,
+    overwrite: bool = False,
+) -> tuple[int, int]:
+    """Extract CSVs from one ZIP. Returns (extracted, skipped)."""
+    extracted = 0
+    skipped = 0
+
+    with zipfile.ZipFile(zip_path) as archive:
+        for info, rel_path in _safe_csv_members(archive):
+            dest = output_dir.joinpath(*rel_path.parts)
+            if dest.exists() and not overwrite:
+                skipped += 1
+                continue
+
+            dest.parent.mkdir(parents=True, exist_ok=True)
+            with archive.open(info) as source, dest.open("wb") as target:
+                shutil.copyfileobj(source, target)
+            extracted += 1
+
+    return extracted, skipped
+
+
+def write_manifest(
+    output_dir: Path, archive_url: str, archives: list[CrimeArchive]
+) -> None:
+    manifest = {
+        "source": archive_url,
+        "fetched_at": datetime.now(UTC).isoformat(),
+        "archives": [asdict(archive) for archive in archives],
+    }
+    path = output_dir / "archive_manifest.json"
+    path.write_text(json.dumps(manifest, indent=2) + "\n")
+
+
+def _month_arg(value: str) -> str:
+    if not MONTH_RE.fullmatch(value):
+        raise argparse.ArgumentTypeError("month must be in YYYY-MM format")
+    return value
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Download all monthly police.uk crime archive ZIPs"
+    )
+    parser.add_argument(
+        "--output",
+        type=Path,
+        required=True,
+        help="Directory for extracted CSVs; ZIPs are kept under _archives/",
+    )
+    parser.add_argument(
+        "--archive-url",
+        default=ARCHIVE_URL,
+        help=f"Archive index URL (default: {ARCHIVE_URL})",
+    )
+    parser.add_argument(
+        "--from-month",
+        type=_month_arg,
+        help="Only download archives from this YYYY-MM onwards",
+    )
+    parser.add_argument(
+        "--to-month",
+        type=_month_arg,
+        help="Only download archives up to this YYYY-MM",
+    )
+    parser.add_argument(
+        "--limit",
+        type=int,
+        help="Download at most this many archives after filtering",
+    )
+    parser.add_argument(
+        "--list",
+        action="store_true",
+        help="Print the archive URLs that would be downloaded and exit",
+    )
+    parser.add_argument(
+        "--no-extract",
+        dest="extract",
+        action="store_false",
+        help="Download ZIPs only; do not extract CSVs",
+    )
+    parser.add_argument(
+        "--overwrite-extracted",
+        action="store_true",
+        help="Overwrite CSVs when extracting overlapping archive snapshots",
+    )
+    parser.add_argument(
+        "--no-verify",
+        dest="verify",
+        action="store_false",
+        help="Skip MD5 verification",
+    )
+    parser.add_argument(
+        "--force",
+        action="store_true",
+        help="Redownload archives even if ZIP files already exist",
+    )
+    parser.add_argument(
+        "--timeout",
+        type=float,
+        default=600.0,
+        help="Per-read timeout in seconds for large ZIP downloads",
+    )
+    args = parser.parse_args()
+
+    print("Fetching police.uk archive index...")
+    archives = filter_archives(
+        fetch_archives(args.archive_url),
+        from_month=args.from_month,
+        to_month=args.to_month,
+        limit=args.limit,
+    )
+    if not archives:
+        raise SystemExit("No archives matched the requested filters")
+
+    bad_md5 = [
+        archive.filename for archive in archives if archive.raw_md5 and not archive.md5
+    ]
+    if bad_md5:
+        print(
+            "Warning: ignoring malformed MD5 values for "
+            + ", ".join(bad_md5[:5])
+            + ("..." if len(bad_md5) > 5 else ""),
+            file=sys.stderr,
+        )
+
+    print(f"Found {len(archives)} monthly archive ZIPs")
+    if args.list:
+        for archive in archives:
+            print(f"{archive.month}\t{archive.url}\t{archive.raw_md5}")
+        return
+
+    args.output.mkdir(parents=True, exist_ok=True)
+    archive_dir = args.output / "_archives"
+    archive_dir.mkdir(parents=True, exist_ok=True)
+    write_manifest(args.output, args.archive_url, archives)
+
+    total_extracted = 0
+    total_skipped = 0
+    for index, archive in enumerate(archives, start=1):
+        print(f"[{index}/{len(archives)}] {archive.label} ({archive.size})")
+        zip_path = download_archive(
+            archive,
+            archive_dir,
+            verify=args.verify,
+            force=args.force,
+            timeout=args.timeout,
+        )
+        if args.extract:
+            extracted, skipped = extract_csvs(
+                zip_path,
+                args.output,
+                overwrite=args.overwrite_extracted,
+            )
+            total_extracted += extracted
+            total_skipped += skipped
+            print(
+                f"{archive.filename}: extracted {extracted} CSVs"
+                + (f", skipped {skipped} existing CSVs" if skipped else "")
+            )
+
+    if args.extract:
+        print(
+            f"Done. ZIPs saved in {archive_dir}; extracted {total_extracted} CSVs"
+            + (f" and skipped {total_skipped} existing CSVs" if total_skipped else "")
+            + "."
+        )
+    else:
+        print(f"Done. ZIPs saved in {archive_dir}.")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/pipeline/download/map_assets.py b/pipeline/download/map_assets.py
index 1de7d84..9456177 100644
--- a/pipeline/download/map_assets.py
+++ b/pipeline/download/map_assets.py
@@ -1,9 +1,15 @@
 import argparse
+import base64
+import json
+import re
 import sys
 import urllib.request
 from concurrent.futures import ThreadPoolExecutor, as_completed
+from io import BytesIO
 from pathlib import Path
 
+from PIL import Image, ImageDraw
+
 from pipeline.transform.transform_poi import NAPTAN_EMOJIS, _CATEGORIES
 
 GLYPHS_BASE = "https://protomaps.github.io/basemaps-assets/fonts"
@@ -14,53 +20,80 @@ POI_ICON_BASE = "https://geolytix.github.io/MapIcons"
 # Font stacks used by @protomaps/basemaps with lang='en'
 FONT_STACKS = ["Noto Sans Regular", "Noto Sans Italic", "Noto Sans Medium"]
 
-# Fallback emoji not in any category
-_FALLBACK_EMOJIS = ["📍"]
-
 POI_ICON_PATHS = [
-    "asda/asda_express_24px.svg",
-    "asda/asda_green_basket_24px.svg",
-    "asda/asda_green_trolley_24px.svg",
-    "asda/asda_living_24px.svg",
-    "asda/asda_pfs_24px.svg",
-    "asda/asda_primary.svg",
-    "asda/asda_superstore_green_trolley_24px.svg",
-    "brands/aldi_24px.svg",
-    "brands/amazon_fresh_alt_24px.svg",
-    "brands/booths_24px.svg",
-    "brands/budgens_24px.svg",
-    "brands/centra_24px.svg",
-    "brands/cook.svg",
-    "brands/coop_24px.svg",
-    "brands/costco_24px.svg",
-    "brands/dunnes_stores_24px.svg",
-    "brands/farmfoods_updated_24px.svg",
-    "brands/heron_24px.svg",
-    "brands/iceland_24px.svg",
-    "brands/iceland_food_warehouse_24px.svg",
-    "brands/lidl_24px.svg",
-    "brands/little_waitrose_24px.svg",
-    "brands/makro_24px.svg",
-    "brands/mns_24px.svg",
-    "brands/mns_food_24px.svg",
-    "brands/mns_high_street_24px.svg",
-    "brands/mns_hospital_24px.svg",
-    "brands/mns_moto_24px.svg",
-    "brands/mns_outlet_24px.svg",
-    "brands/morrisons_24px.svg",
-    "brands/morrisons_daily_24px.svg",
-    "brands/sainsburys_24px.svg",
-    "brands/sainsburys_local_24px.svg",
-    "brands/spar_24px.svg",
-    "brands/tesco_24px.svg",
-    "brands/tesco_express_24px.svg",
-    "brands/tesco_extra_24px.svg",
-    "brands/waitrose_24px.svg",
-    "brands/wholefoods_24px.svg",
-    "logos/planet_organic_24px.svg",
+    "brands_2023/supermarkets/farmfoods.svg",
+    "brands_2023/supermarkets/heron_foods.svg",
+    "brands_2023/supermarkets/little_waitrose.svg",
+    "brands_2024/amazon_fresh.svg",
+    "brands_2024/booths.svg",
+    "brands_2024/budgens.svg",
+    "brands_2024/cook.svg",
+    "brands_2024/dunnes_stores.svg",
+    "brands_2024/iceland.svg",
+    "brands_2024/makro.svg",
+    "brands_2024/mns.svg",
+    "brands_2024/morrisons_daily.svg",
+    "brands_2024/sainsburys_local.svg",
+    "brands_2024/wholefoods.svg",
+    "logos/aldi.svg",
+    "logos/asda.svg",
+    "logos/centra.svg",
+    "logos/coop.svg",
+    "logos/lidl.svg",
+    "logos/morrisons.svg",
+    "logos/planet_organic.svg",
+    "logos/sainsburys.svg",
+    "logos/spar.svg",
+    "logos/tesco.svg",
+    "logos/tesco_express.svg",
+    "logos/tesco_extra.svg",
+    "logos/waitrose.svg",
     "public_transport/london_tube.svg",
+    "visuals/mns.svg",
 ]
 
+DERIVED_POI_ICON_PATHS = [
+    ("costco_logo", "brands/costco.svg", "logos/costco.svg"),
+    (
+        "embedded_png",
+        "brands/iceland_food_warehouse_24px.svg",
+        "logos/the_food_warehouse.png",
+    ),
+]
+
+POI_ICON_SVG_CROPS = {
+    "brands_2023/supermarkets/farmfoods.svg": (1.293, 7.314, 15.48, 3.293),
+    "brands_2023/supermarkets/heron_foods.svg": (0.062, 6.68, 17.995, 5.325),
+    "brands_2023/supermarkets/little_waitrose.svg": (0.916, 5.645, 16.365, 6.719),
+    "brands_2024/amazon_fresh.svg": (3.817, 1.646, 16.367, 16.358),
+    "brands_2024/booths.svg": (1.456, 7.143, 15.313, 3.512),
+    "brands_2024/budgens.svg": (2.251, 2.278, 13.6, 13.612),
+    "brands_2024/cook.svg": (5.028, 5.493, 13.945, 9.648),
+    "brands_2024/dunnes_stores.svg": (4.375, 7.732, 15.249, 5.055),
+    "brands_2024/iceland.svg": (1.136, 6.823, 16.067, 4.302),
+    "brands_2024/makro.svg": (4.411, 6.098, 16.397, 5.428),
+    "brands_2024/mns.svg": (4.042, 6.986, 16.171, 6.724),
+    "brands_2024/morrisons_daily.svg": (3.341, 4.414, 17.317, 8.248),
+    "brands_2024/sainsburys_local.svg": (4.58, 1.61, 14.84, 14.849),
+    "brands_2024/wholefoods.svg": (4.17, 2.193, 15.659, 15.668),
+    "logos/aldi.svg": (4.813, 2.563, 14.374, 14.383),
+    "logos/asda.svg": (3.91, 7.135, 16.181, 5.442),
+    "logos/centra.svg": (3.36, 7.35, 17.28, 4.651),
+    "logos/coop.svg": (6.407, 4.658, 11.187, 11.793),
+    "logos/costco.svg": (70.61, 144.908, 256.67, 85.825),
+    "logos/lidl.svg": (4.938, 2.973, 13.985, 13.985),
+    "logos/morrisons.svg": (5.231, 2.985, 13.538, 13.398),
+    "logos/planet_organic.svg": (5.528, 3.564, 12.943, 12.943),
+    "logos/sainsburys.svg": (7.502, 3.572, 8.996, 12.646),
+    "logos/spar.svg": (4.933, 2.968, 14.133, 13.853),
+    "logos/tesco.svg": (4.338, 6.865, 15.324, 5.359),
+    "logos/tesco_express.svg": (5.231, 5.933, 13.538, 8.345),
+    "logos/tesco_extra.svg": (4.933, 5.775, 14.133, 8.519),
+    "logos/waitrose.svg": (5.528, 6.09, 12.943, 9.855),
+}
+
+POI_ICON_SVG_INTRINSIC_MAX = 512
+
 
 def collect_twemoji_codes() -> list[str]:
     """Derive twemoji hex codes from transform_poi categories.
@@ -76,9 +109,6 @@ def collect_twemoji_codes() -> list[str]:
     for emoji in NAPTAN_EMOJIS.values():
         emojis.add(emoji)
 
-    for emoji in _FALLBACK_EMOJIS:
-        emojis.add(emoji)
-
     # First codepoint hex, matching frontend logic
     return sorted({f"{ord(e[0]):x}" for e in emojis})
 
@@ -97,6 +127,214 @@ def download_file(url: str, dest: Path) -> tuple[bool, str]:
         return False, url
 
 
+def download_text(url: str) -> str:
+    with urllib.request.urlopen(url) as response:
+        return response.read().decode("utf-8")
+
+
+def build_costco_logo(marker_svg: str) -> str:
+    start = marker_svg.find('<g><path d=" M 316.312')
+    end = marker_svg.rfind("</g></g></svg>")
+    if start < 0 or end < 0:
+        raise ValueError("Costco marker SVG layout changed")
+
+    logo_group = marker_svg[start : end + 4]
+    return (
+        '<?xml version="1.0" encoding="UTF-8"?>\n'
+        '<svg xmlns="http://www.w3.org/2000/svg" viewBox="70 145 260 90" '
+        'width="260pt" height="90pt" preserveAspectRatio="xMidYMid meet">\n'
+        f"{logo_group}\n"
+        "</svg>\n"
+    )
+
+
+def trim_white_png(png_bytes: bytes) -> bytes:
+    image = Image.open(BytesIO(png_bytes)).convert("RGBA")
+    pixels = image.load()
+
+    for y in range(image.height):
+        for x in range(image.width):
+            red, green, blue, alpha = pixels[x, y]
+            if red > 245 and green > 245 and blue > 245:
+                pixels[x, y] = (red, green, blue, 0)
+
+    alpha_box = image.getchannel("A").getbbox()
+    if alpha_box:
+        image = image.crop(alpha_box)
+
+    out = BytesIO()
+    image.save(out, format="PNG")
+    return out.getvalue()
+
+
+def extract_embedded_png(marker_svg: str) -> bytes:
+    match = re.search(r"base64,([^\"']+)", marker_svg)
+    if not match:
+        raise ValueError("POI marker SVG did not contain an embedded PNG")
+    return trim_white_png(base64.b64decode(match.group(1)))
+
+
+def svg_intrinsic_size(width: float, height: float) -> tuple[int, int]:
+    if width <= 0 or height <= 0:
+        return (POI_ICON_SVG_INTRINSIC_MAX, POI_ICON_SVG_INTRINSIC_MAX)
+    if width >= height:
+        return (
+            POI_ICON_SVG_INTRINSIC_MAX,
+            max(1, round(POI_ICON_SVG_INTRINSIC_MAX * height / width)),
+        )
+    return (
+        max(1, round(POI_ICON_SVG_INTRINSIC_MAX * width / height)),
+        POI_ICON_SVG_INTRINSIC_MAX,
+    )
+
+
+def set_svg_geometry(svg_text: str, crop: tuple[float, float, float, float]) -> str:
+    x, y, width, height = crop
+    view_box = f"{x:g} {y:g} {width:g} {height:g}"
+    intrinsic_width, intrinsic_height = svg_intrinsic_size(width, height)
+
+    svg_text = re.sub(r'viewBox="[^"]+"', f'viewBox="{view_box}"', svg_text, count=1)
+    if 'viewBox="' not in svg_text:
+        svg_text = re.sub(r"<svg\b", f'<svg viewBox="{view_box}"', svg_text, count=1)
+
+    svg_text = re.sub(r'width="[^"]+"', f'width="{intrinsic_width}"', svg_text, count=1)
+    if 'width="' not in svg_text:
+        svg_text = re.sub(
+            r"<svg\b", f'<svg width="{intrinsic_width}"', svg_text, count=1
+        )
+
+    svg_text = re.sub(
+        r'height="[^"]+"', f'height="{intrinsic_height}"', svg_text, count=1
+    )
+    if 'height="' not in svg_text:
+        svg_text = re.sub(
+            r"<svg\b", f'<svg height="{intrinsic_height}"', svg_text, count=1
+        )
+
+    return svg_text
+
+
+def get_svg_view_box(svg_text: str) -> tuple[float, float, float, float] | None:
+    match = re.search(r'viewBox="([^"]+)"', svg_text)
+    if not match:
+        return None
+    parts = [
+        float(part) for part in re.split(r"[\s,]+", match.group(1).strip()) if part
+    ]
+    if len(parts) != 4:
+        return None
+    return (parts[0], parts[1], parts[2], parts[3])
+
+
+def crop_poi_svg_icons(poi_icons_dir: Path) -> None:
+    for icon_path, crop in POI_ICON_SVG_CROPS.items():
+        dest = poi_icons_dir / icon_path
+        if not dest.exists():
+            continue
+        svg_text = dest.read_text(encoding="utf-8")
+        if icon_path == "brands_2024/dunnes_stores.svg":
+            svg_text = svg_text.replace('fill="#fffcfc"', 'fill="#111111"')
+            svg_text = svg_text.replace('fill="#fcfcfc"', 'fill="#111111"')
+        dest.write_text(set_svg_geometry(svg_text, crop), encoding="utf-8")
+
+    for dest in poi_icons_dir.rglob("*.svg"):
+        svg_text = dest.read_text(encoding="utf-8")
+        view_box = get_svg_view_box(svg_text)
+        if view_box:
+            dest.write_text(set_svg_geometry(svg_text, view_box), encoding="utf-8")
+
+
+def download_derived_poi_icon(
+    kind: str, source_path: str, dest: Path
+) -> tuple[bool, str]:
+    url = f"{POI_ICON_BASE}/{source_path}"
+    dest.parent.mkdir(parents=True, exist_ok=True)
+
+    try:
+        source = download_text(url)
+        if kind == "costco_logo":
+            dest.write_text(build_costco_logo(source), encoding="utf-8")
+        elif kind == "embedded_png":
+            dest.write_bytes(extract_embedded_png(source))
+        else:
+            raise ValueError(f"Unknown derived POI icon kind: {kind}")
+        return True, url
+    except urllib.error.HTTPError as e:
+        print(f"  {e.code} {url}", file=sys.stderr)
+        return False, url
+    except Exception as e:
+        print(f"  ERROR {url}: {e}", file=sys.stderr)
+        return False, url
+
+
+# Slategray accent used by civic POI icons (school, library, building, …) in
+# protomaps' v4 sprite. We match it so the townhall blends in with its peers.
+_TOWNHALL_COLOR = {
+    "light": (135, 128, 171),
+    "dark": (118, 118, 127),
+}
+_TOWNHALL_LOGICAL_SIZE = 17
+
+
+def _render_townhall_glyph(size_px: int, color: tuple[int, int, int]) -> Image.Image:
+    # Draw at 8× resolution and downsample with Lanczos so the pediment's
+    # diagonals come out anti-aliased; PIL's polygon fill is otherwise aliased.
+    super_factor = 8
+    canvas = size_px * super_factor
+    img = Image.new("RGBA", (canvas, canvas), (0, 0, 0, 0))
+    draw = ImageDraw.Draw(img)
+    fill = (*color, 255)
+
+    def s(v: float) -> float:
+        return v * canvas / _TOWNHALL_LOGICAL_SIZE
+
+    draw.polygon([(s(8.5), s(1)), (s(15), s(6.5)), (s(2), s(6.5))], fill=fill)
+    draw.rectangle([(s(1), s(6.5)), (s(16), s(8.5))], fill=fill)
+    for column_x in (3, 8, 13):
+        draw.rectangle([(s(column_x), s(8.5)), (s(column_x + 1.5), s(14))], fill=fill)
+    draw.rectangle([(s(0), s(14)), (s(17), s(15.5))], fill=fill)
+
+    return img.resize((size_px, size_px), Image.LANCZOS)
+
+
+def inject_townhall_sprite(sprites_dir: Path) -> None:
+    """Append a townhall glyph to each downloaded sprite sheet.
+
+    Protomaps' v4 sprite omits `townhall` even though the basemap style
+    references it; we add the icon here so MapLibre can resolve the name
+    natively at runtime.
+    """
+    for theme in ("light", "dark"):
+        color = _TOWNHALL_COLOR[theme]
+        for suffix, scale in (("", 1), ("@2x", 2)):
+            json_path = sprites_dir / f"{theme}{suffix}.json"
+            png_path = sprites_dir / f"{theme}{suffix}.png"
+            if not json_path.exists() or not png_path.exists():
+                continue
+
+            manifest = json.loads(json_path.read_text())
+            sheet = Image.open(png_path).convert("RGBA")
+
+            glyph_size = _TOWNHALL_LOGICAL_SIZE * scale
+            glyph = _render_townhall_glyph(glyph_size, color)
+
+            new_width = max(sheet.width, glyph_size)
+            new_height = sheet.height + glyph_size
+            extended = Image.new("RGBA", (new_width, new_height), (0, 0, 0, 0))
+            extended.paste(sheet, (0, 0))
+            extended.paste(glyph, (0, sheet.height))
+            extended.save(png_path, optimize=True)
+
+            manifest["townhall"] = {
+                "x": 0,
+                "y": sheet.height,
+                "width": glyph_size,
+                "height": glyph_size,
+                "pixelRatio": scale,
+            }
+            json_path.write_text(json.dumps(manifest))
+
+
 def main():
     parser = argparse.ArgumentParser(description=__doc__)
     parser.add_argument(
@@ -147,7 +385,7 @@ def main():
     # Skip already-downloaded files
     remaining = [(url, dest) for url, dest in tasks]
 
-    print(f"Downloading {len(remaining)} assets")
+    print(f"Downloading {len(remaining) + len(DERIVED_POI_ICON_PATHS)} assets")
 
     ok = 0
     fail = 0
@@ -162,6 +400,18 @@ def main():
             else:
                 fail += 1
 
+    for kind, source_path, dest_path in DERIVED_POI_ICON_PATHS:
+        success, _url = download_derived_poi_icon(
+            kind, source_path, poi_icons_dir / dest_path
+        )
+        if success:
+            ok += 1
+        else:
+            fail += 1
+
+    crop_poi_svg_icons(poi_icons_dir)
+    inject_townhall_sprite(sprites_dir)
+
     print(f"Done: {ok} downloaded, {fail} failed")
 
 
diff --git a/pipeline/download/noise.py b/pipeline/download/noise.py
index 0433f14..9a947ec 100644
--- a/pipeline/download/noise.py
+++ b/pipeline/download/noise.py
@@ -18,6 +18,7 @@ endpoint is broken for that coverage).
 
 import argparse
 import tempfile
+import time
 from concurrent.futures import ThreadPoolExecutor, as_completed
 from pathlib import Path
 
@@ -29,8 +30,10 @@ from pyproj import Transformer
 from rasterio.merge import merge
 from rasterio.transform import rowcol
 
-# Noise sources: (label, column_name, WCS base URL, coverage ID, WCS version)
-# Road/rail work with WCS 1.0.0; airport requires WCS 2.0.1.
+# Noise sources:
+# (label, column_name, WCS base URL, coverage ID, WCS version, allow_missing_tiles)
+# Road/rail work with WCS 1.0.0; airport requires WCS 2.0.1 and returns 500
+# for many sparse/no-coverage tiles, which should become nulls.
 NOISE_SOURCES = [
     (
         "Road",
@@ -38,6 +41,7 @@ NOISE_SOURCES = [
         "https://environment.data.gov.uk/spatialdata/road-noise-all-metrics-england-round-4/wcs",
         "Road_Noise_Lden_England_Round_4_All",
         "1.0.0",
+        False,
     ),
     (
         "Rail",
@@ -45,6 +49,7 @@ NOISE_SOURCES = [
         "https://environment.data.gov.uk/spatialdata/noise-data/wcs",
         "Rail_Noise_Lden_England_Round_4_All",
         "1.0.0",
+        False,
     ),
     (
         "Airport",
@@ -52,6 +57,7 @@ NOISE_SOURCES = [
         "https://environment.data.gov.uk/spatialdata/airport-noise-all-metrics-england-round-4/wcs",
         "dac9cba4-abe7-43bd-b8e9-8a83da52edd8__Airport_Noise_ALL_Lden",
         "2.0.1",
+        True,
     ),
 ]
 
@@ -74,6 +80,14 @@ NATIVE_RESOLUTION = 10
 # and keeps download size ~100x smaller than native 10m)
 RESOLUTION = 100
 
+# Retry/split behaviour for slow Defra WCS requests. Some 100km eastern tiles
+# intermittently return 504s; smaller fallback requests usually succeed.
+MAX_RETRIES = 3
+RETRY_BACKOFF_SECONDS = 5
+MIN_TILE_SIZE = 25_000
+
+type Tile = tuple[int, int, int, int]
+
 
 def _wcs_get_coverage_url(
     wcs_base: str,
@@ -117,6 +131,53 @@ def _bng_from_latlon(lat: np.ndarray, lon: np.ndarray) -> tuple[np.ndarray, np.n
     return _TO_BNG.transform(lon, lat)  # pyproj takes (x=lon, y=lat)
 
 
+def _looks_like_tiff(response: httpx.Response) -> bool:
+    content_type = response.headers.get("content-type", "")
+    return "tiff" in content_type or response.content[:4] in (b"II*\x00", b"MM\x00*")
+
+
+def _fetch_tile_bytes(
+    wcs_base: str,
+    coverage_id: str,
+    min_e: int,
+    min_n: int,
+    max_e: int,
+    max_n: int,
+    wcs_version: str = "1.0.0",
+) -> bytes | None:
+    """Fetch one WCS tile. Returns None when the server reports no GeoTIFF."""
+    url = _wcs_get_coverage_url(
+        wcs_base, coverage_id, min_e, min_n, max_e, max_n, wcs_version
+    )
+    with httpx.Client(timeout=300, follow_redirects=True) as client:
+        resp = client.get(url)
+        resp.raise_for_status()
+
+    if not _looks_like_tiff(resp):
+        return None
+    return resp.content
+
+
+def _split_tile(min_e: int, min_n: int, max_e: int, max_n: int) -> list[Tile]:
+    e_edges = [min_e, max_e]
+    n_edges = [min_n, max_n]
+    if max_e - min_e > MIN_TILE_SIZE:
+        e_edges.insert(1, (min_e + max_e) // 2)
+    if max_n - min_n > MIN_TILE_SIZE:
+        n_edges.insert(1, (min_n + max_n) // 2)
+
+    subtiles: list[Tile] = []
+    for e0, e1 in zip(e_edges, e_edges[1:]):
+        for n0, n1 in zip(n_edges, n_edges[1:]):
+            if e1 > e0 and n1 > n0:
+                subtiles.append((e0, n0, e1, n1))
+    return subtiles
+
+
+def _tile_path(tile_dir: Path, min_e: int, min_n: int, max_e: int, max_n: int) -> Path:
+    return tile_dir / f"tile_{min_e}_{min_n}_{max_e}_{max_n}.tif"
+
+
 def _download_tile(
     wcs_base: str,
     coverage_id: str,
@@ -124,30 +185,63 @@ def _download_tile(
     min_n: int,
     max_e: int,
     max_n: int,
-    tile_path: Path,
+    tile_dir: Path,
     wcs_version: str = "1.0.0",
-) -> Path | None:
-    """Download a single WCS tile. Returns path if successful, None otherwise."""
-    url = _wcs_get_coverage_url(
-        wcs_base, coverage_id, min_e, min_n, max_e, max_n, wcs_version
+    split_failures: bool = True,
+) -> tuple[list[Path], list[Tile]]:
+    """Download a WCS tile, splitting on repeated server failures."""
+    tile_path = _tile_path(tile_dir, min_e, min_n, max_e, max_n)
+    if tile_path.exists() and tile_path.stat().st_size > 0:
+        return [tile_path], []
+
+    last_error: Exception | None = None
+    for attempt in range(1, MAX_RETRIES + 1):
+        try:
+            content = _fetch_tile_bytes(
+                wcs_base, coverage_id, min_e, min_n, max_e, max_n, wcs_version
+            )
+            if content is None:
+                return [], []
+            tile_path.write_bytes(content)
+            return [tile_path], []
+        except (httpx.HTTPStatusError, httpx.TimeoutException, httpx.ConnectError) as e:
+            last_error = e
+            if attempt < MAX_RETRIES:
+                sleep_for = RETRY_BACKOFF_SECONDS * attempt
+                print(
+                    f"  Retrying tile ({min_e},{min_n})-({max_e},{max_n}) "
+                    f"after {type(e).__name__} ({attempt}/{MAX_RETRIES})"
+                )
+                time.sleep(sleep_for)
+
+    subtiles = _split_tile(min_e, min_n, max_e, max_n) if split_failures else []
+    if len(subtiles) > 1:
+        print(
+            f"  Splitting tile ({min_e},{min_n})-({max_e},{max_n}) "
+            f"into {len(subtiles)} smaller requests after: {last_error}"
+        )
+        paths: list[Path] = []
+        failures: list[Tile] = []
+        for e0, n0, e1, n1 in subtiles:
+            child_paths, child_failures = _download_tile(
+                wcs_base,
+                coverage_id,
+                e0,
+                n0,
+                e1,
+                n1,
+                tile_dir,
+                wcs_version,
+                split_failures,
+            )
+            paths.extend(child_paths)
+            failures.extend(child_failures)
+        return paths, failures
+
+    print(
+        f"  Failed to download tile ({min_e},{min_n})-({max_e},{max_n}): {last_error}"
     )
-    try:
-        with httpx.Client(timeout=300, follow_redirects=True) as client:
-            resp = client.get(url)
-            resp.raise_for_status()
-
-        content_type = resp.headers.get("content-type", "")
-        if "tiff" not in content_type and resp.content[:4] not in (
-            b"II*\x00",
-            b"MM\x00*",
-        ):
-            return None
-
-        tile_path.write_bytes(resp.content)
-        return tile_path
-    except (httpx.HTTPStatusError, httpx.TimeoutException, httpx.ConnectError) as e:
-        print(f"  Failed to download tile ({min_e},{min_n})-({max_e},{max_n}): {e}")
-        return None
+    return [], [(min_e, min_n, max_e, max_n)]
 
 
 def download_raster(
@@ -156,6 +250,7 @@ def download_raster(
     coverage_id: str,
     label: str,
     wcs_version: str = "1.0.0",
+    allow_missing_tiles: bool = False,
 ) -> list[Path]:
     """Download noise GeoTIFF raster covering England, returning paths to saved files."""
     tiles = []
@@ -168,13 +263,13 @@ def download_raster(
     print(
         f"[{label}] Downloading {len(tiles)} tiles at {RESOLUTION}m resolution ({MAX_WORKERS} workers)..."
     )
-    paths = []
+    paths: list[Path] = []
+    failures: list[Tile] = []
     completed = 0
 
     with ThreadPoolExecutor(max_workers=MAX_WORKERS) as executor:
         futures = {}
         for min_e, min_n, max_e, max_n in tiles:
-            tile_path = tile_dir / f"tile_{min_e}_{min_n}.tif"
             fut = executor.submit(
                 _download_tile,
                 wcs_base,
@@ -183,23 +278,41 @@ def download_raster(
                 min_n,
                 max_e,
                 max_n,
-                tile_path,
+                tile_dir,
                 wcs_version,
+                not allow_missing_tiles,
             )
             futures[fut] = (min_e, min_n)
 
         for fut in as_completed(futures):
             completed += 1
-            result = fut.result()
-            if result is not None:
-                paths.append(result)
+            result_paths, result_failures = fut.result()
+            paths.extend(result_paths)
+            failures.extend(result_failures)
             print(
-                f"\r  [{completed}/{len(tiles)}] Downloaded {len(paths)} valid tiles",
+                f"\r  [{completed}/{len(tiles)}] Downloaded {len(paths)} GeoTIFFs",
                 end="",
                 flush=True,
             )
 
-    print(f"\n[{label}] Downloaded {len(paths)}/{len(tiles)} tiles")
+    if failures:
+        preview = ", ".join(
+            f"({e0},{n0})-({e1},{n1})" for e0, n0, e1, n1 in failures[:5]
+        )
+        suffix = "..." if len(failures) > 5 else ""
+        if allow_missing_tiles:
+            print(
+                f"\n[{label}] WARNING: skipped {len(failures)} missing/no-data "
+                f"tile requests: {preview}{suffix}"
+            )
+            print(f"[{label}] Downloaded {len(paths)} GeoTIFFs from {len(tiles)} tiles")
+            return paths
+        raise RuntimeError(
+            f"[{label}] Failed to download {len(failures)} tile requests after "
+            f"retries and splitting: {preview}{suffix}"
+        )
+
+    print(f"\n[{label}] Downloaded {len(paths)} GeoTIFFs from {len(tiles)} tiles")
     return paths
 
 
@@ -281,11 +394,23 @@ def main() -> None:
     result = postcodes.select("postcode")
 
     with tempfile.TemporaryDirectory() as tmp:
-        for label, col_name, wcs_base, coverage_id, wcs_version in NOISE_SOURCES:
+        for (
+            label,
+            col_name,
+            wcs_base,
+            coverage_id,
+            wcs_version,
+            allow_missing_tiles,
+        ) in NOISE_SOURCES:
             tile_dir = Path(tmp) / label.lower()
             tile_dir.mkdir()
             tile_paths = download_raster(
-                tile_dir, wcs_base, coverage_id, label, wcs_version
+                tile_dir,
+                wcs_base,
+                coverage_id,
+                label,
+                wcs_version,
+                allow_missing_tiles,
             )
 
             if not tile_paths:
diff --git a/pipeline/download/places.py b/pipeline/download/places.py
index e2a24b8..369a3b9 100644
--- a/pipeline/download/places.py
+++ b/pipeline/download/places.py
@@ -6,6 +6,7 @@ Reuses the same england-latest.osm.pbf as pois.py.
 """
 
 import argparse
+import re
 from pathlib import Path
 
 import osmium
@@ -44,11 +45,37 @@ _STATION_STRIP = (
     " underground station",
     " railway station",
     " dlr station",
+    " station dlr",
+    " dlr",
     " overground station",
     " tram stop",
     " station",
 )
 
+_DLR_CODE_RE = re.compile(r"ZZDL([A-Z0-9]{3})")
+
+
+def _is_dlr_station(tags: dict[str, str]) -> bool:
+    name = tags.get("name", "").lower()
+    network = tags.get("network", "").lower()
+    operator = tags.get("operator", "").lower()
+    return (
+        "docklands" in network
+        or "dlr" in network
+        or "docklands" in operator
+        or "dlr" in operator
+        or name.endswith(" dlr")
+        or " dlr " in name
+    )
+
+
+def _is_tram_station(tags: dict[str, str]) -> bool:
+    if _is_dlr_station(tags):
+        return False
+    station_tag = tags.get("station", "")
+    network = tags.get("network", "").lower()
+    return station_tag == "light_rail" or "tramlink" in network or "tram" in network
+
 
 def _station_display_name(name: str, tags: dict[str, str]) -> str:
     """Build a descriptive station name like 'Bank tube station'."""
@@ -78,6 +105,96 @@ def _station_display_name(name: str, tags: dict[str, str]) -> str:
     return f"{name} {suffix}"
 
 
+def _station_name_score(name: str) -> tuple[int, int]:
+    lower = name.lower()
+    suffix_penalty = int(
+        lower.endswith(
+            (
+                " underground station",
+                " tube station",
+                " dlr station",
+                " railway station",
+                " rail station",
+                " station dlr",
+                " station",
+            )
+        )
+        or lower.endswith(" dlr")
+    )
+    return (suffix_penalty, len(name))
+
+
+def _naptan_dlr_stations(naptan_path: Path) -> list[dict]:
+    """Extract station-level DLR destinations from NaPTAN access nodes."""
+    df = pl.read_parquet(naptan_path)
+    required = {"id", "name", "category", "lat", "lng"}
+    missing = required - set(df.columns)
+    if missing:
+        raise ValueError(f"NaPTAN file is missing columns: {sorted(missing)}")
+
+    rows: dict[str, dict] = {}
+    for row in df.iter_rows(named=True):
+        atco_id = str(row["id"] or "")
+        match = _DLR_CODE_RE.search(atco_id)
+        if not match:
+            continue
+        if row["category"] not in {"Tube station", "Rail station"}:
+            continue
+
+        code = match.group(1)
+        raw_name = str(row["name"] or "")
+        if not raw_name:
+            continue
+
+        lat = float(row["lat"])
+        lon = float(row["lng"])
+        current = rows.get(code)
+        if current is None:
+            rows[code] = {
+                "raw_name": raw_name,
+                "lat_sum": lat,
+                "lon_sum": lon,
+                "count": 1,
+            }
+            continue
+
+        current["lat_sum"] += lat
+        current["lon_sum"] += lon
+        current["count"] += 1
+        if _station_name_score(raw_name) < _station_name_score(current["raw_name"]):
+            current["raw_name"] = raw_name
+
+    stations = []
+    for station in rows.values():
+        count = station["count"]
+        display_name = _station_display_name(station["raw_name"], {"network": "DLR"})
+        stations.append(
+            {
+                "name": display_name,
+                "place_type": "station",
+                "lat": station["lat_sum"] / count,
+                "lon": station["lon_sum"] / count,
+                "population": 0,
+                "travel_destination": True,
+            }
+        )
+
+    return sorted(stations, key=lambda station: station["name"])
+
+
+def _append_naptan_dlr_stations(places: list[dict], naptan_path: Path) -> int:
+    existing_names = {str(place["name"]).casefold() for place in places}
+    added = 0
+    for station in _naptan_dlr_stations(naptan_path):
+        key = station["name"].casefold()
+        if key in existing_names:
+            continue
+        places.append(station)
+        existing_names.add(key)
+        added += 1
+    return added
+
+
 class PlaceHandler(osmium.SimpleHandler):
     def __init__(self, progress: tqdm, england_polygon) -> None:
         super().__init__()
@@ -145,14 +262,7 @@ class PlaceHandler(osmium.SimpleHandler):
         # Railway stations (tube, national rail, DLR, overground, Elizabeth line)
         if n.tags.get("railway") == "station":
             tags = dict(n.tags)
-            station_tag = tags.get("station", "")
-            network = tags.get("network", "").lower()
-            # Skip tram stops
-            if (
-                station_tag == "light_rail"
-                or "tramlink" in network
-                or "tram" in network
-            ):
+            if _is_tram_station(tags):
                 return
             display_name = _station_display_name(name, tags)
             self._add(
@@ -178,6 +288,11 @@ def main() -> None:
         required=True,
         help="England boundary GeoJSON file",
     )
+    parser.add_argument(
+        "--naptan",
+        type=Path,
+        help="Optional NaPTAN parquet file used to add DLR station destinations",
+    )
     args = parser.parse_args()
 
     pbf_file = args.pbf
@@ -195,6 +310,9 @@ def main() -> None:
         handler.apply_file(str(pbf_file), locations=True)
 
     print(f"Extracted {len(handler.places):,} place nodes")
+    if args.naptan:
+        added = _append_naptan_dlr_stations(handler.places, args.naptan)
+        print(f"Added {added:,} DLR station destinations from NaPTAN")
 
     if handler.places:
         df = pl.DataFrame(handler.places)
diff --git a/pipeline/download/test_crime.py b/pipeline/download/test_crime.py
new file mode 100644
index 0000000..b986004
--- /dev/null
+++ b/pipeline/download/test_crime.py
@@ -0,0 +1,60 @@
+from zipfile import ZipFile
+
+from pipeline.download.crime import extract_csvs, parse_archives
+
+
+def test_parse_archives_reads_monthly_zip_links_only():
+    html = """
+    <p><a href="/data/archive/latest.zip">latest.zip</a></p>
+    <div class="archive crime">
+      <div class="download">
+        <i class="icon-file"></i> <span><a href="/data/archive/2026-03.zip">March 2026</a> (1.6&nbsp;GB)</span>
+        <p class="contained-range">Contains data from Apr 2023 to Mar 2026</p>
+        <p class="md5sum">6dde462489389445877f3988ef3f4f4b</p>
+      </div>
+      <div class="download">
+        <i class="icon-file"></i> <span><a href="/data/archive/2019-06.zip">June 2019</a> (1.6&nbsp;GB)</span>
+        <p class="contained-range">Contains data from Jul 2016 to Jun 2019</p>
+        <p class="md5sum">d6494297b24c1434bdb2504e95261bf8-100</p>
+      </div>
+    </div>
+    <div class="archive neighbourhood">
+      <div class="download">
+        <span><a href="/data/neighbourhood.zip">Neighbourhood crime</a> (2.2 MB)</span>
+        <small class="md5sum">6b80e2b97d87f6668b7a45953924d191</small>
+      </div>
+    </div>
+    """
+
+    archives = parse_archives(html, "https://data.police.uk/data/archive/")
+
+    assert [archive.filename for archive in archives] == [
+        "2026-03.zip",
+        "2019-06.zip",
+    ]
+    assert archives[0].url == "https://data.police.uk/data/archive/2026-03.zip"
+    assert archives[0].md5 == "6dde462489389445877f3988ef3f4f4b"
+    assert archives[1].md5 is None
+    assert archives[1].raw_md5 == "d6494297b24c1434bdb2504e95261bf8-100"
+
+
+def test_extract_csvs_preserves_existing_newer_files(tmp_path):
+    zip_path = tmp_path / "older.zip"
+    output = tmp_path / "crime"
+    existing = output / "2023-01" / "2023-01-city-street.csv"
+    existing.parent.mkdir(parents=True)
+    existing.write_text("newer\n")
+
+    with ZipFile(zip_path, "w") as archive:
+        archive.writestr("2023-01/2023-01-city-street.csv", "older\n")
+        archive.writestr("2022-12/2022-12-city-street.csv", "old\n")
+        archive.writestr("../escape.csv", "bad\n")
+        archive.writestr("notes.txt", "ignored\n")
+
+    extracted, skipped = extract_csvs(zip_path, output)
+
+    assert extracted == 1
+    assert skipped == 1
+    assert existing.read_text() == "newer\n"
+    assert (output / "2022-12" / "2022-12-city-street.csv").read_text() == "old\n"
+    assert not (tmp_path / "escape.csv").exists()
diff --git a/pipeline/download/test_noise.py b/pipeline/download/test_noise.py
new file mode 100644
index 0000000..59b2695
--- /dev/null
+++ b/pipeline/download/test_noise.py
@@ -0,0 +1,89 @@
+import httpx
+import pytest
+
+from pipeline.download import noise
+
+
+def test_download_tile_splits_after_retries(monkeypatch, tmp_path):
+    monkeypatch.setattr(noise, "MAX_RETRIES", 1)
+    monkeypatch.setattr(noise, "MIN_TILE_SIZE", 50)
+
+    def fake_fetch_tile_bytes(
+        wcs_base,
+        coverage_id,
+        min_e,
+        min_n,
+        max_e,
+        max_n,
+        wcs_version="1.0.0",
+    ):
+        if max_e - min_e > 50 or max_n - min_n > 50:
+            raise httpx.TimeoutException("too large")
+        return b"II*\x00fake-tiff"
+
+    monkeypatch.setattr(noise, "_fetch_tile_bytes", fake_fetch_tile_bytes)
+
+    paths, failures = noise._download_tile("base", "coverage", 0, 0, 100, 100, tmp_path)
+
+    assert failures == []
+    assert len(paths) == 4
+    assert sorted(path.name for path in paths) == [
+        "tile_0_0_50_50.tif",
+        "tile_0_50_50_100.tif",
+        "tile_50_0_100_50.tif",
+        "tile_50_50_100_100.tif",
+    ]
+
+
+def test_download_tile_reports_unsplittable_failure(monkeypatch, tmp_path):
+    monkeypatch.setattr(noise, "MAX_RETRIES", 1)
+    monkeypatch.setattr(noise, "MIN_TILE_SIZE", 100)
+
+    def fake_fetch_tile_bytes(*args, **kwargs):
+        raise httpx.ConnectError("offline")
+
+    monkeypatch.setattr(noise, "_fetch_tile_bytes", fake_fetch_tile_bytes)
+
+    paths, failures = noise._download_tile("base", "coverage", 0, 0, 100, 100, tmp_path)
+
+    assert paths == []
+    assert failures == [(0, 0, 100, 100)]
+
+
+def test_download_raster_tolerates_missing_tiles_when_allowed(monkeypatch, tmp_path):
+    monkeypatch.setattr(noise, "BNG_MIN_E", 0)
+    monkeypatch.setattr(noise, "BNG_MAX_E", 100)
+    monkeypatch.setattr(noise, "BNG_MIN_N", 0)
+    monkeypatch.setattr(noise, "BNG_MAX_N", 100)
+    monkeypatch.setattr(noise, "TILE_SIZE", 100)
+
+    def fake_download_tile(*args, **kwargs):
+        return [], [(0, 0, 100, 100)]
+
+    monkeypatch.setattr(noise, "_download_tile", fake_download_tile)
+
+    paths = noise.download_raster(
+        tmp_path,
+        "base",
+        "coverage",
+        "Airport",
+        allow_missing_tiles=True,
+    )
+
+    assert paths == []
+
+
+def test_download_raster_raises_on_missing_strict_tiles(monkeypatch, tmp_path):
+    monkeypatch.setattr(noise, "BNG_MIN_E", 0)
+    monkeypatch.setattr(noise, "BNG_MAX_E", 100)
+    monkeypatch.setattr(noise, "BNG_MIN_N", 0)
+    monkeypatch.setattr(noise, "BNG_MAX_N", 100)
+    monkeypatch.setattr(noise, "TILE_SIZE", 100)
+
+    def fake_download_tile(*args, **kwargs):
+        return [], [(0, 0, 100, 100)]
+
+    monkeypatch.setattr(noise, "_download_tile", fake_download_tile)
+
+    with pytest.raises(RuntimeError, match=r"\[Road\] Failed to download"):
+        noise.download_raster(tmp_path, "base", "coverage", "Road")
diff --git a/pipeline/download/test_places.py b/pipeline/download/test_places.py
new file mode 100644
index 0000000..978fa30
--- /dev/null
+++ b/pipeline/download/test_places.py
@@ -0,0 +1,81 @@
+import polars as pl
+
+from pipeline.download.places import (
+    _is_dlr_station,
+    _is_tram_station,
+    _naptan_dlr_stations,
+    _station_display_name,
+)
+
+
+def test_dlr_light_rail_is_not_treated_as_tram():
+    dlr_tags = {
+        "name": "Lewisham DLR",
+        "railway": "station",
+        "station": "light_rail",
+        "network": "Docklands Light Railway",
+    }
+
+    assert _is_dlr_station(dlr_tags)
+    assert not _is_tram_station(dlr_tags)
+    assert _station_display_name("Lewisham DLR", dlr_tags) == "Lewisham DLR station"
+    assert (
+        _station_display_name("Tower Gateway Station DLR", dlr_tags)
+        == "Tower Gateway DLR station"
+    )
+
+
+def test_tram_light_rail_is_still_excluded():
+    tram_tags = {
+        "name": "East Croydon",
+        "railway": "station",
+        "station": "light_rail",
+        "network": "London Trams",
+    }
+
+    assert not _is_dlr_station(tram_tags)
+    assert _is_tram_station(tram_tags)
+
+
+def test_naptan_dlr_stations_are_deduplicated_by_atco_code(tmp_path):
+    naptan = tmp_path / "naptan.parquet"
+    pl.DataFrame(
+        {
+            "id": [
+                "4900ZZDLSHA3",
+                "9400ZZDLSHA",
+                "4900ZZDLGRE1",
+                "490002076RV",
+                "4900ZZLUBNK",
+            ],
+            "name": [
+                "Shadwell DLR",
+                "Shadwell DLR Station",
+                "Greenwich Station",
+                "Tower Gateway Station DLR",
+                "Bank",
+            ],
+            "category": [
+                "Tube station",
+                "Tube station",
+                "Rail station",
+                "Bus stop",
+                "Tube station",
+            ],
+            "lat": [51.51156, 51.511693, 51.47794, 51.510575, 51.5131],
+            "lng": [-0.055595, -0.056643, -0.01442, -0.07514, -0.0894],
+        }
+    ).write_parquet(naptan)
+
+    stations = _naptan_dlr_stations(naptan)
+
+    assert [station["name"] for station in stations] == [
+        "Greenwich DLR station",
+        "Shadwell DLR station",
+    ]
+    shadwell = next(
+        station for station in stations if station["name"].startswith("Shadwell")
+    )
+    assert shadwell["lat"] == (51.51156 + 51.511693) / 2
+    assert shadwell["place_type"] == "station"
+    assert shadwell["travel_destination"] is True
diff --git a/pipeline/download/transit_network.py b/pipeline/download/transit_network.py
index 94dd5f5..4cbf5f8 100644
--- a/pipeline/download/transit_network.py
+++ b/pipeline/download/transit_network.py
@@ -56,6 +56,7 @@ NR_AUTH_URL = "https://opendata.nationalrail.co.uk/authenticate"
 NR_TIMETABLE_URL = "https://opendata.nationalrail.co.uk/api/staticfeeds/3.0/timetable"
 
 USER_AGENT = "property-map-pipeline/1.0 (https://github.com)"
+TRANSXCHANGE2GTFS_PACKAGE = "transxchange2gtfs@1.12.0"
 
 
 def _download_http(
@@ -473,10 +474,50 @@ def convert_tfl_to_gtfs(raw_dir: Path, output_dir: Path) -> Path:
     download_naptan()
 
     print("Converting TfL TransXChange → GTFS...")
+    # The shim patches known packaging/runtime issues in the pinned npm package
+    # before loading its CLI from npx's temporary install.
+    shim_path = Path(__file__).with_name("transxchange2gtfs_shim.js")
     subprocess.run(
-        ["npx", "--yes", "transxchange2gtfs", str(txc_path), str(dest)],
+        [
+            "npx",
+            "--yes",
+            "--package",
+            TRANSXCHANGE2GTFS_PACKAGE,
+            "sh",
+            "-c",
+            "\n".join(
+                [
+                    'bin="$(command -v transxchange2gtfs)"',
+                    'script="$(readlink -f "$bin")"',
+                    'pkg_dir="$(dirname "$(dirname "$script")")"',
+                    'shim="$1"',
+                    "shift",
+                    'exec node "$shim" "$pkg_dir" "$@"',
+                ]
+            ),
+            "transxchange2gtfs",
+            str(shim_path.resolve()),
+            str(txc_path.resolve()),
+            str(dest.resolve()),
+        ],
         check=True,
     )
+    required_files = {
+        "agency.txt",
+        "calendar.txt",
+        "calendar_dates.txt",
+        "routes.txt",
+        "stop_times.txt",
+        "stops.txt",
+        "trips.txt",
+    }
+    if not dest.exists() or not zipfile.is_zipfile(dest):
+        raise RuntimeError(f"transxchange2gtfs did not create a valid GTFS zip: {dest}")
+    with zipfile.ZipFile(dest) as z:
+        missing = required_files - set(z.namelist())
+    if missing:
+        missing_str = ", ".join(sorted(missing))
+        raise RuntimeError(f"TfL GTFS zip is missing required files: {missing_str}")
     size_mb = dest.stat().st_size / (1024 * 1024)
     print(f"  Saved to {dest} ({size_mb:.1f} MB)")
     return dest
diff --git a/pipeline/download/transxchange2gtfs_shim.js b/pipeline/download/transxchange2gtfs_shim.js
new file mode 100644
index 0000000..e1952e8
--- /dev/null
+++ b/pipeline/download/transxchange2gtfs_shim.js
@@ -0,0 +1,76 @@
+#!/usr/bin/env node
+"use strict";
+
+const fs = require("fs");
+const path = require("path");
+const { createRequire } = require("module");
+
+const [pkgDirArg, ...converterArgs] = process.argv.slice(2);
+
+if (!pkgDirArg || converterArgs.length < 2) {
+  console.error(
+    "Usage: transxchange2gtfs_shim.js <package-dir> <input...> <output>",
+  );
+  process.exit(2);
+}
+
+const pkgDir = path.resolve(pkgDirArg);
+
+function replaceOnce(relativePath, before, after) {
+  const file = path.join(pkgDir, relativePath);
+  const original = fs.readFileSync(file, "utf8");
+  if (original.includes(before)) {
+    fs.writeFileSync(file, original.replace(before, after));
+  } else if (original.includes(after)) {
+    return;
+  } else {
+    throw new Error(`Could not patch ${relativePath}: expected text not found`);
+  }
+}
+
+// The published 1.12.0 package has a few compatibility issues with current
+// TfL TransXChange exports:
+// - the bin script points at dist/src/cli.js, but the package ships dist/cli.js
+// - the compiled date-holidays import expects a synthetic default export
+// - some TfL journeys reference timing links without matching route-link geometry
+//
+// GTFS shapes are optional for R5 routing. Clear shape references and omit
+// shapes.txt so missing route geometry does not drop otherwise usable trips.
+function patchPackage() {
+  replaceOnce(
+    "dist/transxchange/TransXChangeJourneyStream.js",
+    "distanceSoFarM += routeLink.Distance;",
+    "distanceSoFarM += routeLink ? routeLink.Distance : 0;",
+  );
+  replaceOnce(
+    "dist/gtfs/TripsStream.js",
+    "(0, crypto_1.createHash)('md5').update(JSON.stringify({ routeId: journey.route, routeLinkSeq: journey.routeLinkIds })).digest(\"hex\"));",
+    "\"\");",
+  );
+  replaceOnce(
+    "dist/gtfs/StopTimesStream.js",
+    "stop.shapeDistTraveled, stop.exactTime ? \"1\" : \"0\");",
+    "\"\", stop.exactTime ? \"1\" : \"0\");",
+  );
+  replaceOnce(
+    "dist/Container.js",
+    "\"stops.txt\": transxchange.pipe(new StopsStream_1.StopsStream(naptanIndex)),\n            \"shapes.txt\": journeyStream.pipe(new ShapesStream_1.ShapesStream())",
+    "\"stops.txt\": transxchange.pipe(new StopsStream_1.StopsStream(naptanIndex))",
+  );
+  replaceOnce(
+    "dist/Container.js",
+    "\"routes.txt\": transxchange.pipe(new RoutesStream_1.RoutesStream()),\n            \"transfers.txt\": transxchange.pipe(new TransfersStream_1.TransfersStream(naptanIndex, locationIndex)),\n            \"stops.txt\": transxchange.pipe(new StopsStream_1.StopsStream(naptanIndex))",
+    "\"routes.txt\": transxchange.pipe(new RoutesStream_1.RoutesStream()),\n            \"stops.txt\": transxchange.pipe(new StopsStream_1.StopsStream(naptanIndex))",
+  );
+}
+
+patchPackage();
+
+const pkgRequire = createRequire(path.join(pkgDir, "package.json"));
+const Holidays = pkgRequire("date-holidays");
+if (!Holidays.default) {
+  Holidays.default = Holidays;
+}
+
+process.argv = [process.argv[0], "transxchange2gtfs", ...converterArgs];
+require(path.join(pkgDir, "dist", "cli.js"));
diff --git a/pipeline/transform/crime.py b/pipeline/transform/crime.py
index acd4010..a9cf808 100644
--- a/pipeline/transform/crime.py
+++ b/pipeline/transform/crime.py
@@ -1,12 +1,32 @@
 import argparse
+import re
 from pathlib import Path
 
 import polars as pl
 
+STREET_CRIME_CSV_RE = re.compile(r"^\d{4}-\d{2}-.+-street\.csv$")
+
+
+def find_street_crime_csvs(crime_dir: Path) -> tuple[list[Path], int]:
+    csvs = sorted(crime_dir.rglob("*.csv"))
+    street_csvs = [path for path in csvs if STREET_CRIME_CSV_RE.fullmatch(path.name)]
+    return street_csvs, len(csvs) - len(street_csvs)
+
 
 def transform_crime(crime_dir: Path, output_path: Path) -> None:
-    csvs = sorted(crime_dir.rglob("*.csv"))
-    print(f"Found {len(csvs)} CSV files across {len(list(crime_dir.iterdir()))} months")
+    csvs, ignored_csv_count = find_street_crime_csvs(crime_dir)
+    if not csvs:
+        raise FileNotFoundError(f"No street crime CSV files found in {crime_dir}")
+
+    month_count = len({path.parent.name for path in csvs})
+    print(
+        f"Found {len(csvs)} street crime CSV files across {month_count} months"
+        + (
+            f" (ignored {ignored_csv_count} non-street CSVs)"
+            if ignored_csv_count
+            else ""
+        )
+    )
 
     df = pl.scan_csv(
         csvs,
diff --git a/pipeline/transform/join_epc_pp.py b/pipeline/transform/join_epc_pp.py
index 2f90389..95c9769 100644
--- a/pipeline/transform/join_epc_pp.py
+++ b/pipeline/transform/join_epc_pp.py
@@ -1,6 +1,15 @@
 import argparse
-import polars as pl
+import csv
+import io
+import tempfile
+import zipfile
 from pathlib import Path
+
+import polars as pl
+import pyarrow as pa
+import pyarrow.csv as pa_csv
+import pyarrow.parquet as pq
+
 from ..utils import fuzzy_join_on_postcode
 
 
@@ -8,12 +17,168 @@ pl.Config.set_tbl_cols(-1)
 
 RATING_RANK = {"A": 1, "B": 2, "C": 3, "D": 4, "E": 5, "F": 6, "G": 7}
 MIN_PRICE = 50_000
+EPC_SOURCE_COLUMNS = [
+    "address",
+    "postcode",
+    "current_energy_rating",
+    "potential_energy_rating",
+    "property_type",
+    "built_form",
+    "inspection_date",
+    "total_floor_area",
+    "number_habitable_rooms",
+    "floor_height",
+    "construction_age_band",
+    "tenure",
+]
+
+
+def _normalise_csv_columns(columns: list[str]) -> list[str]:
+    return [column.strip().lower() for column in columns]
+
+
+def _clean_string(column: str) -> pl.Expr:
+    stripped = pl.col(column).cast(pl.String).str.strip_chars()
+    return pl.when(stripped == "").then(None).otherwise(stripped)
+
+
+def _clean_number(column: str, dtype: pl.DataType) -> pl.Expr:
+    return _clean_string(column).cast(dtype, strict=False)
+
+
+def _select_epc_columns(raw: pl.LazyFrame) -> pl.LazyFrame:
+    return (
+        raw.select(
+            _clean_string("address").alias("epc_address"),
+            _clean_string("postcode").str.to_uppercase().alias("epc_postcode"),
+            _clean_string("current_energy_rating")
+            .str.to_uppercase()
+            .alias("current_energy_rating"),
+            _clean_string("potential_energy_rating")
+            .str.to_uppercase()
+            .alias("potential_energy_rating"),
+            _clean_string("property_type").alias("epc_property_type"),
+            _clean_string("built_form").alias("built_form"),
+            _clean_string("inspection_date").alias("inspection_date"),
+            _clean_number("total_floor_area", pl.Float64).alias("total_floor_area"),
+            _clean_number("number_habitable_rooms", pl.Int16).alias(
+                "number_habitable_rooms"
+            ),
+            _clean_number("floor_height", pl.Float64).alias("floor_height"),
+            _clean_string("construction_age_band").alias("construction_age_band"),
+            _clean_string("tenure").alias("tenure"),
+        )
+        .filter(pl.col("epc_address").is_not_null())
+        .with_columns(
+            pl.when(pl.col("number_habitable_rooms") == 0)
+            .then(None)
+            .otherwise(pl.col("number_habitable_rooms"))
+            .alias("number_habitable_rooms"),
+        )
+    )
+
+
+def _certificate_member_names(zip_file: zipfile.ZipFile) -> list[str]:
+    return sorted(
+        name
+        for name in zip_file.namelist()
+        if not name.endswith("/")
+        and Path(name).name.lower().startswith("certificates")
+        and name.lower().endswith(".csv")
+    )
+
+
+def _read_zip_csv_header(zip_file: zipfile.ZipFile, member_name: str) -> list[str]:
+    with zip_file.open(member_name) as member:
+        text = io.TextIOWrapper(member, encoding="utf-8-sig", newline="")
+        try:
+            return next(csv.reader(text))
+        except StopIteration as exc:
+            raise ValueError(f"EPC CSV member is empty: {member_name}") from exc
+
+
+def _source_columns_for_header(header: list[str]) -> list[str]:
+    columns_by_normalised_name = {
+        normalised: source
+        for source, normalised in zip(header, _normalise_csv_columns(header))
+    }
+    return [
+        columns_by_normalised_name.get(column, column) for column in EPC_SOURCE_COLUMNS
+    ]
+
+
+def _zip_certificates_to_parquet(zip_path: Path, output_path: Path) -> None:
+    schema = pa.schema((column, pa.string()) for column in EPC_SOURCE_COLUMNS)
+    writer = pq.ParquetWriter(output_path, schema=schema, compression="zstd")
+
+    try:
+        try:
+            zip_file = zipfile.ZipFile(zip_path)
+        except zipfile.BadZipFile as exc:
+            raise ValueError(
+                f"{zip_path} is not a readable EPC zip archive; re-download "
+                "domestic-csv.zip and try again"
+            ) from exc
+
+        with zip_file:
+            member_names = _certificate_member_names(zip_file)
+            if not member_names:
+                raise ValueError(f"No certificate CSV files found in {zip_path}")
+
+            for member_name in member_names:
+                print(f"Reading EPC certificates from {member_name}")
+                source_columns = _source_columns_for_header(
+                    _read_zip_csv_header(zip_file, member_name)
+                )
+                convert_options = pa_csv.ConvertOptions(
+                    include_columns=source_columns,
+                    include_missing_columns=True,
+                    column_types={
+                        source_column: pa.string() for source_column in source_columns
+                    },
+                    strings_can_be_null=True,
+                )
+                read_options = pa_csv.ReadOptions(block_size=64 * 1024 * 1024)
+
+                with zip_file.open(member_name) as member:
+                    reader = pa_csv.open_csv(
+                        member,
+                        read_options=read_options,
+                        convert_options=convert_options,
+                    )
+                    while True:
+                        try:
+                            batch = reader.read_next_batch()
+                        except StopIteration:
+                            break
+
+                        if batch.num_rows == 0:
+                            continue
+
+                        writer.write_batch(batch.rename_columns(EPC_SOURCE_COLUMNS))
+    finally:
+        writer.close()
+
+
+def _scan_epc_certificates(epc_path: Path, temp_dir: Path) -> pl.LazyFrame:
+    if epc_path.suffix.lower() == ".zip":
+        parquet_path = temp_dir / "epc-certificates.parquet"
+        _zip_certificates_to_parquet(epc_path, parquet_path)
+        raw = pl.scan_parquet(parquet_path)
+    else:
+        raw = pl.scan_csv(
+            epc_path,
+            infer_schema=False,
+            with_column_names=_normalise_csv_columns,
+        )
+
+    return _select_epc_columns(raw)
 
 
 def main():
     parser = argparse.ArgumentParser(description="Fuzzy join EPC and Price Paid data")
     parser.add_argument(
-        "--epc", type=Path, required=True, help="EPC certificates CSV file"
+        "--epc", type=Path, required=True, help="EPC certificates CSV file or zip"
     )
     parser.add_argument(
         "--price-paid", type=Path, required=True, help="Price paid parquet file"
@@ -23,74 +188,56 @@ def main():
     )
     args = parser.parse_args()
 
-    epc_base = (
-        pl.scan_csv(args.epc)
-        .select(
-            pl.col("ADDRESS").alias("epc_address"),
-            "POSTCODE",
-            "CURRENT_ENERGY_RATING",
-            "POTENTIAL_ENERGY_RATING",
-            pl.col("PROPERTY_TYPE").alias("epc_property_type"),
-            "BUILT_FORM",
-            "INSPECTION_DATE",
-            "TOTAL_FLOOR_AREA",
-            "NUMBER_HABITABLE_ROOMS",
-            "FLOOR_HEIGHT",
-            "CONSTRUCTION_AGE_BAND",
-            "TENURE",
-        )
-        .filter(pl.col("epc_address").is_not_null())
-        .with_columns(
-            pl.when(pl.col("NUMBER_HABITABLE_ROOMS") == 0)
-            .then(None)
-            .otherwise(pl.col("NUMBER_HABITABLE_ROOMS"))
-            .alias("NUMBER_HABITABLE_ROOMS"),
-        )
-    )
+    with tempfile.TemporaryDirectory(prefix="epc_certificates_") as tmpdir:
+        _run(args.epc, args.price_paid, args.output, Path(tmpdir))
+
+
+def _run(epc_path: Path, price_paid_path: Path, output_path: Path, temp_dir: Path):
+    epc_base = _scan_epc_certificates(epc_path, temp_dir)
 
     # Dedup fork: keep latest certificate per property (existing logic)
     epc = (
-        epc_base.sort("INSPECTION_DATE", descending=True)
-        .group_by("epc_address", "POSTCODE")
+        epc_base.sort("inspection_date", descending=True)
+        .group_by("epc_address", "epc_postcode")
         .first()
-        .drop("TENURE")
+        .drop("tenure")
     )
 
     # Events fork: detect renovation events between consecutive certificates
     # Collect eagerly because .over() window functions don't work in streaming
     # engine (fuzzy_join.py:50 uses sink_parquet which requires streaming).
     events = (
-        epc_base.sort("INSPECTION_DATE")
+        epc_base.sort("inspection_date")
         .with_columns(
-            pl.col("CURRENT_ENERGY_RATING")
+            pl.col("current_energy_rating")
             .replace_strict(RATING_RANK, default=None, return_dtype=pl.Int32)
             .alias("_rating_rank"),
         )
         .with_columns(
-            pl.col("NUMBER_HABITABLE_ROOMS")
+            pl.col("number_habitable_rooms")
             .shift(1)
-            .over("epc_address", "POSTCODE")
+            .over("epc_address", "epc_postcode")
             .alias("_prev_rooms"),
-            pl.col("TOTAL_FLOOR_AREA")
+            pl.col("total_floor_area")
             .shift(1)
-            .over("epc_address", "POSTCODE")
+            .over("epc_address", "epc_postcode")
             .alias("_prev_area"),
             pl.col("_rating_rank")
             .shift(1)
-            .over("epc_address", "POSTCODE")
+            .over("epc_address", "epc_postcode")
             .alias("_prev_rating_rank"),
         )
         .with_columns(
             pl.when(
-                pl.col("NUMBER_HABITABLE_ROOMS").is_not_null()
+                pl.col("number_habitable_rooms").is_not_null()
                 & pl.col("_prev_rooms").is_not_null()
-                & (pl.col("NUMBER_HABITABLE_ROOMS") != pl.col("_prev_rooms"))
+                & (pl.col("number_habitable_rooms") != pl.col("_prev_rooms"))
             )
             .then(pl.lit("Remodelling"))
             .when(
-                pl.col("TOTAL_FLOOR_AREA").is_not_null()
+                pl.col("total_floor_area").is_not_null()
                 & pl.col("_prev_area").is_not_null()
-                & (pl.col("TOTAL_FLOOR_AREA") > pl.col("_prev_area"))
+                & (pl.col("total_floor_area") > pl.col("_prev_area"))
             )
             .then(pl.lit("Extension"))
             .when(
@@ -104,13 +251,13 @@ def main():
         )
         .filter(pl.col("_event").is_not_null())
         .with_columns(
-            pl.col("INSPECTION_DATE")
+            pl.col("inspection_date")
             .cast(pl.String)
             .str.slice(0, 4)
             .cast(pl.Int32)
             .alias("_event_year"),
         )
-        .group_by("epc_address", "POSTCODE")
+        .group_by("epc_address", "epc_postcode")
         .agg(
             pl.struct(
                 pl.col("_event_year").alias("year"),
@@ -128,8 +275,8 @@ def main():
 
     # Social tenure fork: flag properties that were ever social housing
     social_tenure = (
-        epc_base.filter(pl.col("TENURE").str.to_lowercase().str.contains("social"))
-        .select("epc_address", "POSTCODE")
+        epc_base.filter(pl.col("tenure").str.to_lowercase().str.contains("social"))
+        .select("epc_address", "epc_postcode")
         .unique()
         .with_columns(pl.lit("Yes").alias("was_council_house"))
         .collect()
@@ -140,12 +287,12 @@ def main():
     epc = (
         epc.join(
             events.lazy(),
-            on=["epc_address", "POSTCODE"],
+            on=["epc_address", "epc_postcode"],
             how="left",
         )
         .join(
             social_tenure.lazy(),
-            on=["epc_address", "POSTCODE"],
+            on=["epc_address", "epc_postcode"],
             how="left",
         )
         .with_columns(
@@ -167,7 +314,7 @@ def main():
     duration_map = {"F": "Freehold", "L": "Leasehold"}
 
     price_paid = (
-        pl.scan_parquet(args.price_paid)
+        pl.scan_parquet(price_paid_path)
         .select(
             "price",
             "date_of_transfer",
@@ -219,9 +366,9 @@ def main():
             left_address_col="pp_address",
             right_address_col="epc_address",
             left_postcode_col="postcode",
-            right_postcode_col="POSTCODE",
+            right_postcode_col="epc_postcode",
         )
-        .drop("POSTCODE")
+        .drop("epc_postcode")
         .collect(engine="streaming")
     )
 
@@ -236,7 +383,7 @@ def main():
     # For new-builds (old_new == "Y"), use the first transaction date year as
     # the exact construction date; otherwise fall back to the EPC age band.
     epc_band_year = (
-        pl.col("CONSTRUCTION_AGE_BAND")
+        pl.col("construction_age_band")
         .str.replace("England and Wales: ", "")
         .str.replace(" onwards", "")
         .str.extract(r"(\d{4})", 1)
@@ -251,7 +398,7 @@ def main():
         pl.when(is_new_build & transfer_year.is_not_null())
         .then(transfer_year)
         .otherwise(epc_band_year)
-        .alias("CONSTRUCTION_AGE_BAND"),
+        .alias("construction_age_band"),
         pl.when(is_new_build & transfer_year.is_not_null())
         .then(pl.lit(0, dtype=pl.UInt8))
         .when(epc_band_year.is_not_null())
@@ -263,8 +410,8 @@ def main():
     joined = joined.rename({col: col.lower() for col in joined.columns})
 
     print(joined.head())
-    joined.write_parquet(args.output)
-    print(f"Wrote {args.output}")
+    joined.write_parquet(output_path)
+    print(f"Wrote {output_path}")
 
 
 if __name__ == "__main__":
diff --git a/pipeline/transform/merge.py b/pipeline/transform/merge.py
index 8b485a2..56b6b90 100644
--- a/pipeline/transform/merge.py
+++ b/pipeline/transform/merge.py
@@ -7,6 +7,15 @@ from pipeline.utils.postcode_mapping import build_postcode_mapping
 
 MIN_FLOOR_AREA_M2 = 10
 
+_IOD_PERCENTILE_COLUMNS = [
+    "Education, Skills and Training Score",
+    "Income Score (rate)",
+    "Employment Score (rate)",
+    "Health Deprivation and Disability Score",
+    "Indoors Sub-domain Score",
+    "Outdoors Sub-domain Score",
+]
+
 
 _AREA_COLUMNS = [
     "Postcode",
@@ -76,6 +85,24 @@ _AREA_COLUMNS = [
 ]
 
 
+def _less_deprived_percentile_expr(column: str) -> pl.Expr:
+    """Convert an IoD deprivation score to a 0-100 less-deprived percentile."""
+    non_null_count = pl.col(column).count()
+    descending_rank = pl.col(column).rank("average", descending=True)
+    return (
+        pl.when(pl.col(column).is_null())
+        .then(None)
+        .when(pl.col(column) == pl.col(column).min())
+        .then(100.0)
+        .when(pl.col(column) == pl.col(column).max())
+        .then(0.0)
+        .when(non_null_count > 1)
+        .then(((descending_rank - 1) / (non_null_count - 1) * 100).round(1))
+        .otherwise(100.0)
+        .alias(column)
+    )
+
+
 def _build(
     epc_pp_path: Path,
     arcgis_path: Path,
@@ -134,20 +161,11 @@ def _build(
     )
     wide = wide.join(arcgis, on="postcode", how="left")
 
-    iod = pl.scan_parquet(iod_path)
+    iod = pl.scan_parquet(iod_path).with_columns(
+        *(_less_deprived_percentile_expr(c) for c in _IOD_PERCENTILE_COLUMNS)
+    )
     wide = wide.join(iod, left_on="lsoa21", right_on="LSOA code (2021)", how="left")
 
-    # Invert deprivation scores so that higher values = less deprived (better)
-    iod_score_cols = [
-        "Education, Skills and Training Score",
-        "Income Score (rate)",
-        "Employment Score (rate)",
-        "Health Deprivation and Disability Score",
-        "Indoors Sub-domain Score",
-        "Outdoors Sub-domain Score",
-    ]
-    wide = wide.with_columns(*(pl.col(c).max() - pl.col(c) for c in iod_score_cols))
-
     ethnicity = pl.scan_parquet(ethnicity_path)
     wide = wide.join(
         ethnicity,
diff --git a/pipeline/transform/poi_proximity.py b/pipeline/transform/poi_proximity.py
index f3a1204..4634d7f 100644
--- a/pipeline/transform/poi_proximity.py
+++ b/pipeline/transform/poi_proximity.py
@@ -1,6 +1,8 @@
 """Compute POI proximity counts and distances per postcode from ArcGIS + filtered POIs."""
 
 import argparse
+import re
+import unicodedata
 from pathlib import Path
 
 import polars as pl
@@ -15,9 +17,25 @@ POI_GROUPS_2KM = {
     "groceries": ["Greengrocer", "Supermarket", "Convenience Store"],
 }
 
-# Groups for which to compute distance to nearest POI (from filtered POIs)
+# Groups for which to compute distance to nearest POI (from filtered POIs).
+# Keep `train_tube` for the existing backend feature; the individual POI
+# distance filters below power the frontend dropdown.
 DISTANCE_GROUPS = {
     "train_tube": ["Tube station", "Rail station"],
+    "grocery_store": [
+        "Greengrocer",
+        "Supermarket",
+        "Convenience Store",
+        "Waitrose",
+        "Tesco",
+    ],
+    "tube_station": ["Tube station"],
+    "rail_station": ["Rail station"],
+    "waitrose": ["Waitrose"],
+    "tesco": ["Tesco"],
+    "cafe": ["Café"],
+    "pub": ["Pub"],
+    "restaurant": ["Restaurant"],
 }
 
 # OS Open Greenspace function types used for park counts and distance calculation.
@@ -27,6 +45,69 @@ GREENSPACE_PARK_FUNCTIONS = {
     "parks": ["Public Park Or Garden", "Playing Field", "Play Space"],
 }
 
+GROCERY_DYNAMIC_FILTER_MIN_POIS = 100
+DYNAMIC_FILTER_ALL_GROUPS = {"Public Transport", "Leisure"}
+DYNAMIC_FILTER_COUNT_THRESHOLD_GROUPS = {"Groceries"}
+
+
+def _poi_category_slug(category: str) -> str:
+    ascii_text = (
+        unicodedata.normalize("NFKD", category)
+        .encode("ascii", "ignore")
+        .decode("ascii")
+        .lower()
+    )
+    slug = re.sub(r"[^a-z0-9]+", "_", ascii_text).strip("_")
+    return slug or "poi"
+
+
+def _build_poi_category_groups(
+    pois: pl.DataFrame,
+) -> tuple[dict[str, list[str]], dict[str, str]]:
+    """Build one proximity group for each POI category selected for filters."""
+    if "group" not in pois.columns:
+        raise ValueError("POI dataframe must include a 'group' column")
+
+    categories = (
+        pois.group_by("group", "category")
+        .len()
+        .filter(
+            pl.col("group").is_in(list(DYNAMIC_FILTER_ALL_GROUPS))
+            | (
+                pl.col("group").is_in(list(DYNAMIC_FILTER_COUNT_THRESHOLD_GROUPS))
+                & (pl.col("len") > GROCERY_DYNAMIC_FILTER_MIN_POIS)
+            )
+        )
+        .select("category")
+        .sort("category")
+        .to_series()
+        .to_list()
+    )
+    used_slugs: dict[str, int] = {}
+    groups: dict[str, list[str]] = {}
+    display_names: dict[str, str] = {}
+
+    for category in categories:
+        if not isinstance(category, str) or not category:
+            continue
+        base_slug = f"poi_{_poi_category_slug(category)}"
+        slug_count = used_slugs.get(base_slug, 0)
+        used_slugs[base_slug] = slug_count + 1
+        group_key = base_slug if slug_count == 0 else f"{base_slug}_{slug_count + 1}"
+        groups[group_key] = [category]
+        display_names[group_key] = category
+
+    return groups, display_names
+
+
+def _dynamic_poi_metric_renames(display_names: dict[str, str]) -> dict[str, str]:
+    renames: dict[str, str] = {}
+    for group_key, category in display_names.items():
+        renames[f"{group_key}_nearest_km"] = f"Distance to nearest {category} POI (km)"
+        renames[f"{group_key}_2km"] = f"Number of {category} POIs within 2km"
+        renames[f"{group_key}_5km"] = f"Number of {category} POIs within 5km"
+    return renames
+
 
 def main():
     parser = argparse.ArgumentParser(
@@ -56,12 +137,35 @@ def main():
     )
 
     pois = pl.read_parquet(args.pois)
+    poi_category_groups, poi_display_names = _build_poi_category_groups(pois)
 
     # Count amenity POIs within 2km
     counts_2km = count_pois_per_postcode(
         postcodes, pois, groups=POI_GROUPS_2KM, radius_km=2
     )
 
+    # Dynamic POI filters: nearest distance plus counts within 2km and 5km for
+    # the selected public transport, grocery, and leisure categories.
+    dynamic_counts_2km = count_pois_per_postcode(
+        postcodes, pois, groups=poi_category_groups, radius_km=2
+    )
+    dynamic_counts_5km = count_pois_per_postcode(
+        postcodes, pois, groups=poi_category_groups, radius_km=5
+    )
+    dynamic_distances = min_distance_per_postcode(
+        postcodes, pois, groups=poi_category_groups
+    )
+    dynamic_renames = _dynamic_poi_metric_renames(poi_display_names)
+    dynamic_counts_2km = dynamic_counts_2km.rename(
+        {k: v for k, v in dynamic_renames.items() if k in dynamic_counts_2km.columns}
+    )
+    dynamic_counts_5km = dynamic_counts_5km.rename(
+        {k: v for k, v in dynamic_renames.items() if k in dynamic_counts_5km.columns}
+    )
+    dynamic_distances = dynamic_distances.rename(
+        {k: v for k, v in dynamic_renames.items() if k in dynamic_distances.columns}
+    )
+
     # Distance to nearest train/tube station (from filtered POIs)
     distances = min_distance_per_postcode(postcodes, pois, groups=DISTANCE_GROUPS)
 
@@ -77,6 +181,9 @@ def main():
     # Join all results on postcode
     result = (
         counts_2km.join(distances, on="postcode")
+        .join(dynamic_counts_2km, on="postcode")
+        .join(dynamic_counts_5km, on="postcode")
+        .join(dynamic_distances, on="postcode")
         .join(park_counts_1km, on="postcode")
         .join(park_distances, on="postcode")
     )
diff --git a/pipeline/transform/test_crime.py b/pipeline/transform/test_crime.py
new file mode 100644
index 0000000..fa5d40c
--- /dev/null
+++ b/pipeline/transform/test_crime.py
@@ -0,0 +1,47 @@
+import polars as pl
+
+from pipeline.transform.crime import find_street_crime_csvs, transform_crime
+
+
+def test_find_street_crime_csvs_ignores_archive_sidecars(tmp_path):
+    crime_dir = tmp_path / "crime"
+    month_dir = crime_dir / "2024-01"
+    month_dir.mkdir(parents=True)
+    street = month_dir / "2024-01-test-force-street.csv"
+    street.touch()
+    (month_dir / "2024-01-test-force-outcomes.csv").touch()
+    (month_dir / "2024-01-test-force-stop-and-search.csv").touch()
+    (crime_dir / "notes.csv").touch()
+
+    csvs, ignored_count = find_street_crime_csvs(crime_dir)
+
+    assert csvs == [street]
+    assert ignored_count == 3
+
+
+def test_transform_crime_reads_only_street_crime_csvs(tmp_path):
+    crime_dir = tmp_path / "crime"
+    month_dir = crime_dir / "2024-01"
+    month_dir.mkdir(parents=True)
+
+    (month_dir / "2024-01-test-force-street.csv").write_text(
+        "\n".join(
+            [
+                "Crime ID,Month,Reported by,Falls within,Longitude,Latitude,Location,LSOA code,LSOA name,Crime type,Last outcome category,Context",
+                "1,2024-01,Test Force,Test Force,-0.1,51.5,On or near Test Street,E01000001,Test LSOA,Burglary,Under investigation,",
+                "2,2024-01,Test Force,Test Force,-0.1,51.5,On or near Test Street,E01000001,Test LSOA,Burglary,Under investigation,",
+                "3,2024-01,Test Force,Test Force,-0.1,51.5,On or near Test Street,,No LSOA,Robbery,Under investigation,",
+            ]
+        )
+        + "\n"
+    )
+    (month_dir / "2024-01-test-force-outcomes.csv").write_text(
+        "Crime ID,Month,Reported by,Outcome type\n1,2024-01,Test Force,Charged\n"
+    )
+
+    output = tmp_path / "crime.parquet"
+    transform_crime(crime_dir, output)
+
+    result = pl.read_parquet(output).to_dicts()
+
+    assert result == [{"LSOA code": "E01000001", "Burglary (avg/yr)": 2.0}]
diff --git a/pipeline/transform/test_join_epc_pp.py b/pipeline/transform/test_join_epc_pp.py
new file mode 100644
index 0000000..7980ea9
--- /dev/null
+++ b/pipeline/transform/test_join_epc_pp.py
@@ -0,0 +1,174 @@
+import csv
+import io
+import zipfile
+from datetime import date
+from pathlib import Path
+
+import polars as pl
+
+from pipeline.transform.join_epc_pp import (
+    EPC_SOURCE_COLUMNS,
+    _run,
+    _scan_epc_certificates,
+)
+
+
+def _write_csv(path: Path, fieldnames: list[str], rows: list[dict[str, str]]) -> None:
+    with path.open("w", newline="") as file:
+        writer = csv.DictWriter(file, fieldnames=fieldnames)
+        writer.writeheader()
+        writer.writerows(rows)
+
+
+def _row(**overrides: str) -> dict[str, str]:
+    row = {
+        "address": "1 Example Street",
+        "postcode": " aa1 1aa ",
+        "current_energy_rating": "c",
+        "potential_energy_rating": "b",
+        "property_type": "House",
+        "built_form": "Mid-Terrace",
+        "inspection_date": "2024-01-02",
+        "total_floor_area": "84.5",
+        "number_habitable_rooms": "5",
+        "floor_height": "2.4",
+        "construction_age_band": "England and Wales: 1950-1966",
+        "tenure": "owner-occupied",
+    }
+    row.update(overrides)
+    return row
+
+
+def test_scan_epc_certificates_supports_legacy_uppercase_csv(tmp_path: Path):
+    csv_path = tmp_path / "certificates.csv"
+    fieldnames = [column.upper() for column in EPC_SOURCE_COLUMNS]
+    row = {column.upper(): value for column, value in _row().items()}
+    row["NUMBER_HABITABLE_ROOMS"] = "0"
+    _write_csv(csv_path, fieldnames, [row])
+
+    df = _scan_epc_certificates(csv_path, tmp_path).collect()
+
+    assert df.to_dicts() == [
+        {
+            "epc_address": "1 Example Street",
+            "epc_postcode": "AA1 1AA",
+            "current_energy_rating": "C",
+            "potential_energy_rating": "B",
+            "epc_property_type": "House",
+            "built_form": "Mid-Terrace",
+            "inspection_date": "2024-01-02",
+            "total_floor_area": 84.5,
+            "number_habitable_rooms": None,
+            "floor_height": 2.4,
+            "construction_age_band": "England and Wales: 1950-1966",
+            "tenure": "owner-occupied",
+        }
+    ]
+
+
+def test_scan_epc_certificates_supports_domestic_zip(tmp_path: Path):
+    zip_path = tmp_path / "domestic-csv.zip"
+    rows_2023 = [_row(address="2 Example Street", inspection_date="2023-03-04")]
+    rows_2024 = [
+        _row(
+            address="3 Example Street",
+            postcode="BB2 2BB",
+            inspection_date="2024-05-06",
+            total_floor_area="",
+            tenure="Rented (social)",
+        )
+    ]
+
+    with zipfile.ZipFile(zip_path, "w", compression=zipfile.ZIP_DEFLATED) as archive:
+        for member_name, rows in [
+            ("certificates-2023.csv", rows_2023),
+            ("nested/certificates-2024.csv", rows_2024),
+        ]:
+            csv_text = [",".join(EPC_SOURCE_COLUMNS)]
+            csv_text.extend(
+                ",".join(row[column] for column in EPC_SOURCE_COLUMNS) for row in rows
+            )
+            archive.writestr(member_name, "\n".join(csv_text) + "\n")
+        archive.writestr("recommendations-2024.csv", "address,postcode\nignored,X\n")
+
+    df = _scan_epc_certificates(zip_path, tmp_path).sort("inspection_date").collect()
+
+    assert df.select("epc_address", "epc_postcode", "total_floor_area").to_dicts() == [
+        {
+            "epc_address": "2 Example Street",
+            "epc_postcode": "AA1 1AA",
+            "total_floor_area": 84.5,
+        },
+        {
+            "epc_address": "3 Example Street",
+            "epc_postcode": "BB2 2BB",
+            "total_floor_area": None,
+        },
+    ]
+    assert df.get_column("tenure").to_list() == ["owner-occupied", "Rented (social)"]
+    assert df.schema["number_habitable_rooms"] == pl.Int16
+
+
+def test_run_joins_domestic_zip_with_price_paid(tmp_path: Path):
+    zip_path = tmp_path / "domestic-csv.zip"
+    with zipfile.ZipFile(zip_path, "w", compression=zipfile.ZIP_DEFLATED) as archive:
+        csv_buffer = io.StringIO()
+        writer = csv.DictWriter(csv_buffer, fieldnames=EPC_SOURCE_COLUMNS)
+        writer.writeheader()
+        writer.writerows(
+            [
+                _row(
+                    current_energy_rating="d",
+                    inspection_date="2023-01-01",
+                    total_floor_area="80",
+                    tenure="Rented (social)",
+                ),
+                _row(
+                    current_energy_rating="c",
+                    inspection_date="2024-01-01",
+                    total_floor_area="85",
+                    tenure="owner-occupied",
+                ),
+            ]
+        )
+        archive.writestr("certificates-2024.csv", csv_buffer.getvalue())
+
+    price_paid_path = tmp_path / "price-paid.parquet"
+    pl.DataFrame(
+        {
+            "price": [250_000],
+            "date_of_transfer": [date(2024, 2, 3)],
+            "property_type": ["T"],
+            "postcode": ["AA1 1AA"],
+            "paon": ["1"],
+            "saon": [None],
+            "street": ["Example Street"],
+            "locality": [None],
+            "town_city": ["Exampletown"],
+            "duration": ["F"],
+            "old_new": ["N"],
+        }
+    ).write_parquet(price_paid_path)
+
+    output_path = tmp_path / "epc-pp.parquet"
+    _run(zip_path, price_paid_path, output_path, tmp_path)
+
+    df = pl.read_parquet(output_path)
+
+    assert df.height == 1
+    assert df.select(
+        "epc_address",
+        "current_energy_rating",
+        "total_floor_area",
+        "construction_age_band",
+        "was_council_house",
+    ).to_dicts() == [
+        {
+            "epc_address": "1 Example Street",
+            "current_energy_rating": "C",
+            "total_floor_area": 85.0,
+            "construction_age_band": 1950,
+            "was_council_house": "Yes",
+        }
+    ]
+    assert df.get_column("renovation_history").list.len().to_list() == [1]
diff --git a/pipeline/transform/test_merge.py b/pipeline/transform/test_merge.py
new file mode 100644
index 0000000..3964eca
--- /dev/null
+++ b/pipeline/transform/test_merge.py
@@ -0,0 +1,33 @@
+import polars as pl
+
+from pipeline.transform.merge import (
+    _is_dynamic_poi_metric_column,
+    _less_deprived_percentile_expr,
+)
+
+
+def test_less_deprived_percentile_expr_preserves_direction_and_nulls() -> None:
+    df = pl.DataFrame({"Income Score (rate)": [1.0, 2.0, 3.0, None]})
+
+    result = df.lazy().with_columns(
+        _less_deprived_percentile_expr("Income Score (rate)")
+    ).collect()
+
+    assert result["Income Score (rate)"].to_list() == [100.0, 50.0, 0.0, None]
+
+
+def test_less_deprived_percentile_expr_uses_exact_scale_endpoints() -> None:
+    df = pl.DataFrame({"Income Score (rate)": [1.0, 1.0, 2.0, 3.0, 3.0]})
+
+    result = df.lazy().with_columns(
+        _less_deprived_percentile_expr("Income Score (rate)")
+    ).collect()
+
+    assert result["Income Score (rate)"].to_list() == [100.0, 100.0, 50.0, 0.0, 0.0]
+
+
+def test_dynamic_poi_metric_columns_are_area_level() -> None:
+    assert _is_dynamic_poi_metric_column("Distance to nearest Cafe POI (km)")
+    assert _is_dynamic_poi_metric_column("Number of Cafe POIs within 2km")
+    assert _is_dynamic_poi_metric_column("Number of Cafe POIs within 5km")
+    assert not _is_dynamic_poi_metric_column("Number of restaurants within 2km")
diff --git a/pipeline/transform/test_poi_proximity.py b/pipeline/transform/test_poi_proximity.py
new file mode 100644
index 0000000..68fb9c7
--- /dev/null
+++ b/pipeline/transform/test_poi_proximity.py
@@ -0,0 +1,41 @@
+import polars as pl
+
+from pipeline.transform.poi_proximity import _build_poi_category_groups
+
+
+def test_dynamic_poi_groups_include_requested_categories_only() -> None:
+    pois = pl.DataFrame(
+        {
+            "group": (
+                ["Public Transport"] * 2
+                + ["Leisure"] * 2
+                + ["Groceries"] * 101
+                + ["Groceries"] * 100
+                + ["Education"] * 200
+                + ["Health"] * 200
+            ),
+            "category": (
+                ["Rail station", "Bus stop"]
+                + ["Café", "Restaurant"]
+                + ["Tesco"] * 101
+                + ["Waitrose"] * 100
+                + ["School"] * 200
+                + ["Pharmacy"] * 200
+            ),
+            "lat": [51.5] * 605,
+            "lng": [-0.1] * 605,
+        }
+    )
+
+    groups, display_names = _build_poi_category_groups(pois)
+
+    assert set(display_names.values()) == {
+        "Bus stop",
+        "Café",
+        "Rail station",
+        "Restaurant",
+        "Tesco",
+    }
+    assert "poi_waitrose" not in groups
+    assert "poi_school" not in groups
+    assert "poi_pharmacy" not in groups
diff --git a/pipeline/transform/test_transform_poi.py b/pipeline/transform/test_transform_poi.py
index 2d14716..a9e0a0a 100644
--- a/pipeline/transform/test_transform_poi.py
+++ b/pipeline/transform/test_transform_poi.py
@@ -79,6 +79,33 @@ def test_transform_grocery_retail_points_keeps_fascia_icon_category():
     ]
 
 
+def test_transform_grocery_retail_points_accepts_base_fascias():
+    raw = pl.DataFrame(
+        {
+            "id": [101, 102, 103, 104],
+            "retailer": ["Aldi", "Asda", "Booths", "Whole Foods Market"],
+            "fascia": ["Aldi", "Asda Superstore", "Booths", "Whole Foods Market"],
+            "store_name": [
+                "Aldi Test",
+                "Asda Test Superstore",
+                "Booths Test",
+                "Whole Foods Test",
+            ],
+            "long_wgs": [-0.141, -0.142, -0.143, -0.144],
+            "lat_wgs": [51.515, 51.516, 51.517, 51.518],
+        }
+    )
+
+    pois = transform_grocery_retail_points(raw)
+
+    assert pois.select("category", "icon_category").to_dicts() == [
+        {"category": "Aldi", "icon_category": "Aldi"},
+        {"category": "Asda", "icon_category": "Asda Superstore"},
+        {"category": "Booths", "icon_category": "Booths"},
+        {"category": "Whole Foods Market", "icon_category": "Whole Foods Market"},
+    ]
+
+
 def test_transform_grocery_retail_points_drops_invalid_rows():
     raw = pl.DataFrame(
         {
diff --git a/pipeline/transform/transform_poi.py b/pipeline/transform/transform_poi.py
index 831d6d2..5c9b237 100644
--- a/pipeline/transform/transform_poi.py
+++ b/pipeline/transform/transform_poi.py
@@ -1078,19 +1078,40 @@ COOP_RETAILERS = {
 }
 
 GROCERY_RETAILER_DISPLAY_NAMES: dict[str, str] = {
+    "Aldi": "Aldi",
+    "Asda": "Asda",
+    "Booths": "Booths",
+    "Budgens": "Budgens",
+    "Centra": "Centra",
     "Cook": "COOK",
+    "Costco": "Costco",
+    "Dunnes Stores": "Dunnes Stores",
+    "Farmfoods": "Farmfoods",
     "Heron": "Heron Foods",
+    "Iceland": "Iceland",
+    "Lidl": "Lidl",
+    "Makro": "Makro",
     "Marks and Spencer": "M&S",
+    "Morrisons": "Morrisons",
+    "Planet Organic": "Planet Organic",
     "Sainsburys": "Sainsbury's",
+    "Spar": "Spar",
+    "Tesco": "Tesco",
+    "Waitrose": "Waitrose",
+    "Whole Foods Market": "Whole Foods Market",
     **{retailer: "Co-op" for retailer in COOP_RETAILERS},
 }
 
 
 GROCERY_FASCIA_ICON_NAMES: dict[str, str] = {
+    **GROCERY_RETAILER_DISPLAY_NAMES,
     "Aldi Local": "Aldi",
     "Asda Express": "Asda Express",
     "Asda Living": "Asda Living",
     "Asda PFS": "Asda PFS",
+    "Asda Supercentre": "Asda Supercentre",
+    "Asda Supermarket": "Asda Supermarket",
+    "Asda Superstore": "Asda Superstore",
     "Cooltrader": "Heron Foods",
     "Co-op Food": "Co-op",
     "Cook": "COOK",
@@ -1112,6 +1133,7 @@ GROCERY_FASCIA_ICON_NAMES: dict[str, str] = {
     "Marks and Spencer Travel SF": "M&S Food",
     "Morrisons Daily": "Morrisons Daily",
     "Morrisons Select": "Morrisons",
+    "Sainsbury's Local": "Sainsbury's Local",
     "Sainsburys": "Sainsbury's",
     "Sainsburys Local": "Sainsbury's Local",
     "Spar PFS": "Spar",
@@ -1128,12 +1150,18 @@ GROCERY_FASCIA_ICON_NAMES: dict[str, str] = {
 def normalize_grocery_retailer(retailer: str | None) -> str:
     if retailer is None:
         return ""
-    return GROCERY_RETAILER_DISPLAY_NAMES.get(retailer, retailer)
+    display_name = GROCERY_RETAILER_DISPLAY_NAMES.get(retailer)
+    if display_name is None:
+        raise ValueError(f"Missing grocery retailer display name for {retailer!r}")
+    return display_name
 
 
 def normalize_grocery_icon_category(fascia: str | None, retailer: str | None) -> str:
     if fascia:
-        return GROCERY_FASCIA_ICON_NAMES.get(fascia, normalize_grocery_retailer(fascia))
+        icon_name = GROCERY_FASCIA_ICON_NAMES.get(fascia)
+        if icon_name is None:
+            raise ValueError(f"Missing grocery fascia icon name for {fascia!r}")
+        return icon_name
     return normalize_grocery_retailer(retailer)
 
 
diff --git a/pipeline/utils/poi_counts.py b/pipeline/utils/poi_counts.py
index 8685d66..cdf9bfa 100644
--- a/pipeline/utils/poi_counts.py
+++ b/pipeline/utils/poi_counts.py
@@ -2,9 +2,12 @@
 
 import numpy as np
 import polars as pl
+from scipy.spatial import cKDTree
 
 from .haversine import haversine_km
 
+EARTH_RADIUS_KM = 6371.0088
+
 
 def _build_poi_grid(
     pois: pl.DataFrame, grid_size: float = 0.05
@@ -49,6 +52,21 @@ def _get_nearby_indices(
     return np.concatenate(nearby_indices)
 
 
+def _project_lat_lng_km(
+    lats: np.ndarray, lngs: np.ndarray, origin_lat: float
+) -> np.ndarray:
+    """Project WGS84 coordinates to local km coordinates for nearest-neighbour lookup."""
+    lat_rad = np.radians(lats)
+    lng_rad = np.radians(lngs)
+    origin_lat_rad = np.radians(origin_lat)
+    return np.column_stack(
+        (
+            EARTH_RADIUS_KM * lng_rad * np.cos(origin_lat_rad),
+            EARTH_RADIUS_KM * lat_rad,
+        )
+    )
+
+
 def count_pois_per_postcode(
     postcodes_df: pl.DataFrame,
     pois: pl.DataFrame,
@@ -136,7 +154,7 @@ def min_distance_per_postcode(
 ) -> pl.DataFrame:
     """
     For each postcode, compute the distance (km) to the closest POI per group.
-    Returns NaN where no POI of that group exists within the grid search range (~5.5km).
+    Returns NaN where no POI of that group exists.
     """
     print("Computing minimum POI distances per postcode...")
 
@@ -144,51 +162,84 @@ def min_distance_per_postcode(
     n_pois = len(pois)
     print(f"  {n_postcodes:,} postcodes, {n_pois:,} POIs")
 
-    grid_size = 0.05
-    print("  Building POI spatial grid...")
-    poi_lats, poi_lngs, poi_cats, poi_grid = _build_poi_grid(pois, grid_size)
-    print(f"  POI grid has {len(poi_grid):,} occupied cells")
-
-    category_masks = {}
-    for group, categories in groups.items():
-        mask = np.isin(poi_cats, categories)
-        category_masks[group] = mask
-        print(f"  {group}: {mask.sum():,} POIs")
-
     pc_lats = postcodes_df["lat"].to_numpy()
     pc_lons = postcodes_df["lon"].to_numpy()
     pc_codes = postcodes_df["postcode"].to_list()
+    valid_pc_mask = np.isfinite(pc_lats) & np.isfinite(pc_lons)
+    valid_pc_indices = np.flatnonzero(valid_pc_mask)
 
     result_min_dist = {
         group: np.full(n_postcodes, np.nan, dtype=np.float32) for group in groups
     }
 
-    batch_size = 50000
-    n_batches = (n_postcodes + batch_size - 1) // batch_size
-    print(f"  Processing {n_postcodes:,} postcodes in {n_batches} batches...")
+    if n_pois == 0 or len(valid_pc_indices) == 0:
+        print("  No valid postcode/POI coordinates; returning NaN distances")
+        return pl.DataFrame(
+            {
+                "postcode": pc_codes,
+                **{
+                    f"{group}_nearest_km": values
+                    for group, values in result_min_dist.items()
+                },
+            }
+        )
 
-    for batch_idx in range(n_batches):
-        start_idx = batch_idx * batch_size
-        end_idx = min(start_idx + batch_size, n_postcodes)
+    poi_lats = pois["lat"].to_numpy()
+    poi_lngs = pois["lng"].to_numpy()
+    poi_cats = pois["category"].to_numpy()
+    valid_poi_mask = np.isfinite(poi_lats) & np.isfinite(poi_lngs)
+    origin_lat = float(np.nanmean(pc_lats[valid_pc_mask]))
+    query_xy = _project_lat_lng_km(
+        pc_lats[valid_pc_indices], pc_lons[valid_pc_indices], origin_lat
+    )
 
-        if batch_idx % 5 == 0:
-            print(
-                f"  Batch {batch_idx + 1}/{n_batches}: postcodes {start_idx:,} - {end_idx:,}"
-            )
+    batch_size = 200_000
+    n_batches = (len(valid_pc_indices) + batch_size - 1) // batch_size
 
-        for i in range(start_idx, end_idx):
-            nearby = _get_nearby_indices(pc_lats[i], pc_lons[i], poi_grid, grid_size)
-            if nearby is None:
-                continue
+    for group, categories in groups.items():
+        group_indices = np.flatnonzero(valid_poi_mask & np.isin(poi_cats, categories))
+        print(f"  {group}: {len(group_indices):,} POIs")
+        if len(group_indices) == 0:
+            continue
 
-            distances = haversine_km(
-                poi_lats[nearby], poi_lngs[nearby], pc_lats[i], pc_lons[i]
-            )
+        poi_xy = _project_lat_lng_km(
+            poi_lats[group_indices], poi_lngs[group_indices], origin_lat
+        )
+        tree = cKDTree(poi_xy)
+        k = min(8, len(group_indices))
 
-            for group, cat_mask in category_masks.items():
-                group_mask = cat_mask[nearby]
-                if group_mask.any():
-                    result_min_dist[group][i] = distances[group_mask].min()
+        for batch_idx in range(n_batches):
+            start_idx = batch_idx * batch_size
+            end_idx = min(start_idx + batch_size, len(valid_pc_indices))
+            batch_pc_indices = valid_pc_indices[start_idx:end_idx]
+            batch_xy = query_xy[start_idx:end_idx]
+
+            if batch_idx == 0 or (batch_idx + 1) % 5 == 0:
+                print(
+                    f"    Batch {batch_idx + 1}/{n_batches}: postcodes {start_idx:,} - {end_idx:,}"
+                )
+
+            _, nearest = tree.query(batch_xy, k=k)
+            nearest = np.asarray(nearest)
+
+            if k == 1:
+                candidate_indices = group_indices[nearest]
+                distances = haversine_km(
+                    poi_lats[candidate_indices],
+                    poi_lngs[candidate_indices],
+                    pc_lats[batch_pc_indices],
+                    pc_lons[batch_pc_indices],
+                )
+            else:
+                candidate_indices = group_indices[nearest]
+                distances = haversine_km(
+                    poi_lats[candidate_indices],
+                    poi_lngs[candidate_indices],
+                    pc_lats[batch_pc_indices, None],
+                    pc_lons[batch_pc_indices, None],
+                ).min(axis=1)
+
+            result_min_dist[group][batch_pc_indices] = distances.astype(np.float32)
 
     result_data = {"postcode": pc_codes}
     for group in groups:
diff --git a/pipeline/utils/test_poi_counts.py b/pipeline/utils/test_poi_counts.py
index 7e7ad96..6d61ccb 100644
--- a/pipeline/utils/test_poi_counts.py
+++ b/pipeline/utils/test_poi_counts.py
@@ -113,9 +113,9 @@ def test_min_distance_finds_nearest(postcodes, pois):
     # Restaurant is co-located — distance ~0
     assert ec1a["restaurants_nearest_km"][0] < 0.01
 
-    # Far-away postcode should have NaN (no POIs within grid range)
+    # Far-away postcode should still get the global nearest distance.
     zz99 = result.filter(pl.col("postcode") == "ZZ99 9ZZ")
-    assert np.isnan(zz99["train_tube_nearest_km"][0])
+    assert zz99["train_tube_nearest_km"][0] > 300
 
 
 def test_min_distance_no_pois_returns_nan(postcodes):
diff --git a/r5-java/run.sh b/r5-java/run.sh
index 10adce7..88209be 100755
--- a/r5-java/run.sh
+++ b/r5-java/run.sh
@@ -111,20 +111,24 @@ fi
 # R5 writes .mapdb temp files next to OSM/GTFS files during network construction.
 # Copy source data to a writable build dir to avoid polluting the originals.
 mkdir -p "$NETWORK_DIR"
+OSM_PBF="property-data/england-latest.osm.pbf"
 TRANSIT_SRC="property-data/transit"
-NETWORK_DATA_DIR="$TRANSIT_SRC"
+NETWORK_DATA_DIR="$NETWORK_DIR/build"
 
 if [ ! -f "$NETWORK_DIR/network.dat" ]; then
     BUILD_DIR="$NETWORK_DIR/build"
     echo "--- No cached network — copying transit data to build dir ---"
     mkdir -p "$BUILD_DIR"
-    if ! cp "$TRANSIT_SRC"/raw/*.osm.pbf "$BUILD_DIR/" 2>/dev/null; then
-        echo "Warning: no .osm.pbf files found in $TRANSIT_SRC/raw/"
+    if [ ! -f "$OSM_PBF" ]; then
+        echo "Error: OSM PBF not found at $OSM_PBF"
+        echo "Download it from https://download.geofabrik.de/europe/united-kingdom/england-latest.osm.pbf"
+        exit 1
     fi
+    cp "$OSM_PBF" "$BUILD_DIR/"
     if ! cp "$TRANSIT_SRC"/*.zip "$BUILD_DIR/" 2>/dev/null; then
-        echo "Warning: no .zip files found in $TRANSIT_SRC/"
+        echo "Warning: no GTFS .zip files found in $TRANSIT_SRC/ — transit routing would be unavailable"
+        exit 1
     fi
-    NETWORK_DATA_DIR="$BUILD_DIR"
 fi
 
 # --- Step 5: Run batch ---
diff --git a/r5-java/src/main/java/propertymap/Parquet.java b/r5-java/src/main/java/propertymap/Parquet.java
index cb16f6c..f784508 100644
--- a/r5-java/src/main/java/propertymap/Parquet.java
+++ b/r5-java/src/main/java/propertymap/Parquet.java
@@ -32,7 +32,7 @@ public class Parquet {
     static Postcodes loadEnglandPostcodes(String parquetPath, Path refOut) throws Exception {
         try (DuckDBConnection conn = connect(); Statement stmt = conn.createStatement()) {
             stmt.execute("CREATE TABLE postcodes AS SELECT pcds, lat, \"long\" FROM read_parquet('"
-                + escapePath(parquetPath) + "') WHERE ctry = 'E92000001' AND doterm IS NULL");
+                + escapePath(parquetPath) + "') WHERE ctry25cd = 'E92000001' AND doterm IS NULL");
             copyToParquet(stmt, "SELECT * FROM postcodes", refOut);
 
             try (ResultSet rs = stmt.executeQuery("SELECT COUNT(*) FROM postcodes")) {