diff --git a/etl/join_building_data/load_csv_to_staging.py b/etl/join_building_data/load_csv_to_staging.py
new file mode 100644
index 00000000..d1d9bbe1
--- /dev/null
+++ b/etl/join_building_data/load_csv_to_staging.py
@@ -0,0 +1,108 @@
+"""Join csv data to buildings
+Example usage (replace URL with test/staging/localhost as necessary, API key with real key for
+the appropriate site):
+    python load_csv.py \
+        https://colouring.london \
+        a0a00000-0a00-0aaa-a0a0-0000aaaa0000 \
+        data.csv
+This script uses the HTTP API, and can process CSV files which identify buildings by id, TOID,
+UPRN.
+The process:
+    - assume first line of the CSV is a header, where column names are either
+        - building identifiers - one of:
+            - building_id
+            - toid
+            - uprn
+        - building data field names
+    - read through lines of CSV:
+        - use building id if provided
+            - else lookup by toid
+            - else lookup by uprn
+            - else locate building by representative point
+        - update building
+TODO extend to allow latitude,longitude or easting,northing columns and lookup by location.
+"""
+import csv
+import json
+import os
+import sys
+
+import requests
+session = requests.Session()
+session.verify = False
+
+def main(base_url, api_key, source_file):
+    """Read from file, update buildings
+    """
+    with open(source_file, 'r') as source:
+        reader = csv.DictReader(source)
+        for line in reader:
+            building_id = find_building(line, base_url)
+
+            if building_id is None:
+                continue
+
+            response_code, response_data = update_building(building_id, line, api_key, base_url)
+            if response_code != 200:
+                print('ERROR', building_id, response_code, response_data)
+
+
+def update_building(building_id, data, api_key, base_url):
+    """Save data to a building
+    """
+    r = requests.post(
+        "{}/api/buildings/{}.json".format(base_url, building_id),
+        params={'api_key': api_key},
+        json=data,
+        verify=False
+    )
+    return r.status_code, r.json()
+
+
+def find_building(data, base_url):
+    if 'toid' in data:
+        building_id = find_by_reference(base_url, 'toid', data['toid'])
+        if building_id is not None:
+            print("match_by_toid", data['toid'], building_id)
+            return building_id
+
+    if 'uprn' in data:
+        building_id =  find_by_reference(base_url, 'uprn', data['uprn'])
+        if building_id is not None:
+            print("match_by_uprn", data['uprn'], building_id)
+            return building_id
+
+    print("no_match", data)
+    return None
+
+
+def find_by_reference(base_url, ref_key, ref_id):
+    """Find building_id by TOID or UPRN
+    """
+    r = requests.get("{}/api/buildings/reference".format(base_url), params={
+        'key': ref_key,
+        'id': ref_id,
+    },
+        verify=False
+    )
+    buildings = r.json()
+
+    if buildings and 'error' not in buildings and len(buildings) == 1:
+        building_id = buildings[0]['building_id']
+    else:
+        building_id = None
+
+    return building_id
+
+
+if __name__ == '__main__':
+    try:
+        url, api_key, filename = sys.argv[1], sys.argv[2], sys.argv[3]
+    except IndexError:
+        print(
+            "Usage: {} <URL> <api_key> ./path/to/data.csv".format(
+            os.path.basename(__file__)
+        ))
+        exit()
+
+    main(url, api_key, filename)