diff --git a/exploration/admin_lookup.md b/exploration/admin_lookup.md
new file mode 100644
index 0000000..0a8a961
--- /dev/null
+++ b/exploration/admin_lookup.md
@@ -0,0 +1,135 @@
+---
+jupyter:
+  jupytext:
+    text_representation:
+      extension: .md
+      format_name: markdown
+      format_version: '1.3'
+      jupytext_version: 1.16.3
+  kernelspec:
+    display_name: Python 3 (ipykernel)
+    language: python
+    name: python3
+---
+
+# Creating an admin lookup table
+
+This notebook processes our saved Fieldmaps CODs to create a lookup table matching pcodes and places names across all relevant admin levels for which we have raster stats data. We select the **first available** column in the original CODs with the place name. In most cases this is the English, French, or Spanish name.
+
+We want an output table with the following columns:
+
+```python
+DEFAULT_COLS = ["ISO3", "ADM0_PCODE", "ADM0_NAME",  "ADM1_PCODE", "ADM1_NAME", "ADM2_PCODE", "ADM2_NAME", "NAME_LANGUAGE", "ADM_LEVEL"]
+```
+
+```python
+import pandas as pd
+from sqlalchemy import create_engine
+import tempfile
+import geopandas as gpd
+import os
+from typing import Literal
+
+from src.utils.database_utils import db_engine_url
+from src.utils.iso3_utils import get_iso3_data, load_shp_from_azure
+from src.utils.metadata_utils import select_name_column
+from azure.storage.blob import ContainerClient, ContentSettings
+from dotenv import load_dotenv
+
+load_dotenv()
+
+PROD_BLOB_SAS = os.getenv("DSCI_AZ_SAS_PROD")
+DEV_BLOB_SAS = os.getenv("DSCI_AZ_SAS_DEV")
+
+PROJECT_PREFIX = "polygon"
+
+MODE = "dev"
+engine = create_engine(db_engine_url(MODE))
+df_iso3s = get_iso3_data(iso3_codes=None, engine=engine)
+```
+
+```python
+dfs = []
+
+with tempfile.TemporaryDirectory() as temp_dir:
+    for _, row in df_iso3s.iterrows():
+        iso3 = row["iso3"]
+        max_adm_level = row["max_adm_level"]
+        load_shp_from_azure(iso3, temp_dir, MODE)
+        gdf = gpd.read_file(f"{temp_dir}/{iso3.lower()}_adm{max_adm_level}.shp")
+
+        # Get name column and its language code
+        name_column = select_name_column(gdf, max_adm_level)
+        language_code = name_column[-2:]
+        name_columns = [f"ADM{i}_{language_code}" for i in range(0, max_adm_level + 1)]
+
+        # Standardize column names and add language info
+        new_columns = [x.replace(f"_{language_code}", "_NAME") for x in name_columns]
+        gdf = gdf.rename(columns=dict(zip(name_columns, new_columns)))
+        gdf["NAME_LANGUAGE"] = language_code
+        gdf["ISO3"] = iso3
+        gdf["ADM_LEVEL"] = max_adm_level
+
+        # Keep only relevant columns
+        matching_cols = [col for col in gdf.columns if col in DEFAULT_COLS]
+        dfs.append(gdf[matching_cols])
+
+df_all = pd.concat(dfs, ignore_index=True)
+```
+
+Now writing this to Azure...
+
+```python
+def get_container_client(
+    container_name: str = "projects", stage: Literal["prod", "dev"] = "dev"
+):
+    sas = DEV_BLOB_SAS if stage == "dev" else PROD_BLOB_SAS
+    container_url = (
+        f"https://imb0chd0{stage}.blob.core.windows.net/"
+        f"{container_name}?{sas}"
+    )
+    return ContainerClient.from_container_url(container_url)
+
+
+def upload_parquet_to_blob(
+    blob_name,
+    df,
+    stage: Literal["prod", "dev"] = "dev",
+    container_name: str = "projects",
+    **kwargs,
+):
+    upload_blob_data(
+        blob_name,
+        df.to_parquet(**kwargs),
+        stage=stage,
+        container_name=container_name,
+    )
+
+
+def upload_blob_data(
+    blob_name,
+    data,
+    stage: Literal["prod", "dev"] = "dev",
+    container_name: str = "projects",
+    content_type: str = None,
+):
+    container_client = get_container_client(
+        stage=stage, container_name=container_name
+    )
+
+    if content_type is None:
+        content_settings = ContentSettings(
+            content_type="application/octet-stream"
+        )
+    else:
+        content_settings = ContentSettings(content_type=content_type)
+
+    blob_client = container_client.get_blob_client(blob_name)
+    blob_client.upload_blob(
+        data, overwrite=True, content_settings=content_settings
+    )
+```
+
+```python
+upload_parquet_to_blob("admin_lookup.parquet", df_all, MODE, PROJECT_PREFIX)
+```