diff --git a/dags/extract_manual_actors_updates.py b/dags/extract_manual_actors_updates.py
new file mode 100644
index 0000000..981a25e
--- /dev/null
+++ b/dags/extract_manual_actors_updates.py
@@ -0,0 +1,180 @@
+from datetime import datetime, timedelta
+from airflow import DAG
+from airflow.operators.python import PythonOperator
+from airflow.providers.postgres.hooks.postgres import PostgresHook
+
+from utils.utils import load_table, normalize_nom, normalize_url, normalize_phone_number, normalize_email, \
+    find_differences, save_to_database
+
+
+def read_data_from_postgres(**kwargs):
+    table_name = kwargs["table_name"]
+    pg_hook = PostgresHook(postgres_conn_id='lvao-preprod')
+    engine = pg_hook.get_sqlalchemy_engine()
+    df = load_table(table_name, engine)
+    return df
+
+
+def normalize_and_find_differences(**kwargs):
+    ti = kwargs['ti']
+    df_act = ti.xcom_pull(task_ids='load_imported_actors')
+    df_rev_act = ti.xcom_pull(task_ids='load_imported_revision_actors')
+
+    columns_to_exclude = ["identifiant_unique", "statut", "cree_le", "modifie_le"]
+    normalization_map = {
+        "nom": normalize_nom,
+        "nom_commercial": normalize_nom,
+        "ville": normalize_nom,
+        "url": normalize_url,
+        "adresse": normalize_nom,
+        "adresse_complement": normalize_nom,
+        "email": normalize_email,
+        "telephone": normalize_phone_number,
+    }
+
+    df_differences = find_differences(df_act, df_rev_act, columns_to_exclude, normalization_map)
+
+    return df_differences
+
+
+def save_results_to_database(**kwargs):
+    df_cleaned = kwargs['ti'].xcom_pull(task_ids='normalize_and_find_differences')
+    pg_hook = PostgresHook(postgres_conn_id='lvao-preprod')
+    engine = pg_hook.get_sqlalchemy_engine()
+    save_to_database(df_cleaned, "lvao_manual_actors_updates", engine)
+
+
+default_args = {
+    'owner': 'airflow',
+    'depends_on_past': False,
+    'start_date': datetime(2024, 2, 7),
+    'email_on_failure': False,
+    'email_on_retry': False,
+    'retries': 1,
+    'retry_delay': timedelta(minutes=5),
+}
+
+dag = DAG(
+    'manual_actor_updates',
+    default_args=default_args,
+    description='DAG for manually updated LVAO actors data',
+    schedule_interval=None,
+)
+
+t1 = PythonOperator(
+    task_id='load_imported_actors',
+    python_callable=read_data_from_postgres,
+    op_kwargs={"table_name": "qfdmo_acteur"},
+    dag=dag,
+)
+
+t2 = PythonOperator(
+    task_id='load_imported_revision_actors',
+    python_callable=read_data_from_postgres,
+    op_kwargs={"table_name": "qfdmo_revisionacteur"},
+    dag=dag,
+)
+
+t3 = PythonOperator(
+    task_id='normalize_and_find_differences',
+    python_callable=normalize_and_find_differences,
+    provide_context=True,
+    dag=dag,
+)
+
+t4 = PythonOperator(
+    task_id='save_results_to_database',
+    python_callable=save_results_to_database,
+    provide_context=True,
+    dag=dag,
+)
+
+[t1, t2] >> t3 >> t4
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
diff --git a/utils/utils.py b/utils/utils.py
index 0670744..962c362 100644
--- a/utils/utils.py
+++ b/utils/utils.py
@@ -59,7 +59,7 @@ def apply_normalization(df, normalization_map):
     """
     # print(df['adresse'])
 
-    #df = normalize_address(df)
+    # df = normalize_address(df)
     #   print(df['adresse'])
     for column, normalize_func in normalization_map.items():
         if column in df.columns:
@@ -183,3 +183,62 @@ def normalize_url(url):
     except Exception as e:
         print(f"Error parsing URL {url}: {e}")
         return None
+
+
+def find_differences(
+        df_act, df_rev_act, columns_to_exclude, normalization_map
+):
+    # Join the DataFrames on 'identifiant_unique'
+    df_merged = pd.merge(
+        df_act, df_rev_act, on="identifiant_unique", suffixes=("_act", "_rev_act")
+    )
+    df_merged["code_postal_act"] = pd.to_numeric(
+        df_merged["code_postal_act"], errors="coerce"
+    ).astype(pd.Int64Dtype())
+    df_merged["code_postal_rev_act"] = pd.to_numeric(
+        df_merged["code_postal_rev_act"], errors="coerce"
+    ).astype(pd.Int64Dtype())
+
+    suffixes = ["_act", "_rev_act"]
+
+    # Apply the functions to the respective columns
+    for base_col, func in normalization_map.items():
+        for suffix in suffixes:
+            col = base_col + suffix
+            if col in df_merged.columns:
+                df_merged[col] = df_merged[col].apply(func)
+            else:
+                print(f"Column {col} not found in DataFrame.")
+    # Initialize a DataFrame to hold the differences
+    df_differences = pd.DataFrame()
+    df_differences["identifiant_unique"] = df_merged["identifiant_unique"]
+
+    columns_to_exclude = columns_to_exclude
+
+    for col in df_act.columns:
+        if col not in columns_to_exclude:
+            # Define the column names for act and rev_act
+            col_act = col + "_act"
+            col_rev_act = col + "_rev_act"
+
+            # Create masks for non-empty and differing values
+            mask_non_empty = (
+                    df_merged[col_rev_act].notnull()
+                    & (df_merged[col_act] != "")
+                    & (df_merged[col_rev_act] != "")
+            )
+            mask_different = df_merged[col_act] != df_merged[col_rev_act]
+
+            # Apply masks and add to df_differences
+            df_differences[col] = df_merged[col_rev_act].where(
+                mask_non_empty & mask_different, None
+            )
+
+    # Remove rows where all elements are None (no differences found)
+    df_differences = df_differences.dropna(
+        how="all",
+        subset=[
+            col for col in df_differences.columns if col != "identifiant_unique"
+        ],
+    )
+    return df_differences