diff --git a/quotaclimat/data_processing/mediatree/api_import.py b/quotaclimat/data_processing/mediatree/api_import.py
index a13f93e9..d8a95545 100644
--- a/quotaclimat/data_processing/mediatree/api_import.py
+++ b/quotaclimat/data_processing/mediatree/api_import.py
@@ -201,7 +201,7 @@ def extract_api_sub(
 
         if(df is not None):
             df = filter_and_tag_by_theme(df)
-            df["id"] = add_primary_key(df)
+            df["id"] = df.apply(lambda x: add_primary_key(x), axis=1)
             return df
         else:
             None
diff --git a/quotaclimat/data_processing/mediatree/detect_keywords.py b/quotaclimat/data_processing/mediatree/detect_keywords.py
index fd85104e..a5baa28c 100644
--- a/quotaclimat/data_processing/mediatree/detect_keywords.py
+++ b/quotaclimat/data_processing/mediatree/detect_keywords.py
@@ -262,15 +262,14 @@ def filter_and_tag_by_theme(df: pd.DataFrame) -> pd.DataFrame :
 
             return df
 
-def add_primary_key(df):
-    logging.info("Adding primary key to save to PG and have idempotent result")
+def add_primary_key(row):
+    logging.info(f"Adding primary key to save to PG and have idempotent results")
     try:
-        return (
-            df["start"].astype(str) + df["channel_name"]
-        ).apply(get_consistent_hash)
+        return get_consistent_hash(str(row["start"].timestamp()) + row["channel_name"])
+
     except (Exception) as error:
-        logging.error(f"{error} with df {df.head()}")
-        return get_consistent_hash("empty") #  TODO improve - should be a None ?
+        logging.error(f"{error} with df {row}")
+        raise Exception
 
 def filter_indirect_words(keywords_with_timestamp: List[dict]) -> List[dict]:
     return list(filter(lambda kw: indirectes not in kw['theme'], keywords_with_timestamp))
diff --git a/test/sitemap/test_main_import_api.py b/test/sitemap/test_main_import_api.py
index df2c62fa..5d02332d 100644
--- a/test/sitemap/test_main_import_api.py
+++ b/test/sitemap/test_main_import_api.py
@@ -23,7 +23,7 @@ def test_main_api_import():
         start_time = t.time()
         df = parse_reponse_subtitle(json_response)
         df = filter_and_tag_by_theme(df)
-        df["id"] = add_primary_key(df)
+        df["id"] = df.apply(lambda x: add_primary_key(x), axis=1)
         end_time = t.time()
         logging.info(f"Elapsed time for api import {end_time - start_time}")
         # must df._to_pandas() because to_sql does not handle modin dataframe