diff --git a/backend/media_plan_processor.py b/backend/media_plan_processor.py
index d0930e1..976c833 100644
--- a/backend/media_plan_processor.py
+++ b/backend/media_plan_processor.py
@@ -89,7 +89,9 @@ def parse_media_plan(excel_path: str) -> Dict:
     """
     import openpyxl
 
-    wb = openpyxl.load_workbook(excel_path, data_only=True)
+    # read_only=True skips pivot-cache deserialization, which hangs this loader
+    # on workbooks with pivot tables (Amazon media plans use them extensively).
+    wb = openpyxl.load_workbook(excel_path, data_only=True, read_only=True)
     all_assets = []
     channel_counts = {}
 
@@ -106,7 +108,9 @@ def parse_media_plan(excel_path: str) -> Dict:
                 continue
 
         ws = wb[sheet_name]
-        if ws.max_row < 2:
+        # In read_only mode max_row may be None until dimensions are read; treat
+        # that as "has data" and let iter_rows decide.
+        if ws.max_row is not None and ws.max_row < 2:
             continue
 
         # Get header row (row 1)