diff --git a/src/caosadvancedtools/table_json_conversion/fill_xlsx.py b/src/caosadvancedtools/table_json_conversion/fill_xlsx.py
index 380a9d62ea85e6abf92194e7aaf8427d0eaef724..79c7bfea1925adec47cf74af94d99deaf4fabc06 100644
--- a/src/caosadvancedtools/table_json_conversion/fill_xlsx.py
+++ b/src/caosadvancedtools/table_json_conversion/fill_xlsx.py
@@ -65,4 +65,25 @@ def fill_template(template_path: str, json_path: str, result_path: str) -> None:
     ``template_path`` and store the result under ``result_path``.
     """
     template = load_workbook(template_path)
+    # For each top level key in the json we iterate the values (if it is an array). Those are the
+    # root elements that belong to a particular sheet.
+    #       After treating a root element, the row index for the corresponding sheet needs to be
+    #       increased
+    #       When we finished treating an object that goes into a lower ranked sheet (see below), we
+    #       increase the row index of that sheet.
+    #
+
+    # We can generate a hierarchy of sheets in the beginning (using the paths). The lower sheets
+    # are for objects referenced by objects in higher ranked sheets.
+    # We can detect the sheet corresponding to a root element by looking at the first path element:
+    # The first path element must be the root element every where.
+    # Suggestion:
+    # row indices: Dict[str, int] string is the sheet name
+    # sheet_hirarchy: List[Tuple[str]]  elements are sheet names
+    #
+    # Question:
+    # We can create an internal representation where we assign as sheet_names the same names that
+    # are used in table generator. Or should we create another special row that contains this
+    # somehow?
+
     template.save(result_path)