diff --git a/CHANGELOG.md b/CHANGELOG.md
index 7f59fea3..9cee8700 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -18,6 +18,10 @@ Keep it human-readable, your future self will thank you!
 
 - Add `variables_metadata` entry in the dataset metadata
 
+### Changed
+
+- Add `variables_metadata` entry in the dataset metadata
+
 ## [0.5.5](https://github.com/ecmwf/anemoi-datasets/compare/0.5.4...0.5.5) - 2024-10-04
 
 ### Changed
diff --git a/tests/create/data_sources.yaml b/tests/create/data_sources.yaml
deleted file mode 100644
index cf1d7735..00000000
--- a/tests/create/data_sources.yaml
+++ /dev/null
@@ -1,28 +0,0 @@
-common:
-  mars_request: &mars_request
-    expver: "0001"
-    class: ea
-    grid: 20./20.
-
-dates:
-  start: 2020-12-30 00:00:00
-  end: 2021-01-03 12:00:00
-  frequency: 12h
-
-data_sources:
-  my_predefined_source:
-    mars:
-        <<: *mars_request
-        param: [2t]
-        levtype: sfc
-        stream: oper
-        type: an
-
-input:
-  forcings:
-    template: ${data_sources.my_predefined_source}
-    param:
-      - cos_latitude
-
-statistics:
-  end: 2021
diff --git a/tests/create/test_create.py b/tests/create/test_create.py
index 2112cd78..257ab434 100755
--- a/tests/create/test_create.py
+++ b/tests/create/test_create.py
@@ -98,23 +98,22 @@ def compare_dot_zattrs(a, b, path, errors):
     if isinstance(a, dict):
         a_keys = list(a.keys())
         b_keys = list(b.keys())
-        for k in set(a_keys) & set(b_keys):
-            if k in [
-                "licence",
-                "total_number_of_files",
-            ]:
+        for k in set(a_keys) | set(b_keys):
+            if k not in a_keys:
+                errors.append(f"❌ {path}.{k} : missing key (only in reference)")
+                continue
+            if k not in b_keys:
+                errors.append(f"❌ {path}.{k} : additional key (missing in reference)")
                 continue
             if k in [
                 "timestamp",
                 "uuid",
                 "latest_write_timestamp",
-                "_create_yaml_config",
                 "history",
                 "provenance",
                 "provenance_load",
                 "description",
                 "config_path",
-                "dataset_status",
                 "total_size",
             ]:
                 if type(a[k]) is not type(b[k]):
@@ -217,8 +216,8 @@ def compare(self):
             raise AssertionError("Comparison failed")
 
         compare_datasets(self.ds_output, self.ds_reference)
-
         compare_statistics(self.ds_output, self.ds_reference)
+        # do not compare tendencies statistics yet, as we don't know yet if they should stay
 
 
 @pytest.mark.parametrize("name", NAMES)