diff --git a/application/cratedb-toolkit/requirements.txt b/application/cratedb-toolkit/requirements.txt
index 7d1a9c9a..48e247e8 100644
--- a/application/cratedb-toolkit/requirements.txt
+++ b/application/cratedb-toolkit/requirements.txt
@@ -1 +1 @@
-cratedb-toolkit[influxdb,mongodb]==0.0.23
+cratedb-toolkit[influxdb,mongodb]==0.0.24
diff --git a/application/cratedb-toolkit/test_io.py b/application/cratedb-toolkit/test_io.py
index f43e0444..efb727dc 100644
--- a/application/cratedb-toolkit/test_io.py
+++ b/application/cratedb-toolkit/test_io.py
@@ -109,7 +109,7 @@ def test_ctk_load_table_mongodb_json(drop_testing_tables):
     table_cardinalities = {
         "books": 431,
         "city_inspections": 81047,
-        "companies": 2537,
+        "companies": 18801,
         "countries-big": 21640,
         "countries-small": 248,
         "covers": 5071,
@@ -138,7 +138,8 @@ def test_ctk_load_table_mongodb_json(drop_testing_tables):
     command = f"""
 ctk load table \
     "file+bson://{datasets_path}/*.json?batch-size=2500" \
-    --cratedb-sqlalchemy-url="crate://localhost:4200/from-mongodb"
+    --cratedb-sqlalchemy-url="crate://localhost:4200/from-mongodb" \
+    --transformation=zyp-mongodb-json-files.yaml
 """
     print(f"Invoking CTK: {command}", file=sys.stderr)
     subprocess.check_call(shlex.split(command))
diff --git a/application/cratedb-toolkit/zyp-mongodb-json-files.yaml b/application/cratedb-toolkit/zyp-mongodb-json-files.yaml
new file mode 100644
index 00000000..31605164
--- /dev/null
+++ b/application/cratedb-toolkit/zyp-mongodb-json-files.yaml
@@ -0,0 +1,48 @@
+# A Zyp Transformation [1] file to support importing datasets
+# from mongodb-json-files [2] into CrateDB [3].
+#
+# [1] https://commons-codec.readthedocs.io/zyp/
+# [2] https://github.com/ozlerhakan/mongodb-json-files
+# [3] https://cratedb.com/docs/guide/feature/
+
+# Because CrateDB can not store nested arrays into OBJECT(DYNAMIC) columns,
+# this file defines a corresponding transformation to work around the problem.
+#
+# The workaround applied here is to just exclude/omit relevant `available_sizes`
+# elements completely. Converting them right can be implemented on behalf of a
+# later iteration.
+#
+# "image": {
+#   "available_sizes": [
+#     [
+#       [
+#         150,
+#         99
+#       ],
+#       "assets/images/resized/0001/3896/13896v3-max-150x150.jpg"
+#     ],
+#   ]
+#
+# A possible representation could be:
+#
+# "image": {
+#   "available_sizes": [
+#     {
+#       "path": "assets/images/resized/0001/3896/13896v3-max-150x150.jpg",
+#       "size": {"width": 150, "height": 99},
+#     }
+#   ]
+# }
+---
+
+meta:
+  type: zyp-project
+  version: 1
+collections:
+- address:
+    container: datasets
+    name: companies
+  pre:
+    rules:
+    - expression: .[] |= del(.image.available_sizes, .screenshots[].available_sizes)
+      type: jq