diff --git a/docs/sites/features/index.md b/docs/sites/features/index.md
index 31b0ca1..69423be 100644
--- a/docs/sites/features/index.md
+++ b/docs/sites/features/index.md
@@ -1,5 +1,6 @@
 # Features
 
 ```{toctree}
+column-metadata
 primary-keys
 ```
diff --git a/docs/sites/features/metadata.md b/docs/sites/features/metadata.md
new file mode 100644
index 0000000..b8db72b
--- /dev/null
+++ b/docs/sites/features/metadata.md
@@ -0,0 +1,29 @@
+# Column Metadata
+
+Sometimes it can be useful to attach user-provided metadata to columns of tables.
+The `metadata` parameter is available for all column types and accepts a dictionary of arbitrary objects.
+For instance, one may use the `metadata` parameter to mark a column as pseudonymized or provide other context-specific information.
+
+```python
+class UserSchema(dy.Schema):
+    id = dy.String(primary_key=True)
+    # Mark last name column as pseudonymized and (non-docstring) comment on it.
+    last_name = dy.String(metadata={
+        "pseudonymized": True,
+        "comment": "Pseudonymized using cryptographic hash function"
+    })
+    # Add information about database column type.
+    address = dy.String(metadata={"database-type": "VARCHAR(MAX)"})
+```
+
+```python
+>>> print(UserSchema.last_name.metadata)
+{'pseudonymized': True, 'comment': 'Pseudonymized using cryptographic hash function'}
+```
+
+Metadata are never read by `dataframely` and merely enable users to provide custom information
+in a structured way.
+
+```{note}
+User-provided metadata can be useful for code generation. For instance, one could specify metadata for columns, such as database column types or constraints, and override the built-in SQLAlchemy generation for more tailored SQL output.
+```