diff --git a/src/trousse/convert_to_mixed_type.py b/src/trousse/convert_to_mixed_type.py
index 698ecee..7d36b57 100644
--- a/src/trousse/convert_to_mixed_type.py
+++ b/src/trousse/convert_to_mixed_type.py
@@ -2,15 +2,15 @@
 import pandas as pd
 
 
-class _DfConvertToMixedType:
+class _ConvertDfToMixedType:
     """
     Convert values from "object"-typed ``column`` column to appropriate format.
 
     When pandas package reads from CSV file, the columns that are not completely
     consistent with a single type are stored with dtype = "object" and every value
     is converted to "string".
-    This FeatureOperation subclass convert the string values to numeric, boolean
-    or datetime values where possible.
+    This class converts the string values of a column in a pandas DataFrame
+    to numeric, boolean or datetime values where possible.
     The transformed column will still have dtype="object" but the inferred type will
     be "mixed" which allows a correct column categorization by Dataset class.
     By default the converted column overwrites the related original column.
@@ -193,16 +193,27 @@ def _convert_to_datetime_mixed_types(self, col_serie: pd.Series) -> None:
 
     def _set_converted_col_dtype(self, col_serie: pd.Series) -> pd.Series:
         """
-        Set the new dtype to ``col_serie`` after conversion
+        Set the new dtype to ``col_serie`` after conversion.
 
-        This method sets the new col_dtype to ``col_serie`` column
-        TODO: Complete docs
+        This method updates, if possible, the ``col_serie`` column dtype.
+        Particularly, if each value of the column has been consistently interpreted
+        with a single type, the column will be converted to that dtype
+        (and NaNs will be converted coherently with the new dtype).
+        On the other hand, if column values are interpreted with multiple types,
+        the column will maintain the dtype="object" (and the column values will
+        have multiple types).
 
         Parameters
         ----------
         col_serie : pd.Series
             Series containing the values that will be analyzed. It will not be
             modified inplace.
+
+        Returns
+        -------
+        pd.Series
+            Column with the same values as ``col_serie`` and the dtype set
+            according to the value types.
         """
         if self._col_dtype is None:
             # If the _col_dtype is not unique and consistent, convert the column
diff --git a/src/trousse/dataset.py b/src/trousse/dataset.py
index e6aba76..6c79418 100644
--- a/src/trousse/dataset.py
+++ b/src/trousse/dataset.py
@@ -11,9 +11,9 @@
 import pandas as pd
 from joblib import Parallel, delayed
 
-from .convert_to_mixed_type import _DfConvertToMixedType
+from .convert_to_mixed_type import _ConvertDfToMixedType
 from .exceptions import MultipleObjectsInFileError, NotShelveFileError
-from .feature_operations import ConvertToMixedType, FeatureOperation
+from .feature_operations import FeatureOperation
 from .operations_list import OperationsList
 from .settings import CATEG_COL_THRESHOLD
 from .util import lazy_property
@@ -555,7 +555,7 @@ def _data_to_mixed_types(df: pd.DataFrame):
         """
         str_cols = df.select_dtypes(include="object").columns
         for col in str_cols:
-            mixedtype_converter = _DfConvertToMixedType(column=col)
+            mixedtype_converter = _ConvertDfToMixedType(column=col)
             df = mixedtype_converter(df)
 
         return df
diff --git a/src/trousse/feature_operations.py b/src/trousse/feature_operations.py
index 7e1d04a..9e5f69c 100644
--- a/src/trousse/feature_operations.py
+++ b/src/trousse/feature_operations.py
@@ -3,7 +3,7 @@
 from abc import ABC, abstractmethod
 from typing import Any, List, Mapping
 
-from .convert_to_mixed_type import _DfConvertToMixedType
+from .convert_to_mixed_type import _ConvertDfToMixedType
 from .util import is_sequence_and_not_str
 
 if typing.TYPE_CHECKING:
@@ -302,8 +302,11 @@ class ConvertToMixedType(FeatureOperation):
     consistent with a single type are stored with dtype = "object" and every value
     is converted to "string".
     This FeatureOperation subclass convert the string values to numeric, boolean
-    or datetime values where possible.
-    The transformed column will still have dtype="object" but the inferred type will
+    or datetime values where possible. When each value can consistently be
+    interpreted with a single type, the column will be converted to that dtype
+    (and NaNs will be converted appropriately).
+    On the other hand, if column values are interpreted with multiple types,
+    the column will maintain the dtype="object" but the inferred type will
     be "mixed" which allows a correct column categorization by Dataset class.
     By default the converted column overwrites the related original column.
     To store the result of conversion in another column, ``derived_columns``
@@ -377,7 +380,8 @@ def __eq__(self, other: Any) -> bool:
             return False
 
     def _apply(self, dataset: "Dataset") -> "Dataset":
-        """Apply ReplaceStrings operation on a new Dataset instance and return it.
+        """
+        Apply ConvertToMixedType operation on a new Dataset instance and return it.
 
         Parameters
         ----------
@@ -391,7 +395,7 @@ def _apply(self, dataset: "Dataset") -> "Dataset":
         """
         dataset = copy.deepcopy(dataset)
 
-        mixedtype_converter = _DfConvertToMixedType(
+        mixedtype_converter = _ConvertDfToMixedType(
             column=self.columns[0], derived_column=self.derived_columns[0]
         )
         dataset.data = mixedtype_converter(dataset.data)