diff --git a/src/urban_meal_delivery/forecasts/timify.py b/src/urban_meal_delivery/forecasts/timify.py
index 6906d24..0220a58 100644
--- a/src/urban_meal_delivery/forecasts/timify.py
+++ b/src/urban_meal_delivery/forecasts/timify.py
@@ -149,7 +149,7 @@ class OrderHistory:
 
     def make_horizontal_time_series(  # noqa:WPS210
         self, pixel_id: int, predict_at: dt.datetime, train_horizon: int,
-    ) -> Tuple[pd.DataFrame, int, int]:
+    ) -> Tuple[pd.Series, int, pd.Series]:
         """Slice a horizontal time series out of the `.totals`.
 
         Create a time series covering `train_horizon` weeks that can be used
@@ -203,19 +203,20 @@ class OrderHistory:
         frequency = 7
 
         # Take only the counts at the `predict_at` time.
-        training_df = intra_pixel.loc[
-            first_start_at : last_start_at : self._n_daily_time_steps  # type: ignore
+        training_ts = intra_pixel.loc[
+            first_start_at : last_start_at : self._n_daily_time_steps,  # type: ignore
+            'total_orders',
         ]
-        if len(training_df) != frequency * train_horizon:
+        if len(training_ts) != frequency * train_horizon:
             raise RuntimeError('Not enough historic data for `predict_at`')
 
-        actual_df = intra_pixel.loc[[predict_at]]
+        actuals_ts = intra_pixel.loc[[predict_at], 'total_orders']
 
-        return training_df, frequency, actual_df
+        return training_ts, frequency, actuals_ts
 
     def make_vertical_time_series(  # noqa:WPS210
         self, pixel_id: int, predict_day: dt.date, train_horizon: int,
-    ) -> Tuple[pd.DataFrame, int, pd.DataFrame]:
+    ) -> Tuple[pd.Series, int, pd.Series]:
         """Slice a vertical time series out of the `.totals`.
 
         Create a time series covering `train_horizon` weeks that can be used
@@ -268,10 +269,11 @@ class OrderHistory:
         frequency = 7 * self._n_daily_time_steps
 
         # Take all the counts between `first_train_day` and `last_train_day`.
-        training_df = intra_pixel.loc[
-            first_start_at:last_start_at  # type: ignore
+        training_ts = intra_pixel.loc[
+            first_start_at:last_start_at,  # type: ignore
+            'total_orders',
         ]
-        if len(training_df) != frequency * train_horizon:
+        if len(training_ts) != frequency * train_horizon:
             raise RuntimeError('Not enough historic data for `predict_day`')
 
         first_prediction_at = dt.datetime(
@@ -289,15 +291,16 @@ class OrderHistory:
             0,
         ) - dt.timedelta(minutes=self._time_step)
 
-        actuals_df = intra_pixel.loc[
-            first_prediction_at:last_prediction_at  # type: ignore
+        actuals_ts = intra_pixel.loc[
+            first_prediction_at:last_prediction_at,  # type: ignore
+            'total_orders',
         ]
 
-        return training_df, frequency, actuals_df
+        return training_ts, frequency, actuals_ts
 
     def make_real_time_time_series(  # noqa:WPS210
         self, pixel_id: int, predict_at: dt.datetime, train_horizon: int,
-    ) -> Tuple[pd.DataFrame, int, int]:
+    ) -> Tuple[pd.Series, int, pd.Series]:
         """Slice a vertical real-time time series out of the `.totals`.
 
         Create a time series covering `train_horizon` weeks that can be used
@@ -361,8 +364,9 @@ class OrderHistory:
 
         # Take all the counts between `first_train_day` and `last_train_day`,
         # including the ones on the `predict_at` day prior to `predict_at`.
-        training_df = intra_pixel.loc[
-            first_start_at:last_start_at  # type: ignore
+        training_ts = intra_pixel.loc[
+            first_start_at:last_start_at,  # type: ignore
+            'total_orders',
         ]
         n_time_steps_on_predict_day = (
             (
@@ -378,9 +382,9 @@ class OrderHistory:
             // 60  # -> minutes
             // self._time_step
         )
-        if len(training_df) != frequency * train_horizon + n_time_steps_on_predict_day:
+        if len(training_ts) != frequency * train_horizon + n_time_steps_on_predict_day:
             raise RuntimeError('Not enough historic data for `predict_day`')
 
-        actual_df = intra_pixel.loc[[predict_at]]
+        actuals_ts = intra_pixel.loc[[predict_at], 'total_orders']
 
-        return training_df, frequency, actual_df
+        return training_ts, frequency, actuals_ts
diff --git a/tests/forecasts/timify/test_make_time_series.py b/tests/forecasts/timify/test_make_time_series.py
index 98e2c34..4dc187d 100644
--- a/tests/forecasts/timify/test_make_time_series.py
+++ b/tests/forecasts/timify/test_make_time_series.py
@@ -101,22 +101,22 @@ class TestMakeHorizontalTimeSeries:
             )
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
-    def test_time_series_are_dataframes(
+    def test_time_series_are_series(
         self, order_history, good_pixel_id, good_predict_at, train_horizon,
     ):
-        """The time series come in a one-column `pd.DataFrame`."""
+        """The time series come as a `pd.Series`."""
         result = order_history.make_horizontal_time_series(
             pixel_id=good_pixel_id,
             predict_at=good_predict_at,
             train_horizon=train_horizon,
         )
 
-        training_df, _, actual_df = result
+        training_ts, _, actuals_ts = result
 
-        assert isinstance(training_df, pd.DataFrame)
-        assert training_df.columns == ['total_orders']
-        assert isinstance(actual_df, pd.DataFrame)
-        assert actual_df.columns == ['total_orders']
+        assert isinstance(training_ts, pd.Series)
+        assert training_ts.name == 'total_orders'
+        assert isinstance(actuals_ts, pd.Series)
+        assert actuals_ts.name == 'total_orders'
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
     def test_time_series_have_correct_length(
@@ -132,10 +132,10 @@ class TestMakeHorizontalTimeSeries:
             train_horizon=train_horizon,
         )
 
-        training_df, _, actual_df = result
+        training_ts, _, actuals_ts = result
 
-        assert len(training_df) == 7 * train_horizon
-        assert len(actual_df) == 1
+        assert len(training_ts) == 7 * train_horizon
+        assert len(actuals_ts) == 1
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
     def test_frequency_is_number_of_weekdays(
@@ -194,22 +194,22 @@ class TestMakeVerticalTimeSeries:
             )
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
-    def test_time_series_are_dataframes(
+    def test_time_series_are_series(
         self, order_history, good_pixel_id, good_predict_at, train_horizon,
     ):
-        """The time series come in a one-column `pd.DataFrame`."""
+        """The time series come as `pd.Series`."""
         result = order_history.make_vertical_time_series(
             pixel_id=good_pixel_id,
             predict_day=good_predict_at.date(),
             train_horizon=train_horizon,
         )
 
-        training_df, _, actual_df = result
+        training_ts, _, actuals_ts = result
 
-        assert isinstance(training_df, pd.DataFrame)
-        assert training_df.columns == ['total_orders']
-        assert isinstance(actual_df, pd.DataFrame)
-        assert actual_df.columns == ['total_orders']
+        assert isinstance(training_ts, pd.Series)
+        assert training_ts.name == 'total_orders'
+        assert isinstance(actuals_ts, pd.Series)
+        assert actuals_ts.name == 'total_orders'
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
     def test_time_series_have_correct_length(
@@ -229,7 +229,7 @@ class TestMakeVerticalTimeSeries:
             train_horizon=train_horizon,
         )
 
-        training_df, _, actual_df = result
+        training_ts, _, actuals_ts = result
 
         n_daily_time_steps = (
             60
@@ -237,8 +237,8 @@ class TestMakeVerticalTimeSeries:
             // test_config.LONG_TIME_STEP
         )
 
-        assert len(training_df) == 7 * n_daily_time_steps * train_horizon
-        assert len(actual_df) == n_daily_time_steps
+        assert len(training_ts) == 7 * n_daily_time_steps * train_horizon
+        assert len(actuals_ts) == n_daily_time_steps
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
     def test_frequency_is_number_number_of_weekly_time_steps(
@@ -305,22 +305,22 @@ class TestMakeRealTimeTimeSeries:
             )
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
-    def test_time_series_are_dataframes(
+    def test_time_series_are_series(
         self, order_history, good_pixel_id, good_predict_at, train_horizon,
     ):
-        """The time series come in a one-column `pd.DataFrame`."""
+        """The time series come as `pd.Series`."""
         result = order_history.make_real_time_time_series(
             pixel_id=good_pixel_id,
             predict_at=good_predict_at,
             train_horizon=train_horizon,
         )
 
-        training_df, _, actual_df = result
+        training_ts, _, actuals_ts = result
 
-        assert isinstance(training_df, pd.DataFrame)
-        assert training_df.columns == ['total_orders']
-        assert isinstance(actual_df, pd.DataFrame)
-        assert actual_df.columns == ['total_orders']
+        assert isinstance(training_ts, pd.Series)
+        assert training_ts.name == 'total_orders'
+        assert isinstance(actuals_ts, pd.Series)
+        assert actuals_ts.name == 'total_orders'
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
     def test_time_series_have_correct_length1(
@@ -345,7 +345,7 @@ class TestMakeRealTimeTimeSeries:
             pixel_id=good_pixel_id, predict_at=predict_at, train_horizon=train_horizon,
         )
 
-        training_df, _, actual_df = result
+        training_ts, _, actuals_ts = result
 
         n_daily_time_steps = (
             60
@@ -353,8 +353,8 @@ class TestMakeRealTimeTimeSeries:
             // test_config.LONG_TIME_STEP
         )
 
-        assert len(training_df) == 7 * n_daily_time_steps * train_horizon
-        assert len(actual_df) == 1
+        assert len(training_ts) == 7 * n_daily_time_steps * train_horizon
+        assert len(actuals_ts) == 1
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
     def test_time_series_have_correct_length2(
@@ -378,7 +378,7 @@ class TestMakeRealTimeTimeSeries:
             train_horizon=train_horizon,
         )
 
-        training_df, _, actual_df = result
+        training_ts, _, actuals_ts = result
 
         n_daily_time_steps = (
             60
@@ -390,10 +390,10 @@ class TestMakeRealTimeTimeSeries:
         )
 
         assert (
-            len(training_df)
+            len(training_ts)
             == 7 * n_daily_time_steps * train_horizon + n_time_steps_before
         )
-        assert len(actual_df) == 1
+        assert len(actuals_ts) == 1
 
     @pytest.mark.parametrize('train_horizon', test_config.TRAIN_HORIZONS)
     def test_frequency_is_number_number_of_weekly_time_steps(