diff --git a/CHANGELOG.md b/CHANGELOG.md
index 38d07f07f..11c63c454 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -39,7 +39,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Fix missing prophet in docker images ([#767](https://github.com/tinkoff-ai/etna/pull/767))
 - Add `known_future` parameter to CLI ([#758](https://github.com/tinkoff-ai/etna/pull/758))
 - FutureWarning: The frame.append method is deprecated. Use pandas.concat instead ([#764](https://github.com/tinkoff-ai/etna/pull/764))
-- 
+- Correct ordering if multi-index in backtest ([#771](https://github.com/tinkoff-ai/etna/pull/771))
 - 
 - 
 - 
diff --git a/etna/pipeline/base.py b/etna/pipeline/base.py
index 89a446673..b37a23c68 100644
--- a/etna/pipeline/base.py
+++ b/etna/pipeline/base.py
@@ -457,6 +457,7 @@ def _get_backtest_forecasts(self) -> pd.DataFrame:
             forecast = forecast.join(fold_number_df)
             forecasts_list.append(forecast)
         forecasts = pd.concat(forecasts_list)
+        forecasts.sort_index(axis=1, inplace=True)
         return forecasts
 
     def _prepare_fold_masks(self, ts: TSDataset, masks: Union[int, List[FoldMask]], mode: str) -> List[FoldMask]:
diff --git a/tests/test_pipeline/conftest.py b/tests/test_pipeline/conftest.py
index 7fdd6f832..e3dbb3464 100644
--- a/tests/test_pipeline/conftest.py
+++ b/tests/test_pipeline/conftest.py
@@ -162,13 +162,12 @@ def step_ts() -> Tuple[TSDataset, pd.DataFrame, pd.DataFrame]:
         target_forecast += [start_value + i * add_value] * horizon
         fold_number_forecast += [i] * horizon
     forecast_df = pd.DataFrame(
-        {"target": target_forecast, "fold_number": fold_number_forecast},
+        {"fold_number": fold_number_forecast, "target": target_forecast},
         index=timestamp_forecast,
     )
     forecast_df.columns = pd.MultiIndex.from_product(
-        [[segment], ["target", "fold_number"]], names=("segment", "feature")
+        [[segment], ["fold_number", "target"]], names=("segment", "feature")
     )
-
     return ts, metrics_df, forecast_df
 
 
diff --git a/tests/test_pipeline/test_pipeline.py b/tests/test_pipeline/test_pipeline.py
index 104e68951..695a26418 100644
--- a/tests/test_pipeline/test_pipeline.py
+++ b/tests/test_pipeline/test_pipeline.py
@@ -573,3 +573,10 @@ def test_backtest_nans_at_beginning_with_mask(ts_name, request):
         metrics=[MAE()],
         n_folds=[mask],
     )
+
+
+def test_forecast_backtest_correct_ordering(step_ts: TSDataset):
+    ts, _, expected_forecast_df = step_ts
+    pipeline = Pipeline(model=NaiveModel(), horizon=5)
+    _, forecast_df, _ = pipeline.backtest(ts=ts, metrics=[MAE()], n_folds=3)
+    assert np.all(forecast_df.values == expected_forecast_df.values)