diff --git a/docs/source/release_notes.rst b/docs/source/release_notes.rst
index 14d30a0f37..b1e19aab6f 100644
--- a/docs/source/release_notes.rst
+++ b/docs/source/release_notes.rst
@@ -8,6 +8,7 @@ Release Notes
         * Added `explain_predictions` and `explain_predictions_best_worst` for explaining multiple predictions with SHAP :pr:`1016`
         * Added new LSA component for text featurization :pr:`1022`
         * Added guide on installing with conda :pr:`1041`
+        * Added `percent_better_than_baseline` to Automl search rankings and full rankings table :pr:`1050`
     * Fixes
         * Updated TextFeaturizer component to no longer require an internet connection to run :pr:`1022`
         * Fixed non-deterministic element of TextFeaturizer transformations :pr:`1022`
diff --git a/evalml/automl/automl_search.py b/evalml/automl/automl_search.py
index 7c2476c24f..0131967401 100644
--- a/evalml/automl/automl_search.py
+++ b/evalml/automl/automl_search.py
@@ -197,6 +197,7 @@ def __init__(self,
         self.allowed_model_families = allowed_model_families
         self._automl_algorithm = None
         self._start = None
+        self._baseline_score = None
 
         self._validate_problem_type()
 
@@ -528,7 +529,7 @@ def _add_baseline_pipelines(self, X, y):
                                 self._start)
 
                 baseline_results = self._compute_cv_scores(baseline, X, y)
-                self.baseline_score = baseline_results["cv_score_mean"]
+                self._baseline_score = baseline_results["cv_score_mean"]
                 self._add_result(trained_pipeline=baseline,
                                  parameters=baseline.parameters,
                                  training_time=baseline_results['training_time'],
@@ -615,7 +616,7 @@ def _compute_cv_scores(self, pipeline, X, y):
 
     def _add_result(self, trained_pipeline, parameters, training_time, cv_data, cv_scores):
         cv_score = cv_scores.mean()
-        percent_better = np.round(self.objective.calculate_percent_difference(cv_score, self.baseline_score), 2)
+        percent_better = np.round(self.objective.calculate_percent_difference(cv_score, self._baseline_score), 2)
         # calculate high_variance_cv
         # if the coefficient of variance is greater than .2
         with warnings.catch_warnings():
diff --git a/evalml/tests/automl_tests/test_automl.py b/evalml/tests/automl_tests/test_automl.py
index 1fa9f00041..cb9581869b 100644
--- a/evalml/tests/automl_tests/test_automl.py
+++ b/evalml/tests/automl_tests/test_automl.py
@@ -55,7 +55,8 @@ def test_search_results(X_y_regression, X_y_binary, X_y_multi, automl_type):
     assert automl.results['search_order'] == [0, 1]
     assert len(automl.results['pipeline_results']) == 2
     for pipeline_id, results in automl.results['pipeline_results'].items():
-        assert results.keys() == {'id', 'pipeline_name', 'pipeline_class', 'pipeline_summary', 'parameters', 'score', 'high_variance_cv', 'training_time', 'cv_data'}
+        assert results.keys() == {'id', 'pipeline_name', 'pipeline_class', 'pipeline_summary', 'parameters', 'score', 'high_variance_cv', 'training_time',
+                                  'cv_data', 'percent_better_than_baseline'}
         assert results['id'] == pipeline_id
         assert isinstance(results['pipeline_name'], str)
         assert issubclass(results['pipeline_class'], expected_pipeline_class)
@@ -77,11 +78,11 @@ def test_search_results(X_y_regression, X_y_binary, X_y_multi, automl_type):
     assert isinstance(automl.rankings, pd.DataFrame)
     assert isinstance(automl.full_rankings, pd.DataFrame)
     assert np.all(automl.rankings.dtypes == pd.Series(
-        [np.dtype('int64'), np.dtype('O'), np.dtype('float64'), np.dtype('bool'), np.dtype('O')],
-        index=['id', 'pipeline_name', 'score', 'high_variance_cv', 'parameters']))
+        [np.dtype('int64'), np.dtype('O'), np.dtype('float64'), np.dtype('float64'), np.dtype('bool'), np.dtype('O')],
+        index=['id', 'pipeline_name', 'score', 'percent_better_than_baseline', 'high_variance_cv', 'parameters']))
     assert np.all(automl.full_rankings.dtypes == pd.Series(
-        [np.dtype('int64'), np.dtype('O'), np.dtype('float64'), np.dtype('bool'), np.dtype('O')],
-        index=['id', 'pipeline_name', 'score', 'high_variance_cv', 'parameters']))
+        [np.dtype('int64'), np.dtype('O'), np.dtype('float64'), np.dtype('float64'), np.dtype('bool'), np.dtype('O')],
+        index=['id', 'pipeline_name', 'score', 'percent_better_than_baseline', 'high_variance_cv', 'parameters']))
 
 
 @pytest.mark.parametrize("automl_type", [ProblemTypes.BINARY, ProblemTypes.MULTICLASS, ProblemTypes.REGRESSION])
@@ -729,7 +730,8 @@ def test_no_search():
     assert isinstance(automl.rankings, pd.DataFrame)
     assert isinstance(automl.full_rankings, pd.DataFrame)
 
-    df_columns = ["id", "pipeline_name", "score", "high_variance_cv", "parameters"]
+    df_columns = ["id", "pipeline_name", "score", "percent_better_than_baseline",
+                  "high_variance_cv", "parameters"]
     assert (automl.rankings.columns == df_columns).all()
     assert (automl.full_rankings.columns == df_columns).all()