diff --git a/skrub/_reporting/tests/test_table_report.py b/skrub/_reporting/tests/test_table_report.py
index f57b9335a..6268c7e71 100644
--- a/skrub/_reporting/tests/test_table_report.py
+++ b/skrub/_reporting/tests/test_table_report.py
@@ -7,13 +7,29 @@
 
 import numpy as np
 import pytest
+from sklearn.utils import Bunch
 
-from skrub import TableReport, ToDatetime, datasets
+from skrub import TableReport, ToDatetime
 from skrub import _dataframe as sbd
 from skrub._reporting._sample_table import make_table
 from skrub.conftest import polars_installed_without_pyarrow
 
 
+@pytest.fixture
+def simple_df(df_module):
+    return df_module.make_dataframe(
+        {
+            "A": [1, 2, 3, 4, 5],
+            "B": ["a", "b", "a", "b", "c"],
+        }
+    )
+
+
+@pytest.fixture
+def simple_series(df_module):
+    return df_module.make_column(name="A", values=[1, 2, 3, 4, 5])
+
+
 def get_report_id(html):
     return re.search(r'<skrub-table-report.*?id="report_([a-z0-9]+)"', html).group(1)
 
@@ -294,8 +310,8 @@ def test_minimal_mode(pd_module):
     assert 'id="column-associations-panel"' not in html
 
 
-def test_error_input_type():
-    df = datasets.fetch_employee_salaries()
+def test_error_input_type(simple_df, simple_series):
+    df = Bunch(X=simple_df, y=simple_series)
     with pytest.raises(TypeError):
         TableReport(df)
 
diff --git a/skrub/tests/test_config.py b/skrub/tests/test_config.py
index 04018006f..d263ffcf3 100644
--- a/skrub/tests/test_config.py
+++ b/skrub/tests/test_config.py
@@ -4,13 +4,28 @@
 from skrub import TableReport, config_context, get_config, set_config
 from skrub._config import _parse_env_bool
 from skrub._data_ops._evaluation import evaluate
-from skrub.datasets import fetch_employee_salaries
+from skrub.conftest import polars_installed_without_pyarrow
 
 
 def _use_table_report(obj):
     return "SkrubTableReport" in obj._repr_html_()
 
 
+@pytest.fixture
+def simple_df(df_module):
+    return df_module.make_dataframe(
+        {
+            "A": [1, 2, 3, 4, 5],
+            "B": ["a", "b", "a", "b", "c"],
+        }
+    )
+
+
+@pytest.fixture
+def simple_series(df_module):
+    return df_module.make_column(name="A", values=[1, 2, 3, 4, 5])
+
+
 def test_config_context():
     assert get_config() == {
         "use_table_report": False,
@@ -28,38 +43,39 @@ def test_config_context():
     assert get_config()["use_table_report"] is False
 
 
-def test_use_table_report_data_ops():
-    X = skrub.X(fetch_employee_salaries().X)
-
+def test_use_table_report_data_ops(simple_df):
+    X = skrub.X(simple_df)
     with config_context(use_table_report_data_ops=True):
         assert _use_table_report(X)
         with config_context(use_table_report_data_ops=False):
             assert not _use_table_report(X)
 
 
-def test_use_table_report():
-    X = fetch_employee_salaries().X
-    assert not _use_table_report(X)
+@polars_installed_without_pyarrow
+def test_use_table_report(simple_df):
+    assert not _use_table_report(simple_df)
     with config_context(use_table_report=True):
-        assert _use_table_report(X)
+        assert _use_table_report(simple_df)
         with config_context(use_table_report=False):
-            assert not _use_table_report(X)
+            assert not _use_table_report(simple_df)
 
 
-def test_max_plot_columns():
-    X = fetch_employee_salaries().X
-    report = TableReport(X)
+@polars_installed_without_pyarrow
+def test_max_plot_columns(simple_df):
+    report = TableReport(simple_df)
     assert report.max_association_columns == 30
     assert report.max_plot_columns == 30
 
     # Set default to 1
     with config_context(max_plot_columns=1):
-        report = TableReport(X)
+        report = TableReport(simple_df)
         assert report.max_association_columns == 30
         assert report.max_plot_columns == 1
 
         # Argument takes precedence over default configuration
-        report = TableReport(X, max_association_columns="all", max_plot_columns="all")
+        report = TableReport(
+            simple_df, max_association_columns="all", max_plot_columns="all"
+        )
         assert report.max_association_columns == "all"
         assert report.max_plot_columns == "all"
 
@@ -67,15 +83,14 @@ def test_max_plot_columns():
     # repr_html.
     with config_context(use_table_report=True):
         with config_context(max_plot_columns=3):
-            "Plotting was skipped" in X._repr_html_()
+            "Plotting was skipped" in simple_df._repr_html_()
 
 
-def test_enable_subsampling():
-    X = fetch_employee_salaries().X
-    dataop = skrub.X(X)
+def test_enable_subsampling(simple_df):
+    dataop = skrub.X(simple_df)
 
     # No subsampling by default with fit_transform mode
-    assert dataop.skb.subsample(n=3).skb.eval().shape[0] == X.shape[0]
+    assert dataop.skb.subsample(n=3).skb.eval().shape[0] == simple_df.shape[0]
     assert dataop.skb.subsample(n=3).skb.eval(keep_subsampling=True).shape[0] == 3
 
     # Force subsampling during fit_transform
@@ -87,24 +102,24 @@ def test_enable_subsampling():
 
     with config_context(enable_subsampling="disable"):
         assert (
-            evaluate(dataop.skb.subsample(n=3), mode="preview").shape[0] == X.shape[0]
+            evaluate(dataop.skb.subsample(n=3), mode="preview").shape[0]
+            == simple_df.shape[0]
         )
         with config_context(enable_subsampling="default"):
             assert evaluate(dataop.skb.subsample(n=3), mode="preview").shape[0] == 3
 
 
-def test_float_precision():
-    y = fetch_employee_salaries().y
-
+@polars_installed_without_pyarrow
+def test_float_precision(simple_series):
     # Default config: float_precision set to 3
-    report = TableReport(y)
+    report = TableReport(simple_series)
     mean = f"{report._summary['columns'][0]['mean']:#.3g}"
     html = report._repr_html_()
     assert mean in html
 
     # Float precision set to 2
     with config_context(float_precision=2):
-        report_2 = TableReport(y)
+        report_2 = TableReport(simple_series)
         mean_2 = f"{report_2._summary['columns'][0]['mean']:#.2g}"
         html_2 = report_2._repr_html_()
         assert mean_2 in html_2
@@ -131,25 +146,24 @@ def test_error(params):
         set_config(**params)
 
 
-def test_subsampling_seed():
-    X = fetch_employee_salaries().X
-    data_op = skrub.X(X)
+def test_subsampling_seed(simple_df):
+    data_op = skrub.X(simple_df)
 
     with config_context(subsampling_seed=0):
-        index = evaluate(
-            data_op.skb.subsample(n=3, how="random"), mode="preview"
-        ).index.tolist()
-        index_identical = evaluate(
+        col_a = evaluate(data_op.skb.subsample(n=3, how="random"), mode="preview")[
+            "A"
+        ].to_list()
+        col_a_identical = evaluate(
             data_op.skb.subsample(n=3, how="random"), mode="preview"
-        ).index.tolist()
+        )["A"].to_list()
 
     with config_context(subsampling_seed=1):
-        index_different = evaluate(
+        col_a_different = evaluate(
             data_op.skb.subsample(n=3, how="random"), mode="preview"
-        ).index.tolist()
+        )["A"].to_list()
 
-    assert index == index_identical
-    assert index != index_different
+    assert col_a == col_a_identical
+    assert col_a != col_a_different
 
 
 def test_parsing(monkeypatch):