diff --git a/sagemaker_batch_transform/introduction_to_batch_transform/batch_transform_pca_dbscan_movie_clusters.ipynb b/sagemaker_batch_transform/introduction_to_batch_transform/batch_transform_pca_dbscan_movie_clusters.ipynb
index 7393604be8..420935fe05 100644
--- a/sagemaker_batch_transform/introduction_to_batch_transform/batch_transform_pca_dbscan_movie_clusters.ipynb
+++ b/sagemaker_batch_transform/introduction_to_batch_transform/batch_transform_pca_dbscan_movie_clusters.ipynb
@@ -261,7 +261,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Now, we'll setup to split our dataset into train and test.  Dimensionality reduction and clustering don't always require a holdout set to test accuracy, but it will allow us to illustrate how batch prediction might be used when new data arrives.  In this case, our test dataset will be a simple 1% sample of items."
+    "Now, we'll setup to split our dataset into train and test.  Dimensionality reduction and clustering don't always require a holdout set to test accuracy, but it will allow us to illustrate how batch prediction might be used when new data arrives.  In this case, our test dataset will be a simple 0.5% sample of items."
    ]
   },
   {
@@ -270,7 +270,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "test_products = products.sample(frac=0.01)\n",
+    "test_products = products.sample(frac=0.005)\n",
     "train_products = products[~(products.index.isin(test_products.index))]"
    ]
   },