diff --git a/docs/why.md b/docs/why.md
index 648b402d9..3c6f4825c 100644
--- a/docs/why.md
+++ b/docs/why.md
@@ -15,21 +15,20 @@ print(3 in pl.Series([1, 2, 3]))
 Try it out and see ;) Spoiler alert: they don't. pandas checks if `3` is in the index,
 Polars checks if it's in the values.
 
-How about
-```python
-df_left = pd.DataFrame({'a': [1, 2, 3], 'b': [4, 5, 6]})
-df_right = pd.DataFrame({'a': [1, 2, 3], 'c': [4, 5, 6]})
-df_left.merge(df_right, left_on='b', right_on='c', how='left')
-```
-versus
+For another example, try running the code below - note how the outputs have different column names after the join!
 
 ```python
-df_left = pl.DataFrame({'a': [1, 2, 3], 'b': [4, 5, 6]})
-df_right = pl.DataFrame({'a': [1, 2, 3], 'c': [4, 5, 6]})
-df_left.join(df_right, left_on='b', right_on='c', how='left')
-```
+pd_df_left = pd.DataFrame({'a': [1, 2, 3], 'b': [4, 5, 6]})
+pd_df_right = pd.DataFrame({'a': [1, 2, 3], 'c': [4, 5, 6]})
+pd_left_merge = pd_df_left.merge(pd_df_right, left_on='b', right_on='c', how='left')
 
-?
+pl_df_left = pl.DataFrame({'a': [1, 2, 3], 'b': [4, 5, 6]})
+pl_df_right = pl.DataFrame({'a': [1, 2, 3], 'c': [4, 5, 6]})
+pl_left_merge = pl_df_left.join(pl_df_right, left_on='b', right_on='c', how='left')
+
+print(pd_left_merge.columns)
+print(pl_df_right.columns)
+```
 
 There are several such subtle difference between the libraries. Writing dataframe-agnostic code is hard!
 But by having a unified, simple, and predictable API, you can focus on behaviour rather than on subtle