diff --git a/torcharrow/icolumn.py b/torcharrow/icolumn.py
index 8bde374e6..dbcffe2ae 100644
--- a/torcharrow/icolumn.py
+++ b/torcharrow/icolumn.py
@@ -1457,14 +1457,17 @@ def _nunique(self, drop_null=True):
         else:
             return len(set(i for i in self if i is not None))
 
-    # quantile
     @trace
     @expression
     def _quantile(self, q, interpolation="midpoint"):
         """
-        Compute the q-th percentile of non-null data.
+        Compute the q-th quantile of non-null data.
 
         Inefficient prototype implementation.
+
+        Args:
+            q: float or array-like quantiles to compute, value must be within [0, 1]
+            interpolation: interpolation method to use
         """
 
         if interpolation != "midpoint":
@@ -1475,8 +1478,8 @@ def _quantile(self, q, interpolation="midpoint"):
             return []
         out = []
         s = sorted(self)
-        for percent in q:
-            k = (len(self) - 1) * (percent / 100)
+        for p in q:
+            k = (len(self) - 1) * p
             f = math.floor(k)
             c = math.ceil(k)
             if f == c:
diff --git a/torcharrow/inumerical_column.py b/torcharrow/inumerical_column.py
index 131221b4b..0fc961a8f 100644
--- a/torcharrow/inumerical_column.py
+++ b/torcharrow/inumerical_column.py
@@ -189,7 +189,7 @@ def describe(
             res._append(("mean", self.mean()))
             res._append(("std", self.std()))
             res._append(("min", self.min()))
-            values = self._quantile(percentiles, "midpoint")
+            values = self._quantile(percentiles / 100, "midpoint")
             for p, v in zip(percentiles, values):
                 res._append((f"{p}%", v))
             res._append(("max", self.max()))