diff --git a/examples/Inverse_perovskites/Inverse_formate_perovskites.ipynb b/examples/Inverse_perovskites/Inverse_formate_perovskites.ipynb
index c3b8d406..4834a2e7 100644
--- a/examples/Inverse_perovskites/Inverse_formate_perovskites.ipynb
+++ b/examples/Inverse_perovskites/Inverse_formate_perovskites.ipynb
@@ -88,7 +88,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "We now search through the elements of interest (Li-Fr) and find those that are allowed on each site. In this example, we use the F- anion with an increased Shannon radius to simulate the formate anion. We access the Shannon radii data directly from the smact data directory and are interested in the octahedral (6_n) Shannon radius."
+    "We now search through the elements of interest (Li-Fr) and find those that are allowed on each site. In this example, we use the F- anion with an increased Shannon radius to simulate the formate anion. We access the Shannon radii data directly from the smact data directory and are interested in the octahedral (6_n) Shannon radius. "
    ]
   },
   {
@@ -113,6 +113,19 @@
     "                B_list.append([row[0],int(row[1]),float(row[4])])"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "*NB: We access the data directly from the data directory file here for transparency. However, reading the file multiple times would slow down the code if we were looping over many (perhaps millions to billions) of compositions. As such, reading all the data in once into a dictionary, then accessing that dictionary from within a loop, could be preferable, e.g.:*\n",
+    "``` python\n",
+    "for element in search:\n",
+    "    ...\n",
+    "    r_shannon = shannon_radii[element][coordination]\n",
+    "    ...\n",
+    "```"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},