diff --git a/Results.md.in b/Results.md.in
index c241702..89b7212 100644
--- a/Results.md.in
+++ b/Results.md.in
@@ -10,8 +10,15 @@ generated {{ now }}, from results of {{ n_hosts }} machines
 
 ### Time
 
+#### Absolute
+
 ![unique-runningtime](unique-runningtime.png)
 
+#### Normalized to `numpy.unique` for each shape/host combination
+
+![unique-runningtime-normalized](unique-runningtime-normalized.png)
+
+
 ## Bincount
 
 ### Memory usage
@@ -20,5 +27,10 @@ generated {{ now }}, from results of {{ n_hosts }} machines
 
 ### Time
 
+#### Absolute
+
 ![bincount-runningtime](bincount-runningtime.png)
 
+#### Normalized to `numpy.bincount` for each shape/host combination
+
+![bincount-runningtime-normalized](bincount-runningtime-normalized.png)
diff --git a/generate-results.py b/generate-results.py
index 2244e4b..675fe09 100644
--- a/generate-results.py
+++ b/generate-results.py
@@ -38,6 +38,46 @@ def gen_plots():
 
     p.clear()
 
+    # normalize data
+    # method    shape   t   py_version  npy_version pandas_version  vigra_version   host    platform    mem_min mem_max
+    np_unique = results_unique[results_unique["method"] == "numpy.unique(data)"]
+
+    normalization_factors = {}
+
+    for _, row in np_unique.iterrows():
+        key = (
+            row["host"],
+            row["platform"],
+            row["shape"],
+            row["py_version"],
+            row["npy_version"],
+            row["pandas_version"],
+            row["vigra_version"],
+        )
+        normalization_factors[key] = row["t"]
+
+    def normalize(row):
+        key = (
+            row["host"],
+            row["platform"],
+            row["shape"],
+            row["py_version"],
+            row["npy_version"],
+            row["pandas_version"],
+            row["vigra_version"],
+        )
+        return row["t"] / normalization_factors[key]
+
+    results_unique["normalized_t"] = results_unique.apply(normalize, axis=1)
+
+    p = sns.lineplot(x="shape_ind", y="normalized_t", hue="method", data=results_unique)
+    p.set(yscale="log")
+    fig = p.get_figure()
+    fig.savefig("unique-runningtime-normalized.png")
+
+    p.clear()
+
+
     results_bincount = pandas.read_csv("results-bincount.csv", sep="\t")
 
     results_bincount["shape_ind"] = results_bincount["shape"].map(shape_index_mapping)
@@ -56,6 +96,46 @@ def gen_plots():
 
     p.clear()
 
+    # normalize data
+    # method    shape   t   py_version  npy_version pandas_version  vigra_version   host    platform    mem_min mem_max
+    np_bincount = results_bincount[results_bincount["method"] == "numpy.bincount(data.reshape(-1))"]
+
+    normalization_factors = {}
+
+    for _, row in np_bincount.iterrows():
+        key = (
+            row["host"],
+            row["platform"],
+            row["shape"],
+            row["py_version"],
+            row["npy_version"],
+            row["pandas_version"],
+            row["vigra_version"],
+        )
+        normalization_factors[key] = row["t"]
+
+    def normalize(row):
+        key = (
+            row["host"],
+            row["platform"],
+            row["shape"],
+            row["py_version"],
+            row["npy_version"],
+            row["pandas_version"],
+            row["vigra_version"],
+        )
+        return row["t"] / normalization_factors[key]
+
+    results_bincount["normalized_t"] = results_bincount.apply(normalize, axis=1)
+
+    p = sns.lineplot(x="shape_ind", y="normalized_t", hue="method", data=results_bincount)
+    p.set(yscale="log")
+    fig = p.get_figure()
+    fig.savefig("bincount-runningtime-normalized.png")
+
+    p.clear()
+
+
 
 def main():
     template = Template(Path("Results.md.in").read_text())