diff --git a/wta/bench.py b/wta/bench.py
index b92b7ec8bacf84a58a4c9f05ca75a55493a08e05..206032255b43d5a1be75782802e816abd2e5d4c0 100755
--- a/wta/bench.py
+++ b/wta/bench.py
@@ -81,11 +81,15 @@ def run_one(args, i):
 
     copar_args = [copar, 'refine', '--stats-json', f]
 
-    out = subprocess.run(
-        copar_args,
-        stdout=subprocess.DEVNULL,
-        stderr=subprocess.PIPE,
-        check=True)
+    try:
+        out = subprocess.run(
+            copar_args,
+            stdout=subprocess.DEVNULL,
+            stderr=subprocess.PIPE,
+            check=True)
+    except subprocess.CalledProcessError as e:
+        print("error: %s" % e.stderr.decode('utf-8'))
+        raise
 
     stats = json.loads(out.stderr.decode('utf-8'))