diff --git a/benchmarks/bm_runner.py b/benchmarks/bm_runner.py
index 9980e3b6..6b51d433 100644
--- a/benchmarks/bm_runner.py
+++ b/benchmarks/bm_runner.py
@@ -100,7 +100,7 @@ def _setup_common() -> None:
     _echo("Setup complete.")
 
 
-def _asv_compare(*commits: str) -> None:
+def _asv_compare(*commits: str, fail: bool = False) -> None:
     """Run through a list of commits comparing each one to the next."""
     _commits = [commit[:8] for commit in commits]
     for i in range(len(_commits) - 1):
@@ -114,6 +114,11 @@ def _asv_compare(*commits: str) -> None:
         _echo(comparison)
         shifts = _subprocess_runner_capture([*asv_command, "--only-changed"], asv=True)
         _echo(shifts)
+        if fail and shifts:
+            message = (
+                f"Performance shifts detected between commits {before} and {after}.\n"
+            )
+            raise RuntimeError(message)
 
 
 class _SubParserGenerator(ABC):
@@ -199,7 +204,7 @@ def func(args: argparse.Namespace) -> None:
             asv_command = shlex.split(ASV_HARNESS.format(posargs=commit_range))
             _subprocess_runner([*asv_command, *args.asv_args], asv=True)
 
-        _asv_compare(merge_base, head_sha)
+        _asv_compare(merge_base, head_sha, fail=True)
 
 
 class SPerf(_SubParserGenerator):