diff --git a/launchable/commands/subset.py b/launchable/commands/subset.py
index dc4afb115..b25e2b2fb 100644
--- a/launchable/commands/subset.py
+++ b/launchable/commands/subset.py
@@ -248,9 +248,18 @@ class FallbackMode(str, Enum):
     default="run-all",
     help="Behavior when the subset API is unavailable or the model is untrained. "
          "'run-all' (default) runs all tests as usual; 'stop' exits with a non-zero status so CI halts; "
-         "'random-sample' picks a random subset locally based on the count derived from --target "
+         "'random-sample' picks a random subset locally based on the count derived from --fallback-sampling-target "
          "(no duration estimates are available in this path).",
 )
+@click.option(
+    "--fallback-sampling-target",
+    "fallback_sampling_target",
+    hidden=True,
+    type=PERCENTAGE,
+    default=None,
+    help="Sampling ratio (0%–100%) used when --fallback-mode=random-sample. "
+         "Required when --fallback-mode=random-sample is specified.",
+)
 @click.pass_context
 def subset(
     context: click.core.Context,
@@ -283,6 +292,7 @@ def subset(
     similarity: Optional[float] = None,
     subset_id_file: Optional[str] = None,
     fallback_mode: str = "run-all",
+    fallback_sampling_target: Optional[float] = None,
 ):
     fallback_mode_enum = FallbackMode(fallback_mode)
     app = context.obj
@@ -316,6 +326,18 @@ def warn(msg: str):
             stack_trace=msg
         )
 
+    if fallback_mode_enum == FallbackMode.RANDOM_SAMPLE and fallback_sampling_target is None:
+        print_error_and_die(
+            "--fallback-sampling-target is required when --fallback-mode=random-sample",
+            Tracking.ErrorEvent.USER_ERROR
+        )
+
+    if fallback_sampling_target is not None and fallback_mode_enum != FallbackMode.RANDOM_SAMPLE:
+        print_error_and_die(
+            "--fallback-sampling-target can only be used with --fallback-mode=random-sample",
+            Tracking.ErrorEvent.USER_ERROR
+        )
+
     if is_get_tests_from_guess and is_get_tests_from_previous_sessions:
         print_error_and_die(
             "--get-tests-from-guess (list up tests from git ls-files and subset from there) and --get-tests-from-previous-sessions (list up tests from the recent runs and subset from there) are mutually exclusive. Which one do you want to use?",  # noqa E501
@@ -425,6 +447,7 @@ def __init__(self, app: Application):
             self.is_get_tests_from_guess = is_get_tests_from_guess
             self.subset_id_file = subset_id_file
             self.fallback_mode = fallback_mode_enum
+            self.fallback_sampling_target = fallback_sampling_target
             super(Optimize, self).__init__(app=app)
 
         def _default_output_handler(self, output: List[TestPath], rests: List[TestPath]):
@@ -618,7 +641,8 @@ def _fallback_result(self) -> SubsetResult:
                 )
                 sys.exit(1)
             elif self.fallback_mode == FallbackMode.RANDOM_SAMPLE:
-                target_fraction = float(target) if target is not None else 1.0
+                assert self.fallback_sampling_target is not None
+                target_fraction = float(self.fallback_sampling_target)
                 click.echo(
                     "Warning: the service failed to subset. Falling back to local random sample at {:.0%}.".format(
                         target_fraction),
diff --git a/tests/commands/test_api_error.py b/tests/commands/test_api_error.py
index c935fd62b..a3623fd98 100644
--- a/tests/commands/test_api_error.py
+++ b/tests/commands/test_api_error.py
@@ -513,9 +513,11 @@ def test_api_error_fallback_random_sample(self):
             status=500)
 
         with tempfile.NamedTemporaryFile(delete=False) as rest_file:
-            result = self.cli(*self._subset_args(rest_file.name, ("--fallback-mode", "random-sample")), mix_stderr=False)
+            extra = ("--fallback-mode", "random-sample", "--fallback-sampling-target", "100%")
+            result = self.cli(*self._subset_args(rest_file.name, extra), mix_stderr=False)
             self.assert_success(result)
             self.assertIn("example_test.rb", result.stdout)
+            self.assertIn("at 100%", result.stderr)
 
     @responses.activate
     @mock.patch.dict(os.environ, {"LAUNCHABLE_TOKEN": CliTestCase.launchable_token})
@@ -561,7 +563,8 @@ def test_brainless_fallback_random_sample(self):
             status=200)
 
         with tempfile.NamedTemporaryFile(delete=False) as rest_file:
-            result = self.cli(*self._subset_args(rest_file.name, ("--fallback-mode", "random-sample")), mix_stderr=False)
+            extra = ("--fallback-mode", "random-sample", "--fallback-sampling-target", "50%")
+            result = self.cli(*self._subset_args(rest_file.name, extra), mix_stderr=False)
             self.assert_success(result)
             self.assertIn("example_test.rb", result.stdout)
 
@@ -580,3 +583,13 @@ def test_brainless_fallback_run_all_default(self):
             result = self.cli(*self._subset_args(rest_file.name), mix_stderr=False)
             self.assert_success(result)
             self.assertIn("example_test.rb", result.stdout)
+
+    # --- Validation error cases ---
+
+    @responses.activate
+    @mock.patch.dict(os.environ, {"LAUNCHABLE_TOKEN": CliTestCase.launchable_token})
+    def test_random_sample_without_fallback_sampling_target_fails(self):
+        with tempfile.NamedTemporaryFile(delete=False) as rest_file:
+            result = self.cli(*self._subset_args(rest_file.name, ("--fallback-mode", "random-sample")), mix_stderr=False)
+            self.assertNotEqual(result.exit_code, 0)
+            self.assertIn("--fallback-sampling-target", result.stderr)