diff --git a/vizier/_src/algorithms/designers/eagle_strategy/eagle_strategy_utils.py b/vizier/_src/algorithms/designers/eagle_strategy/eagle_strategy_utils.py
index 9c12952a1..6bdcfb4ec 100644
--- a/vizier/_src/algorithms/designers/eagle_strategy/eagle_strategy_utils.py
+++ b/vizier/_src/algorithms/designers/eagle_strategy/eagle_strategy_utils.py
@@ -115,7 +115,7 @@ class EagleStrategyUtils:
   _n_parameters: int = attr.field(init=False)
   _degrees_of_freedom: DefaultDict[vz.ParameterType, int] = attr.field(
       init=False, factory=lambda: collections.defaultdict(int))
-  _original_metric_name: str = attr.field(init=False)
+  _original_metric_name: Optional[str] = attr.field(init=False)
   _goal: vz.ObjectiveMetricGoal = attr.field(init=False)
 
   def __attrs_post_init__(self):
@@ -359,6 +359,8 @@ def get_metric(self, trial: vz.Trial) -> float:
     """Returns the trial metric."""
     if trial.infeasible:
       return np.nan
+    if trial.final_measurement is None:
+      raise ValueError('Trial is not completed.')
     return trial.final_measurement.metrics[OBJECTIVE_NAME]  # pytype: disable=bad-return-type
 
   def is_better_than(
@@ -408,6 +410,8 @@ def standardize_trial_metric_name(self, trial: vz.Trial) -> vz.Trial:
     """Creates a new trial with canonical metric name."""
     if trial.infeasible:
       return trial
+    if trial.final_measurement is None:
+      raise ValueError('Trial is not completed.')
     value = trial.final_measurement.metrics[self._original_metric_name].value
     new_trial = vz.Trial(parameters=trial.parameters, metadata=trial.metadata)
     new_trial.complete(