diff --git a/tests/scripts/analyze_outcomes.py b/tests/scripts/analyze_outcomes.py
index b52952458..4e925a18e 100755
--- a/tests/scripts/analyze_outcomes.py
+++ b/tests/scripts/analyze_outcomes.py
@@ -50,11 +50,7 @@ def execute_reference_driver_tests(results: Results, ref_component, driver_compo
     """Run the tests specified in ref_component and driver_component. Results
     are stored in the output_file and they will be used for the following
     coverage analysis"""
-    # If the outcome file already exists, we assume that the user wants to
-    # perform the comparison analysis again without repeating the tests.
-    if os.path.exists(outcome_file):
-        results.info("Outcome file ({}) already exists. Tests will be skipped.", outcome_file)
-        return
+    results.new_section("Test {} and {}", ref_component, driver_component)
 
     shell_command = "tests/scripts/all.sh --outcome-file " + outcome_file + \
                     " " + ref_component + " " + driver_component
@@ -188,27 +184,18 @@ suite_case = "<suite>;<case>"
 
     return outcomes
 
-def do_analyze_coverage(results: Results, outcomes_or_file, args):
+def do_analyze_coverage(results: Results, outcomes, args):
     """Perform coverage analysis."""
     results.new_section("Analyze coverage")
-    outcomes = read_outcome_file(outcomes_or_file) \
-               if isinstance(outcomes_or_file, str) else outcomes_or_file
     analyze_outcomes(results, outcomes, args)
 
-def do_analyze_driver_vs_reference(results: Results, outcomes_or_file, args):
+def do_analyze_driver_vs_reference(results: Results, outcomes, args):
     """Perform driver vs reference analyze."""
     results.new_section("Analyze driver {} vs reference {}",
                         args['component_driver'], args['component_ref'])
 
     ignored_suites = ['test_suite_' + x for x in args['ignored_suites']]
 
-    if isinstance(outcomes_or_file, str):
-        execute_reference_driver_tests(results, args['component_ref'], \
-                                       args['component_driver'], outcomes_or_file)
-        outcomes = read_outcome_file(outcomes_or_file)
-    else:
-        outcomes = outcomes_or_file
-
     analyze_driver_vs_reference(results, outcomes,
                                 args['component_ref'], args['component_driver'],
                                 ignored_suites, args['ignored_tests'])
@@ -507,17 +494,29 @@ def main():
 
         # If the outcome file exists, parse it once and share the result
         # among tasks to improve performance.
-        # Otherwise, it will be generated by do_analyze_driver_vs_reference.
-        if os.path.exists(options.outcomes):
-            main_results.info("Read outcome file from {}.", options.outcomes)
-            outcomes_or_file = read_outcome_file(options.outcomes)
-        else:
-            outcomes_or_file = options.outcomes
+        # Otherwise, it will be generated by execute_reference_driver_tests.
+        if not os.path.exists(options.outcomes):
+            if len(tasks_list) > 1:
+                sys.stderr.write("mutiple tasks found, please provide a valid outcomes file.\n")
+                sys.exit(2)
+
+            task_name = tasks_list[0]
+            task = KNOWN_TASKS[task_name]
+            if task['test_function'] != do_analyze_driver_vs_reference:
+                sys.stderr.write("please provide valid outcomes file for {}.\n".format(task_name))
+                sys.exit(2)
+
+            execute_reference_driver_tests(main_results,
+                                           task['args']['component_ref'],
+                                           task['args']['component_driver'],
+                                           options.outcomes)
+
+        outcomes = read_outcome_file(options.outcomes)
 
         for task in tasks_list:
             test_function = KNOWN_TASKS[task]['test_function']
             test_args = KNOWN_TASKS[task]['args']
-            test_function(main_results, outcomes_or_file, test_args)
+            test_function(main_results, outcomes, test_args)
 
         main_results.info("Overall results: {} warnings and {} errors",
                           main_results.warning_count, main_results.error_count)