diff --git a/json_test_suite/scripts/analysis.py b/json_test_suite/scripts/analysis.py
new file mode 100644
index 0000000..c83c40e
--- /dev/null
+++ b/json_test_suite/scripts/analysis.py
@@ -0,0 +1,84 @@
+import pandas as pd
+import json
+import yaml
+import argparse
+
+def load_frame(path):
+    # Load the JSON file
+    with open(path, "r") as file:
+        data = json.load(file)
+
+    # Normalize the JSON data
+    rows = []
+    for category_index, category in enumerate(data):
+        category_name = category["category"]
+        for case_index, case in enumerate(category["cases"]):
+            for test_index, test in enumerate(case["tests"]):
+                rows.append({
+                    "category": category_name,
+                    "case": case_index,
+                    "test": test_index,
+                    "valid": test["valid"],
+                    "success": test["success"],
+                })
+
+    # Create the DataFrame
+    df = pd.DataFrame(rows)
+
+    # Set the MultiIndex
+    df.set_index(["category", "case", "test"], inplace=True)
+    
+    return df
+
+
+def load_frames(baseline_path, test_path):
+    # Load the DataFrames
+    df = pd.concat({
+        "baseline": load_frame(baseline_path),
+        "test": load_frame(test_path),
+    }, axis=1)
+
+    return df
+
+def diff(baseline_path, test_path):
+    df = load_frames(baseline_path, test_path)
+    assert df[('baseline', 'valid')].equals(df[('test', 'valid')])
+
+    # Identify regressed and improved tests
+    df['regressed'] = df[("baseline", "success")] & ~df[("test", "success")]
+    df['improved'] = ~df[("baseline", "success")] & df[("test", "success")]
+
+    # Build the result dictionary
+    regressions = {}
+    improvements = {}
+    for category, category_data in df.groupby(level='category'):
+        cat_reg = []
+        cat_imp = []
+        
+        for (case, case_data) in category_data.groupby(level='case'):
+            regressed_tests = case_data[case_data['regressed']].index.get_level_values('test').tolist()
+            improved_tests = case_data[case_data['improved']].index.get_level_values('test').tolist()
+            
+            if regressed_tests:
+                cat_reg.append({"case": case, "tests": regressed_tests})
+            if improved_tests:
+                cat_imp.append({"case": case, "tests": improved_tests})
+        
+        if cat_reg:
+            regressions[category] = cat_reg
+        if cat_imp:
+            improvements[category] = cat_imp
+
+    result = {"regressions": regressions, "improvements": improvements}
+    return result
+
+def main(baseline_path, test_path):
+    delta = diff(baseline_path, test_path)
+    print(yaml.dump(delta))
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("baseline_path", type=str, help="Path to the baseline JSON file")
+    parser.add_argument("test_path", type=str, help="Path to the test JSON file")
+    args = parser.parse_args()
+    main(args.baseline_path, args.test_path)
diff --git a/json_test_suite/scripts/requirements.txt b/json_test_suite/scripts/requirements.txt
new file mode 100644
index 0000000..fb6c7ed
--- /dev/null
+++ b/json_test_suite/scripts/requirements.txt
@@ -0,0 +1 @@
+pandas