diff --git a/statistical_analysis/src/statistical_analysis/core.py b/statistical_analysis/src/statistical_analysis/core.py
new file mode 100644
index 0000000..9940e75
--- /dev/null
+++ b/statistical_analysis/src/statistical_analysis/core.py
@@ -0,0 +1,102 @@
+from __future__ import annotations
+import logging
+from dataclasses import dataclass
+from typing import List, Dict, Any
+import pandas as pd
+import numpy as np
+
+
+logger = logging.getLogger(__name__)
+
+
+class OutlierDetectionError(Exception):
+    """Custom exception raised when data validation or detection fails."""
+    pass
+
+
+@dataclass
+class OutlierAnalysis:
+    column_name: str
+    outlier_value: float
+    drift_signature: str
+    timeout_counts: int
+
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "column_name": self.column_name,
+            "outlier_value": self.outlier_value,
+            "drift_signature": self.drift_signature,
+            "timeout_counts": self.timeout_counts,
+        }
+
+
+def _validate_input_data(log_data: List[Dict[str, Any]]) -> pd.DataFrame:
+    if not isinstance(log_data, list) or not all(isinstance(entry, dict) for entry in log_data):
+        raise OutlierDetectionError("Input log_data must be a list of dictionaries.")
+    if len(log_data) == 0:
+        raise OutlierDetectionError("Input log_data is empty.")
+
+    df = pd.DataFrame(log_data)
+    if df.empty:
+        raise OutlierDetectionError("Converted DataFrame is empty.")
+    return df
+
+
+def analyze_outliers(log_data: List[Dict[str, Any]]) -> OutlierAnalysis:
+    """Analysiert Logdaten und identifiziert Ausreißer mit Fokus auf p99-Region und Latenzverteilungen."""
+    logger.debug("Starting outlier analysis.")
+    df = _validate_input_data(log_data)
+
+    numeric_cols = df.select_dtypes(include=["number"]).columns.tolist()
+    if not numeric_cols:
+        raise OutlierDetectionError("No numeric columns found for outlier analysis.")
+
+    # Compute p99 per column
+    outlier_scores = {}
+    for col in numeric_cols:
+        series = df[col].dropna()
+        if series.empty:
+            continue
+        p99_value = np.percentile(series, 99)
+        outlier_scores[col] = p99_value
+        logger.debug("Computed p99 for %s: %f", col, p99_value)
+
+    if not outlier_scores:
+        raise OutlierDetectionError("No valid numeric data available for outlier computation.")
+
+    outlier_col = max(outlier_scores, key=outlier_scores.get)
+    outlier_value = outlier_scores[outlier_col]
+
+    # Simple drift signature and timeout correlation heuristic
+    drift_signature = "stable"
+    timeout_counts = 0
+
+    if "drift_signature" in df.columns:
+        sig_counts = df["drift_signature"].value_counts()
+        if not sig_counts.empty:
+            drift_signature = sig_counts.idxmax()
+
+    if "timeout_counts" in df.columns:
+        timeout_counts = int(df["timeout_counts"].sum())
+
+    result = OutlierAnalysis(
+        column_name=outlier_col,
+        outlier_value=float(outlier_value),
+        drift_signature=str(drift_signature),
+        timeout_counts=timeout_counts,
+    )
+
+    # CI-ready validation
+    assert isinstance(result.column_name, str)
+    assert isinstance(result.outlier_value, float)
+    assert isinstance(result.drift_signature, str)
+    assert isinstance(result.timeout_counts, int)
+
+    logger.info(
+        "Outlier analysis completed: column=%s, value=%f, drift=%s, timeouts=%d",
+        result.column_name,
+        result.outlier_value,
+        result.drift_signature,
+        result.timeout_counts,
+    )
+    return result