campaign doctor added

app/ads1/campaign_doctor.py

@@ -0,0 +1,101 @@
+import pandas as pd
+import json
+
+def build_campaign_table(dfs: dict) -> pd.DataFrame:
+    df = dfs["campaigns"].copy()
+
+    df["cost"] = df["cost"].fillna(0)
+    df["clicks"] = df["clicks"].fillna(0)
+    df["conversions"] = df["conversions"].fillna(0)
+    df["impressions"] = df["impressions"].fillna(0)
+
+    df["ctr"] = df["clicks"] / df["impressions"].replace(0, 1)
+    df["cpl"] = df["cost"] / df["conversions"].replace(0, 1)
+    df["conv_per_cost"] = df["conversions"] / df["cost"].replace(0, 1)
+
+    return df
+
+def score_campaigns(df: pd.DataFrame) -> pd.DataFrame:
+    df = df.copy()
+
+    # normalize-safe scoring components
+    df["eff_score"] = df["conv_per_cost"]
+    df["eff_score"] = df["eff_score"].fillna(0)
+
+    df["cpl_score"] = 1 / df["cpl"].replace(0, 1)
+    df["cpl_score"] = df["cpl_score"].fillna(0)
+
+    df["volume_score"] = df["conversions"]
+
+    # final blended score (simple + stable)
+    df["final_score"] = (
+        df["eff_score"] * 0.5 +
+        df["cpl_score"] * 0.3 +
+        df["volume_score"] * 0.2
+    )
+
+    return df
+
+def extract_campaign_insights(df: pd.DataFrame):
+    df = score_campaigns(df)
+
+    if df.empty:
+        return None
+
+    best = df.sort_values("final_score", ascending=False).iloc[0]
+    worst = df.sort_values("final_score", ascending=True).iloc[0]
+
+    # scale candidate = good efficiency but not top spender yet
+    scale_pool = df[
+        (df["conversions"] > 0)
+    ].sort_values("eff_score", ascending=False)
+
+    scale = scale_pool.iloc[0] if not scale_pool.empty else best
+
+    return {
+        "best": best.to_dict(),
+        "worst": worst.to_dict(),
+        "scale": scale.to_dict()
+    }
+
+def run_campaign_doctor(dfs: dict) -> dict:
+    print("\n🧠 [campaign_doctor] STARTED", flush=True)
+
+    if not dfs or "campaigns" not in dfs:
+        return {
+            "best_campaign": None,
+            "budget_drain": None,
+            "scale_candidate": None
+        }
+
+    df = build_campaign_table(dfs)
+    insights = extract_campaign_insights(df)
+
+    if not insights:
+        return {
+            "best_campaign": None,
+            "budget_drain": None,
+            "scale_candidate": None
+        }
+
+    best = insights["best"]
+    worst = insights["worst"]
+    scale = insights["scale"]
+
+    result = {
+        "best_campaign": {
+            "name": best["name"],
+            "reason": "Strongest efficiency and conversion performance relative to spend"
+        },
+        "budget_drain": {
+            "name": worst["name"],
+            "reason": "High spend with weak conversion output and poor return efficiency"
+        },
+        "scale_candidate": {
+            "name": scale["name"],
+            "reason": "High conversion efficiency suggests strong potential for scaling"
+        }
+    }
+
+    print("📤 [campaign_doctor] result generated", flush=True)
+    return result