CHURN_ANALYSIS_PROJECT

Customer Churn — End-to-End Project (SQL → Power BI → Python)

This repository contains a complete churn analytics pipeline using SQL Server,Power BI, and Python. It covers data preparation, visualization, and machine learning modeling to understand and predict customer churn.

📂 Repository Structure

• SQLQuery1.sql → Audit of NULL/missing values
• SQLQuery2.sql → Profiling queries (distribution, revenue share, distinct values)
• SQLQuery3.sql → Views for cohorts (vw_ChurnData, vw_JoinData)
• SQLQuery4.sql → Cleansed product table prod_Churn
• project.pbix → Power BI dashboards
• churn_prediction.ipynb → Jupyter Notebook for churn prediction
• prediction_data.xlsx → Excel export from SQL view (vw_JoinData) for ML input
• /images/ → Add screenshots of Power BI dashboards here

---

1. SQL Server — Data Preparation

Run the queries in SQL Server Management Studio (SSMS):

1. SQLQuery4.sql → Creates [dbo].[prod_Churn] (handles nulls/defaults).
2. SQLQuery3.sql → Creates views: vw_ChurnData and vw_JoinData.
3. SQLQuery1.sql → Runs a NULL audit.
4. SQLQuery2.sql → Runs profiling queries for churn insights.

✅ After execution:

• prod_Churn is cleaned.
• Views vw_ChurnData and vw_JoinData are available for analysis.

---

2. Power BI — Visualization

2.1 Connect to SQL Server

• Get Data → SQL Server
• Import prod_Churn, vw_ChurnData, vw_JoinData
• Use Import mode for better performance

2.2 Create Core DAX Measures

Customer Count = COUNTROWS('prod_Churn')
Churned Customers = CALCULATE([Customer Count], 'prod_Churn'[Customer_Status] = "Churned")
Stayed Customers = CALCULATE([Customer Count], 'prod_Churn'[Customer_Status] = "Stayed")
Churn Rate = DIVIDE([Churned Customers], [Customer Count])
Total Revenue = SUM('prod_Churn'[Total_Revenue])
Avg Monthly Charge = AVERAGE('prod_Churn'[Monthly_Charge])

2.3 Suggested Dashboards

• Page 1 (Overview): KPI cards, Donut chart (Churned vs Stayed), Bar chart (Churn Category), Line chart (Tenure vs Charges)
• Page 2 (Profile): Stacked bar (Contract vs Status), Column chart (Churn by State), Matrix (Internet × Security)
• Page 3 (Churn Reasons): Bar chart of top churn reasons

📸 Add screenshots of dashboards to /images/ and reference in README.

---

3. Python — Machine Learning

Notebook: churn_prediction.ipynb

3.1 Environment Setup

python -m venv .venv
.\.venv\Scripts\activate   # Windows
# or source .venv/bin/activate   # Mac/Linux
pip install -U pandas numpy scikit-learn matplotlib seaborn pyodbc joblib openpyxl sqlalchemy

3.2 Load Data

Option A — From SQL

import pyodbc, pandas as pd
conn = pyodbc.connect("DRIVER={ODBC Driver 17 for SQL Server};SERVER=localhost;DATABASE=db_Churn;Trusted_Connection=yes;")
df = pd.read_sql("SELECT * FROM dbo.prod_Churn", conn)

Option B — From Excel (prediction_data.xlsx)

import pandas as pd
df = pd.read_excel("prediction_data.xlsx")

3.3 Train Model

X = df.drop(columns=["Customer_Status"])
y = (df["Customer_Status"] == "Churned").astype(int)

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

from sklearn.linear_model import LogisticRegression
model = LogisticRegression(max_iter=1000)
model.fit(X_train, y_train)

3.4 Evaluate Model

from sklearn.metrics import classification_report, roc_auc_score
pred = model.predict(X_test)
proba = model.predict_proba(X_test)[:, 1]

print(classification_report(y_test, pred))
print("ROC-AUC:", roc_auc_score(y_test, proba))

3.5 Save Model

import joblib
joblib.dump(model, "churn_model.joblib")

3.6 (Optional) Random Forest & SQL Output

from sklearn.ensemble import RandomForestClassifier
rf = RandomForestClassifier(n_estimators=200, random_state=42)
rf.fit(X_train, y_train)
print("RF ROC-AUC:", roc_auc_score(y_test, rf.predict_proba(X_test)[:, 1]))

from sqlalchemy import create_engine
engine = create_engine("mssql+pyodbc://@localhost/db_Churn?driver=ODBC+Driver+17+for+SQL+Server&trusted_connection=yes")
scored = X_test.copy()
scored["Churn_Prob"] = proba
scored.to_sql("ml_Scored", engine, schema="dbo", if_exists="replace", index=False)

---

🔁 Execution Order

1. Run SQL scripts → prod_Churn & views created
2. Build Power BI dashboards → KPIs & visuals from data
3. Run Notebook → Train churn models using SQL or Excel data
4. (Optional) Write predictions to SQL & refresh Power BI

---

✅ Checklist

• SQL scripts executed successfully
• Power BI visuals built with correct measures
• Notebook runs end-to-end without errors
• Model saved and/or predictions written back to SQL