🧠 Smart Price Optimization using XGBoost

The Problem We Solve

Pricing is the toughest decision for any business.

Imagine a shop selling noodles. A few years ago when there was no competition, they could charge ₹5. Then as the market grew, they increased price to ₹10. Customers accepted it because:

• The brand was trusted
• Quality was good
• No better alternatives existed

Then competition arrived.

Other brands offered similar noodles at lower prices. The shop thought: "We're the market leader, let's charge ₹12 instead."

What happened? Customers switched to competitors offering the same product at ₹10. Sales crashed. Profit collapsed.

Why? Because when competition exists, you can't just increase price. Customers will leave.

---

💡 The Real Challenge

This happens to EVERY business:

✗ Too high price → Customers switch to competitors → Low volume → Low profit
✗ Too low price → Volume is good but margins crushed → Low profit
✓ Perfect price → Good volume + Good margins = Maximum profit

But how do you find that perfect price?

Most businesses guess. They try ₹10, see what sells. Then try ₹12, see sales drop. Then go back to ₹10. This is chaos!

What if you could predict the best price using DATA?

---

🚀 What This Project Does

This system uses Machine Learning (XGBoost) to analyze your past sales data and tell you:

• "Your optimal price is ₹10.50, not ₹12"
• "This product can handle a price increase"
• "This product should keep its current price"
• "If competitor goes down to ₹8, you should reconsider"

Real Impact:

BEFORE (Guessing):
Month 1: ₹10 price, ₹3,000 profit
Month 2: Increase to ₹12, ₹1,000 profit (oops!)
Month 3: Back to ₹10, ₹3,000 profit

Average: ₹2,333/month (wasted potential)

AFTER (Smart Pricing):
Month 1: ₹10 price, ₹3,000 profit
Month 2: Tested ₹10.50, ₹3,200 profit ✅
Month 3: Stayed at ₹10.50, ₹3,200 profit ✅

Average: ₹3,133/month (30% GAIN!)

---

📊 How It Works

Step 1: Feed Your Historical Data

Collect your past sales records:
- What price you charged
- How many units sold at that price
- Your cost per unit
- Date/time period

Step 2: ML Model Learns Patterns

The XGBoost model analyzes:
- "When price went from ₹10 to ₹12, sales dropped 80%"
- "When price stayed at ₹10, sales stayed consistent"
- "Customers are price-sensitive to changes above ₹11"

Step 3: Get Recommendations

Model tells you:
- Current price: ₹10, Sales: 100/day, Profit: ₹200/day
- If you try ₹10.50: Predicted sales: 95/day, Profit: ₹220/day ✅
- If you try ₹11: Predicted sales: 40/day, Profit: ₹80/day ❌
- If you try ₹12: Predicted sales: 10/day, Profit: ₹30/day ❌

Recommendation: Stay at ₹10 or try ₹10.50 MAX

Step 4: Make Data-Backed Decisions

No more guessing!
- Implement the recommended price
- Monitor sales
- Adjust based on real market conditions

---

📁 Dataset Format

Place retail_price.csv in the project folder with these columns:

Column	Description
unit_price	Price you're currently charging (₹)
units_sold	Number of units sold at that price
cost_price	Cost to produce/buy the unit
product_category_name	Category (Noodles, Snacks, etc.)
product_id	Unique product identifier
month_year	Month of sale (for trends)

---

📦 Requirements

pip install pandas numpy matplotlib seaborn xgboost joblib streamlit scikit-learn

Or use:

pip install -r requirements.txt

---

🧪 Model Details

Aspect	Details
Algorithm	XGBoost Regressor
Purpose	Predict optimal pricing
Training Data	80% of historical sales
Testing Data	20% for validation
Metrics	RMSE (accuracy) & R² (model quality)
Parameters	max_depth=5, learning_rate=0.1, n_estimators=150

---

🎯 Dashboard Features

When you run the app, you get:

1. 📊 Data Preview - See your sales history at a glance
2. 📈 Model Performance - How accurate are the predictions?
3. 🔍 Feature Importance - What factors matter most for pricing?
4. 💡 Top Opportunities - Which products are underpriced/overpriced?
5. 🧾 Custom Prediction - "What if I price this at ₹12?" Get instant predictions
6. 📉 Price vs Sales Relationship - Visualize how price affects demand

---

💻 How to Run

streamlit run smart_price_optimization.py

Then:

1. Upload your sales data (or use the included retail_price.csv)
2. View the dashboard
3. Check predictions for different price points
4. Get recommendations
5. Implement the optimal price

---

📈 Real-World Use Cases

✅ Retail shops - Price products to beat competition while keeping profit
✅ Food vendors - Find the price customers will happily pay
✅ E-commerce stores - Dynamic pricing based on market data
✅ Restaurants - Menu item pricing strategy
✅ FMCG brands - Wholesale pricing optimization

---

🎓 Key Learning

The Sweet Spot Formula

PROFIT = (Price - Cost) × Quantity Sold

NOT: "Higher price = more profit"
BUT: "Price that keeps volume high = more profit"

EXAMPLE:
₹12 price: (12-6) × 10 units = ₹60 profit
₹10 price: (10-6) × 50 units = ₹200 profit ✅