📊 Marketing Mix Model (MMM)

An end-to-end Marketing Mix Model built in Python using OLS regression and budget optimization. The model measures the revenue contribution of each marketing channel (TV, Digital, Email, Social), estimates ROI per $1 spent, and recommends an optimal budget allocation to maximize revenue.

---

🎯 Business Questions

• Which marketing channels are driving the most revenue?
• What is the ROI of each channel?
• How should we reallocate our budget to maximize revenue?

---

📁 Project Structure

marketing_mix_model/
├── outputs/                     # Generated charts saved here
├── src/
│   ├── data_prep.py             # Synthetic data generation + adstock transformation
│   ├── model.py                 # OLS regression, diagnostics, revenue decomposition
│   ├── optimizer.py             # Budget optimization via scipy
│   └── visualizations.py       # All charts and summary dashboard
├── main.py                      # Run the full pipeline
├── requirements.txt
└── README.md

---

🧪 Methodology

1. Data

Synthetic weekly marketing data (3 years / 156 weeks) with realistic spend patterns:

• TV: bursty campaign flights (~40% of weeks)
• Digital: consistent always-on spend
• Email: low cost, steady spend
• Social: growing investment trend over time

2. Adstock Transformation

Advertising has a carryover effect — a TV ad seen this week still influences purchases next week. We model this with geometric adstock:

adstock[t] = spend[t] + decay_rate × adstock[t-1]

Channel	Decay Rate	Interpretation
TV	0.6	High carryover — brand awareness lingers
Digital	0.3	Moderate — intent fades quickly
Email	0.1	Low — effect is mostly immediate
Social	0.4	Moderate carryover

3. OLS Regression Model

revenue = baseline
        + β_tv       × tv_adstock
        + β_digital  × digital_adstock
        + β_email    × email_adstock
        + β_social   × social_adstock
        + β_promo    × promo
        + ε

Each coefficient β represents revenue generated per unit of adstocked spend — this is the channel ROI.

4. Model Diagnostics

Metric	Description
R²	% of revenue variance explained
Adjusted R²	R² penalized for model complexity
MAPE	Mean Absolute Percentage Error
Durbin-Watson	Checks for autocorrelation in residuals
p-values	Statistical significance of each channel

5. Budget Optimization

Uses scipy.optimize.minimize (SLSQP) with:

• Equality constraint: total spend = budget
• Diminishing returns modeled via square-root transformation
• Minimum/maximum spend bounds per channel

---

📈 Output Charts

File	Description
01_spend_over_time.png	Weekly spend by channel
02_model_fit.png	Actual vs predicted revenue + residuals
03_revenue_decomposition.png	Stacked channel contributions over time
04_channel_roi.png	ROI per $1 spent by channel
05_budget_optimization.png	Current vs optimal budget allocation
06_budget_scenarios.png	Revenue curve across budget levels
07_summary_dashboard.png	Full summary dashboard

---

🚀 Getting Started

git clone https://github.com/yourusername/marketing_mix_model.git
cd marketing_mix_model
pip install -r requirements.txt
python main.py

No data download required — data is generated synthetically by the pipeline.

---

🛠 Tech Stack

• Python 3.14+
• statsmodels — OLS regression with full statistical diagnostics
• scipy — budget optimization (SLSQP)
• pandas / numpy — data wrangling
• matplotlib — visualizations

---

💡 Key Concepts Demonstrated

• Adstock / carryover effect modeling
• OLS regression with statistical inference (p-values, confidence intervals)
• Revenue decomposition by marketing channel
• Constrained budget optimization with diminishing returns
• Marketing ROI measurement and reporting

---

👤 Author

Built as part of a data science portfolio project.
Background: 15+ years in Marketing Analytics | SQL | Python | Statistical Modeling