This project delivers a comprehensive Marketing Mix Model (MMM) to analyze and forecast sales for a global Fast-Moving Consumer Goods (FMCG) brand. The analysis uses five years of monthly data across five international markets to measure the effectiveness of marketing strategies and predict future performance.
An FMCG manufacturer needed to optimize its marketing spend and understand the key drivers behind its sales figures. The primary challenge was to quantify the impact of three core marketing activities (i.e., TV advertisements, online banner ads, and in-store sales promotions) on sales volume.
The company sought data-driven answers to the following critical questions:
- How many additional unit sales are generated by each marketing activity?
- Given an annual spend of £2,000,000 on TV ads and £500,000 on banner ads, which channel is more cost-effective?
- What factors, other than marketing, contribute to significant variations in monthly sales?
- What are the projected sales for the next month across all regions?
The analysis is built on a proprietary dataset containing 5 years of monthly records for 5 distinct international regions. Key features in the dataset include:
- Sales: Number of product units sold.
- Price: The base price of the product.
- Ad1 (TV): TV advertising measured in Gross Rating Points (GRPs).
- Ad2 (Banners): The number of online banner advertisements run per month.
- Prom (Promotions): The number of stores participating in sales promotions.
- Economic Indicators: Monthly wage increase percentages for the working population.
- Categorical Data: Region, month, and product type identifiers.
The project follows a structured analytical pipeline to ensure robust and actionable insights.
- Data Pre-processing: The initial phase involves cleaning the data, handling any anomalies, and applying necessary transformations to prepare variables for modeling.
- Exploratory Data Analysis (EDA): Visualizing data to uncover underlying trends, seasonality, and correlations between marketing activities and sales.
- Model Development: A multiple regression model was constructed to quantify the relationship between marketing inputs, price, economic factors, and sales volume.
- Validation and Robustness Checks: The model's performance was rigorously tested to check for overfitting and ensure key statistical assumptions were met. This step builds confidence in the model's predictive power.
- Forecasting: The validated model was used to generate sales predictions for the upcoming month for all regions. Model accuracy was evaluated using the Mean Absolute Percentage Error (MAPE)