Case Study

Client:

Home Depot

​

Challenge:

The client was investing in advertising across various platforms—YouTube, Facebook, and newspapers—but lacked a clear understanding of which channels were most effective and how to optimize their budget for sales growth. The company needed a data-driven approach to determine the true value of each advertising channel and forecast the impact of future spending decisions.

​

Solution:

We conducted a comprehensive two-part analysis using R and advanced statistical modeling, including adstock theory and marketing mix modeling.

1. Exploratory Data Analysis and Linear Regression: We began by analyzing the relationship between ad spend and sales on each platform. A linear regression model was built to quantify the impact of each channel. We then rigorously evaluated the model for reliability by checking for multicollinearity and heteroscedasticity.

2. Time Series Analysis and Forecasting: In the second phase, we converted the data into a time series object to analyze sales trends over time. We then built a forecasting model to predict sales outcomes based on different budget allocation scenarios. Specifically, we modeled the impact of shifting ad spend from underperforming channels to those with a higher ROI.

​

Results:

The analysis yielded powerful, actionable insights that will guide future marketing strategy and budget allocation:

​

• Platform Effectiveness: The linear regression model revealed that Facebook and YouTube ad spend have a positive, statistically significant relationship with sales. Specifically, sales increased by 118 units per $1,000 of Facebook ad spend and by 45 units per $1,000 of YouTube ad spend.

• Poor Performers: In stark contrast, newspaper advertising was found to have a negative, non-statistically significant relationship with sales, with a projected decrease of six units per $1,000 of ad spend.

• Forecasting Growth: By reallocating ad spend from newspapers to Facebook and YouTube, the forecasting model predicted a tangible and positive shift in sales performance. Visualizations clearly showed a noticeable increase in forecasted sales with the new budget allocation compared to the original plan.

• Model Reliability: The model's reliability was confirmed by a high R-squared value of 0.88, indicating that it accurately explains 88% of the data. Additionally, low Variance Inflation Factor (VIF) scores confirmed a lack of multicollinearity , and tests for heteroscedasticity came back clean, ensuring the results are trustworthy and consistent.

​

Conclusion & Strategic Recommendations:

This project provides a clear roadmap for maximizing ROI. The client can confidently reallocate their ad spend away from newspapers and invest more heavily in Facebook and YouTube, which are proven to drive sales. The predictive model empowers the business to move from guesswork to a data-driven strategy, ensuring every dollar spent on advertising contributes to measurable growth.