Business Context
BrandX, a consumer goods company with $500M in annual revenue, runs multiple TV advertising campaigns throughout the year. The marketing team needs to understand the residual impact of these campaigns on sales, particularly how long the effects last after a campaign ends. This understanding will inform future marketing budget allocations and campaign strategies.
Dataset
| Feature Group | Count | Examples |
|---|
| Time Series Data | 36 | Month (Jan-Dec for 3 years) |
| Advertising Expenditure | 12 | TV spend, digital spend, print spend |
| Sales Data | 36 | Monthly sales figures |
| Control Variables | 5 | Seasonality indicators, economic factors |
- Size: 36 months of data, 55 features total
- Target: Monthly sales figures (continuous variable)
- Class balance: Not applicable (regression problem)
- Missing data: 5% missing in advertising expenditure features due to incomplete records
Requirements
- Build a regression model to estimate the carryover effects of TV advertising on monthly sales.
- Quantify how long the effects of a campaign persist after it ends.
- Include feature importance analysis to identify which advertising channels are most effective.
- Address any missing data appropriately.
- Explain your choice of model and evaluation strategy.
Constraints
- The model must provide interpretable results to guide marketing decisions.
- Inference must be able to run monthly for updates on campaign effectiveness.
- The solution should be scalable to include more advertising channels in the future.
