When dealing with monthly or temporal data, standard random splits (Option D) or cross- validation (Option A) can cause "temporal leakage," where the model inadvertently learns from future data to predict the past. According to ISACA AAIATM principles, "Time Series" validation is the most appropriate method for sequential data. It involves training the model on a specific period (e.g., months 1?0) and testing it on the subsequent period (e.g., month 11). This approach accurately reflects how the model will perform in production and is essential for detecting data drift, as it identifies when seasonal trends or long-term shifts in customer behavior cause the model's accuracy to degrade over time.