Please use the following information to answer the questions that follow:

1. Assume that a cancelled reservation generates $0 in revenue.
2. Assume that the average price per room * the number of nights is the total revenue for customers who did not cancel.
3. Assume 100% of customers offered a discount accept it.

With those assumptions in mind, please answer the following questions (each worth 20 points):

1. One option is to offer customers a reduced rate if they are willing to guarantee their reservations (meaning there is no refund if they cancel). Market research shows that these types of programs generally discount rates by 35%. If you were to offer a 35% discount to all customers, assuming they all accept, would you increase or decrease revenue on the sample provided (use all 36,275 records)?

1. Another option is to use the model you created in HW3 to target the offer toward customers who are more likely to cancel in the first place. If you use the predicted probabilities from your model in HW3 and offer a 35% non-refundable discount only to customers for whom the offer would be revenue positive (meaning the discounted price is greater than the expected revenue without the discounting), how many people would get the offer?

1. What would be the total revenue (across all 36,275 records) utilizing the targeted offer?

1. Please explain the impact of using the model from HW3 to better target the offers? If the model would cost $100,000 to implement across all channels and $20,000 per year to maintain, is the model worth it? Please explain.

1. The assumptions listed in the introduction simplify the problem quite a bit. Please explain which of these assumptions is likely not to hold in the real world...and how additional predictive models might help to further target offers when these assumptions do not hold.