This report explains finding that a lodging property can generally use information on its occupancy to improve the accuracy of cover forecasts in its food and beverage outlets.

We examine twenty-seven forecasting methods. Six of the methods forecast covers using only an outlet's historical data, while the others include information on the property's occupancy.

We conducted our study using four hotels that have a total of thirty-three combinations of food and beverage outlets and dayparts. The food and beverage outlets include room service, lounges, cafés, and main restaurants.

Since we have extensive historical data from one of the properties, we split that into two samples, giving a total of forty-one outlet-daypart scenarios. In all of the cases we used an eight-week holdback data set to test the models.

In thirty-four of the forty-one outlet–daypart scenarios, the best forecast originated with one or another of the models incorporating occupancy data. On average, forecast accuracy improved by over 11 percent when using occupancy data.

In those thirty-four cases where using occupancy data improved the forecasts, the average improvement in accuracy was over 14 percent, while the accuracy improvement exceeded 25 percent in seven of the scenarios.

If this CHR Report made a positive impact on your management approach or business operations, we welcome your commentary. We would like to post your comments on our website. Please submit your comments to js372@sha.cornell.edu and dss18@sha.cornell.edu.

To view the whole report, please click on the link below: www.hotelschool.cornell.edu/chr/pdf/showpdf/chr/research/thompsonfbforecasttopost.pdf