Forecasting Accurate Customer Counts in The Quick Service Restaurant Industry
Open AccessForecasting Accurate Customer Counts in The Quick Service Restaurant Industry Customer or demand forecasting is critical for any restaurant to manage and plan for revenue. Currently, methodology of demand forecasting centers around predicting 1 to 2 weeks ahead. Furthermore, a methodology to accurately forecast customer counts monthly for 1 year ahead does not exist for franchisees (owner/operators) of McDonald’s® restaurants. These independent owner/operators (operators) rely solely on judgmental reasoning to forecast their baseline customer counts for the year ahead. While there is certainly room for operators to incorporate their sense of the area in which they operate, an accurate baseline forecast will assist in planning of staffing, capital expenditures, and inventory. This praxis will explore univariate modeling to determine which univariate model is more accurate and determine the potential savings from applying such model.Data were collected from one franchise of eight restaurants monthly from January 1999 to December 2019. In reviewing the data, some locations had a major rebuild or remodel that resulted in near zero customer counts during construction as well as a large increase in the number of customers the months immediately following the reopening after the construction period. Typically, a rebuild or major remodel improves efficiency in production of product in the kitchen along with improved efficiency in the restaurant service areas. This is done by adding new equipment, changing the layout of the restaurant, and adding points to order either in drive-thru and/or inside the restaurant. Given that this research is to predict a baseline model to forecast customer counts, the data were pared down to represent 5 consecutive years (60 months) of customer traffic for each of the eight locations without any major construction activity. Once the data were prepared, they were tested for required stationarity and differenced if necessary. A model was then selected with the best characteristics and validated by performing 12 months ahead forecasts comparing the forecasted models with the actual values for year 5 customer counts for each location. Univariate methods used in comparison to the operator’s forecast were (1) autoregressive integrated moving average (ARIMA), (2) seasonal autoregressive integrated moving average (SARIMA), (3) exponential smoothing, and (4) census X-13-ARIMA-SEATS. The results indicated univariate modeling was significantly more accurate than all of the operator’s eight locations at forecasting monthly customer counts 1-year-ahead yielding a significant cost savings for operators.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.