In the hospitality sector, demand for accommodation rooms fluctuates daily, seasonally and annually. Analysis of these fluctuations in demand produces unreliable forecasts. Effective revenue management realization requires the development of an appropriate forecasting mechanism. Attainment of reliable forecasts is never easy even though the process of forecasting is simple.
Hence, Cross, R. G. (1997) suggested some rules applicable in forecasting. Cross suggested that the prediction must stay at comprehensive level. Detailed predictions contribute significantly to current Hotel Revenue Management Systems.
The second suggestion is that a large amount of information has to be used in the investigation. Additionally, the predictions must be adjusted regularly to potential changes in the business environment.
Precise predictions are vital in improvement of price and availability suggestions for hotel rooms. In addition, precise forecasting improves decisions made on recruitment of workers, purchase of goods and budget preparations.
In contrast, incorrect forecasting results into adoption of inefficient decisions on price and availability suggestions that the revenue management systems produce. Such inefficient decisions affect the revenue of a hotel negatively.
Talluri (2004) identified two forms of revenue management predictions. The first kind is quantify-dependent revenue management prediction. This is mainly used in aviation and hospitality industries. The other type is price-dependent revenue management prediction.
Apart from demand information, quantify-dependent revenue management prediction needs data on the arrival of reservation requests from different types of clients. This data is obtained at the time when clients make orders. This means that the information utilized in lodge demand prediction depends on present reservation activities, past data that relates to every day arrivals or the number of sold rooms.
Furthermore, this kind of prediction requires guess on cancelled reservations and the number of clients who do not show up. A reservation is grouped as “No Show” if the client who made the request does not terminate it before the deadline of the hotel reaches and does not come to claim it.
Prediction used in hotel revenue management should consider two variables that relate to time. These two variables are reservations and consumption times.
There exist other vital issues considered apart from the prediction method chosen. Some of the key issues are the forecast period, aggregation levels and measurement of the prediction’s precision. Weatherford et al (2001) proved that completely disaggregated predictions provide dependable outcomes compared to partially aggregated strategies.
Therefore, it is appropriate for hotels to examine arrivals by the duration customers spend in the facilities and the category of price. One key concern to take into consideration is the method applicable in time division to produce an appropriate basis for prediction. A common practice is to consider a week as a forecast. In this method, each day is considered differently.
For example, predictions for Mondays rely on information collected on other Mondays and so on. Another vital point to consider is the periodic phenomenon, which has significant influence in the hotel industry. Consideration of limited data set limits managers’ ability to capture seasonality. Conversely, the use of many periods may make predictions made unresponsive and rigid.
Managers physically developed predictions through analysis of historical data before the introduction of revenue management systems. They analyzed the length of stay in hotels and the price categories. The process lacked sophistication and consumed a lot of time. Consequently, the manual system is inappropriate in the current business environment due to stiff competition and fast-paced market activities.
Prediction strategies can be grouped into three categories namely historical, advance and combined forecast models. Historical reservation models only take into consideration total rooms booked on earlier nights. It considers the number of rooms booked and arrivals at a hotel. Advance reservation models take into consideration the booking behavior of customers over a given period.
Finally, joint reservation methods consider features of both chronological and advance methods to establish predictions. Weatherford and Kimes, (2003) showed that exponential smoothing, pickup and moving standard forms are the most dependable prediction strategies.
However, the performance of these strategies depends on the data available. Hence, hotel revenue managers must have adequate prediction models. Additionally, they should use unlike strategies at different times and in diverse markets.
Forecast Accuracy
Accuracy is a vital component in prediction since it determines the method selected. Numerous measures of the performance of the predictions made exist. The Mean Absolute Deviation (MAD) is the easiest and popularly used method. It involves determination of the averages of the absolute values of prediction errors made. The Mean Percent Error (MPE) is the mean of the percentage errors made.
The Mean Absolute Percent Error (MAPE) and MPE are similar. However, MAPE averages the total figures of the prediction errors. The Root Mean Square Error (RMSE) takes into consideration the square root of the averages of the squared prediction errors.
MPE and MAPE are appropriate since they provide limitless figures. Nonetheless, they stay unclassified when the total number of reservations is zero. Hence, Thiel’s Inequality Coefficient is used to overcome the disadvantage. The coefficient is denoted by U in which F* and F are exact and predicted reservations.
Pickup Model
Pickup model is a well-known advance reservation strategy that takes advantage of the unique features of booking information. It relies on reservation information instead of reliance on arrival histories to make dependable predictions.
However, it provides variations in the predictions. The model can be further divided into preservative or multiplicative, standard or superior and straightforward or subjective average prediction strategies.
Judgmental Forecasting
Human judgment plays a significant role in revenue management practice. This is despite the ability of a revenue management system to perform multiple tasks in the process. Conventionally, predictions include human opinions. Judgmental prediction involves the inclusion of human opinions. A different research exhibited that opinions can be used to make adjustments in past predictions.
A revenue manager can then supervise the prediction process using the adjustments. Moreover, hotels can make appropriate use of revenue management systems in educating managers. A well-trained revenue manager can know how to analyze data and make appropriate plans. This can result into efficient operations in a hotel. Notably, business people agree that human judgment play a pivotal role in predictions.
Human judgment plays a vital role in generation of accurate predictions. However, human judgment is inclined towards biasness. Individuals may make decisions or predictions that can result into achievement of their personal goals. In addition, the differences in skills and abilities affect people’s judgment. Hence, well-trained people can make appropriate judgment compared to undertrained persons.
In early prediction studies, Hogarth and Makridakis, (1981) examined the effectiveness of human opinions in prediction generation. They concluded that statistical predictions were dependable compared to judgmental forecasts. Their study found that human predictions had biasness and errors. Human judgment involved much control and humans had overconfidence.
Fischhoff, (1988) noted that appropriate forecasts required human judgment in the selection of the model used, determination of parameters and investigation of study outcomes. There are two reasons for need of accuracy in judgmental prediction. Experts have adequate data and are able to acquire information in time.
There exists some features that can assist revenue managers improve their judgment. Revenue management systems are acceptable in prediction generation. The perception of forecasters on revenue management systems affect decisions or forecasts made. Secondly, revenue management systems are easy to manage, and they enable comparison of statistical predictions made.
Hence, they assist revenue managers to determine errors easily. Additionally, revenue management systems provide flexibility in methods used in generation of forecasts. This makes forecasters feel responsible and involved in revenue management. Revenue management systems also make managers employ correct techniques.
Comprehensible assistance and clear methods eliminate confusion and mistakes in generation of predictions. Revenue management systems also support the integration of human judgment and statistical methods in generation of demand forecasts. Reliable revenue management systems produce dependable statistical predictions. They also support judgmental adjustments in generation of demand forecasts.
An income manager can examine the effectiveness of a revenue management system. The manager can examine the usefulness of a database and the appropriateness of statistical techniques. In addition, the manager can compare the error measurement methods used in generation of demand predictions.
Examination of these factors can assist a manager to establish whether a revenue management system can enable development of appropriate forecasts. Integration of the mentioned features assists managers to design revenue management systems that enable development of precise predictions.