About Statistical KBIs

Statistical KBIs use statistical (regression) formulas to forecast future business volumes by analyzing historical information. Most statistical KBIs are created automatically by the system when you create revenue center periods and market segments. All revenue center periods, market segment groups, rooms, and arrivals are statistical KBIs. However, you must complete their regression formulas. The only way to manually create a Statistical KBI is to select Separate as the Type in the General tab.

Statistical KBIs assess the relationship between dependent (actual data) and independent (forecast data) factors. The relationship between the two is then used to predict a future dependent factor based on the projected independent factor.  For example, the number of guests in a hotel is an independent factor. Assuming that a hotel restaurant gets most of its business from hotel guests, the number of people who eat in the hotel restaurant is dependent on the number of guests in the hotel. (If the restaurant got most of its business from local patrons, regression analysis would not be the appropriate methodology to select, as was mentioned above). Therefore regression uses the historical relationship between the dependent and independent factors to take a forecast for a future independent factor and use it to arrive at a forecast for the dependent factor.

Regression also allows you to use multiple independent factors.  In other words, when forecasting meal covers, you could ask the software to analyze hotel guests, banquet guests, and any other independent factor that may show a relation. Thus, the statistical model determines how effective an independent factor is in predicting a dependent factor and, based on the relationship, predicts it again. Of course, if the prediction of the independent factor is inaccurate, the subsequent prediction of the dependent factor will be inaccurate. In this case, LMS will return a predicted volume forecast zero (o). Keep in mind that the system needs approximately 20 weeks of past history to begin to accurately determine the relationship between the dependant and independent factors. Occasionally, the system will return values for many of the forecast days and zeros for the others. That simply means that there is a relationship on some days of the week and not on others.  This does not mean the use of the statistical model is wrong. It might suggest the need for additional past history to return a value.

Frequently, the KBIs for Transient rooms in a hotel are examples of statistical KBIs. These rooms can be statistically projected by tracking the relationship of actual rooms (actual data) to booked rooms (forecast data).

When you create a statistical KBI, you select one or more related KBIs, which are used to generate a forecast, and you choose an appropriate operation.

Statistical KBIs use the following mathematical operators:

• Independent operators are used to indicate that the associated KBI name is an independent value to be used in the regression calculation.
• Add operators are used to add to the current value of the independent value.
• Subtract operators are used to subtract from the current value of the independent value.

For example, if you selected four related KBIs and assigned modifiers as follows:

• KBIA Independent
• KBIB Add
• KBIC Subtract
• KBID Independent

You would produce the following formula, where A, B, and C are the regression coefficients and KBIX represents the current value of the KBI:

KBIX = A(KBIA + KBIB -- KBIC) + B(KBID) + C

LMS lets you create KBIs for factors whose impact you want to assess without assuming a perfect one-to-one relationship between one guest factor and the available guest market. For example, local guests attending AM meetings may use a restaurant for breakfast, but it is difficult to determine what percentage will do so. For this reason, the local guest market is more difficult to measure than in-house guests. In this case, you can create a statistical KBI to assess the impact of local guests as an independent variable.

Terms & Definitions

Related topics:

• Forecast Structure
  • Calculated KBIs
  • Environmental Tab
  • Input KBIs
  • General Tab
  • Percent of Base KBIs
  • Statistical KBIs