As default, PMI will use historical data from the same month for forecasting purposes, i.e. January will be used to forecast January. If you are satisfied with this basic setting, you don’t have to read any further, but we do recommend you build up your Seasons on your own to maintain and gain the most benefits of PMI and its algorithms.

However, for most properties it will be a good idea to divide the year into seasons rather than calendar months. A season is a longer period during which the business mix, price sensitivity and demand are similar during the entire period. The seasons are therefore, by definition, the same from year to year.

As an additional feature, you can mark certain dates as being different from the standard season; for instance, special cases like large congresses, sporting events, holidays, etc.

Intended Users

General Manager, Hotel Manager, Revenue Manager and Booking Manager.


For Room Forecasting, historical pickup statistics must be available for each season, special period and fixed date that is going to be forecast.

If seasons are used for Room Live Forecast, understanding how the PMI forecasting algorithm works is important.

For other forecasting, historical numbers must be available for each season.


1. To edit seasons and special events, open the Tools – Live Forecast module and click on the ‘Seasons’ icon.

  • It is possible to create several sets of season definitions. A chain could, for instance, create one set to be used in a specific city, and another to be used at their Ski Resorts or luxury five-star hotels.

2. To create a new set, click on the [+] icon.

3. Enter a name for the new season.

4. Decide if this season should be applicable only to a specific property.

5. Decide if this season should be applicable to a group of hotels and use the drop down menu.

6. If Holidays are selected, PMI will include the predefined holidays from the Hierarchy (e.g. Easter) in the season calendar.

7. Select the season you want to use from the drop down menu.

8. To edit or delete the definitions for the chosen set, click on the pen icon (edit), or trash can (delete).

There are 4 different types of entries that can be created for the calendar:

  • Seasons
  • Special Periods (which includes Covid high and Covid low)
  • Outliers
  • Fixed Date (New Year’s Eve and Christmas Day). Please note that moving holidays and holidays that are not always on the same dates are not included in this and therefore do not appear in the color purple in Live Forecast.

9. Seasons

A season is a longer period during which the business mix and price sensitivity are similar. You expect the same booking behavior during this period. The seasons should be very similar year to year.

10. Sort order

By clicking here you are able to get a full overview of the type of business.

11. Special Periods

Within a season, some days could be different from the usual due to special events taking place. It is during a shorter time span, and during this period the booking behavior is significantly different. Examples could be a large convention, sports event, concert, etc. Please note that Special Periods should be used for shorter periods, preferably 2 weeks or less, and only if it’s not normal for the season. If it is a recurring period each year during the same period, you should consider setting up a new Season instead. Below is the Special Period – Covid Rolling. Season generation is based on a 2+ years history with a minimum length of 7 weeks. Calculation of seasons should always start on March 8, 2019, regardless of the current date. If a manual season generation is triggered, and even if the number of years is more/less than 2, the start date should always be March 8 on the first year used. Each time a season generation is executed (manually or scheduled) a new Special Period (SP) replaces the old: Start: Current date -7 weeks and End: Current date +5 weeks

12. Outliers

These are dates or periods that should be disregarded for future calculations of pickup. Examples: Hotel entirely or partially closed due to renovation, extreme pickup/cancellations due to force of nature, or files not exported correctly from PMS. As all of these dates will be disregarded it will not add any value to have several types of outliers. However, if you want to keep track of why dates are not used, you can create as many as needed. Also, Outliers detected within the Special Period as usual, and the Special Period is disregarded for detecting outliers during normal seasons.

13. Fixed Date

Dates that are recurring each year and always have a significantly different booking behavior compared to the normal season. Examples: National day and New Year. Please note that holidays that are recurring but on different dates each year are handled by the Holiday function (6) above and automatically entered in the calendar. It also depends on source date start (first OTB Import): If OTB import started 4 years before the date of the holiday -Covid Rolling SP should not be entered on the dates marked as holidays. If OTB import started later – holiday dates are marked as Covid Rolling SP. Holidays that have passed should never be marked as SP.

14. To add a new season, special period, outlier or fixed date, start by entering the name in the field.

  • Choose the type from the drop-down menu and click on the [+] icon. Choose a color by clicking on the gray square next to the name. You can change the color at any time by clicking on the square and the name by clicking on the pen icon. To delete a season, click on the [-] icon next to the season.

When clicking on the + icon next to the name of the season, the Booking Horizons are shown. The value entered is the shortest lead time when the segment might reasonably be expected to have new bookings. Once inside this period, the algorithm will only take actual changes in consideration and not suggest additional bookings. As a default, these are set to 0, i.e. it will be forecast each day up until arrival date.

Once you have made any kind of changes to the seasons, please remember to save by clicking on the Save icon on the bottom right of the page.

Once Seasons has been established, highlight a season you would like to use by clicking on the season name. Then click directly on the date in the calendar on which that season should be used. By holding down the right button on the mouse, you can apply the season to several days. To change to a new season (or special period, etc.), just select a new season and click on the date again. 

When you start editing, you will receive a message asking if you would like to create a new template. If you answer OK, all changes will be applied to both historical and future years. This is very useful when setting up basic seasons and Fixed Dates. By selecting Cancel, the changes will only apply to this year. 

15. You may see the different seasons more easily in the overview by ticking the check box “Show”.  

When activating the Generate Auto Seasons in the Tools-Live Forecast

16. The Min. Seasons Length (Weeks): The minimum length of a season is 1 week and the maximum length is 10 weeks. The best practice is 3 weeks for a minimum based on historical figures. PMI will not create seasonal periods shorter than the provided number.

17. History (Years): This ranges from 1 year to 8 years. PMI will use the provided number of years of historic data to detect seasonality. If you have chosen 5 years and PMI only has 4 years historical data then it will use 4 years historical data.

18. This icon indicates that PMI could not enough data points to establish the revenue driver ratio from seasonality. The values have been approximated by statistics from the recent 5 weeks.

19. This icon indicates that PMI is updated manually.

20. This icon indicates that the forecast is imported from external source.

Sort order for the Live Forecast

Whenever an automated season overview is displayed in any of the Live Forecast modules or Cockpits, the seasons are shown in alphabetical order. A more intuitive sorting order would be according to the season type, from best to worst.

The ARR gives the first sort value, from best to worst:
1.x Expensive
2.x Above
3.x Normal
4.x Below
5.x Cheap
Under each ARR category, the various occupancy levels are the second sort value, determining how the occupancy levels should be sorted under each ARR level. Shown in the order of high occupancy to low:
x.1 Peak
x.2 High
x.3 Medium
x.4 Low
x.5 Dip

Below is an overview of the sort order for the Live Forecast


Do not create too many seasons and special periods, as there will be less information for the algorithm to base its forecasts upon:

  • Try to combine similar booking patterns into one special period, even though the events themselves could be different. Example: combining “sport events” and “concerts” into “Local Event”.
  • As the algorithm will consider the day of the week, there is no need to break up a week into separate seasons/special periods depending on the day of the week, as it is already considered.
  • For OTB based forecasts: The algorithm considers current OTB and uses historical days with similar OTB for comparison; thus the demand for the day is taken in consideration. Therefore, first see if the suggested forecast is adequate without special periods, as fewer special periods means less maintenance
  • For revenue driver based forecasts: When the actual driver goes up or down, the forecast will show the same swings. As long as the expected relation between the driver and the outcome is the same, the same season and/or special period can be used. E.g. a driver based on room nights will automatically suggest a lower forecast on days with low expected occupancy. No special consideration, therefore, has to be made based on expected occupancy levels.

Was this article helpful?

Related Articles