i-Rates is designed to optimize and automate everyday pricing and other control decisions in hotel room sales. It is a new-generation software based on the fusion of Revenue Managers’ expertise and contemporary Computer Intelligence algorithms of Machine Learning. This document outlines the most important principles of i-Rates, as well as introduces the system and process elements.
Optimal pricing of business nights together with wise control decisions on stay restrictions and distribution channels uncover opportunities for hotel’s revenue growth. The key point is that decisions have to be strategic, to ensure the available capacity for more valuable customers in the future. It is not necessary to sell all available rooms to maximize the overall profit, but how many should be sold? What should be the proper price dynamics in the context of observed sales flow and actions of market competitors? These questions are addressed by modern optimal and adaptive control theory, and hardly can be considered without the computer intelligence and dynamic optimization algorithms.
But we all need to have these algorithms be practically implemented and speaking the comfortable Revenue Management language, such as RevPAR! Here comes i-Rates.
Elements of Optimal Control: Data is first
The control theory approach is to consider the sequence of decisions (such as price changes) in time, and to make the entire sequence be strategically optimal, in terms of final profit.
Note that i-Rates optimization target is Profit estimate, as opposed to frequently used Room Revenue estimate. The difference is significant and is not reduced to a simple shift of optimal prices by variable costs and additional profits per room, as it seems at first sight. Due to customer flow elasticity, any simple shift in prices will lead to changes in sales flow. This is not the case when Adjusted RevPAR target is used in the algorithm: sales frequency stays the same while optimizing the final collected profit.
Adjusted RevPAR is calculated based on each hotel’s financial structure and takes into account the property’s CPOR and outlet revenues. Taking this metric into account, i-Rates achieves the optimal pricing position that maximizes the bottom line.
To establish the practical environment for optimal control, three critical ingredients are of key importance:
- Measurement. Current flow of sales, prices, occupancy, competitors’ actions and activities of distribution channels should be made observable by the computer system. i-Rates is automatically connected to the Property Management System and Rate Shopping services to collect and organize data in its dynamical time-dependent form. i-Rates also computes meaningful progress metrics, such as room revenue adjusted to variable costs and additional revenues from other departments. Aggregated data is displayed to the manager, with additional graphical representation of dynamics.
- Decision engine. i-Rates is concentrated on the most critical control variables - rack price, stay restrictions (length-of-stay, close-to-arrival), and allotments to (properly grouped) distribution channels. Decision engine is based on dynamical optimization (reinforcement learning) algorithms, taking into account the reconstruction of customers’ flow from observed sales flow, with estimated demand elasticity. Decisions take into account the rooms capacity, and competitors’ prices dynamics and distribution. Machine-generated decisions then follow the (optional, but simple and valuable) dialogue with the manager, and are ready to be examined and utilized.
- Decision implementation. Generated (and optionally confirmed) decisions have to be implemented. At this stage, i-Rates communicates with Channel Management services with machine-formatted decisions, as well as presents them in a convenient report form to be used in existing managers’ practice.
- Receives and aggregates relevant data and metrics
- Generates optimal strategic pricing decisions, optionally confirmed in dialogues with the manager
- Implements these decisions via distribution channels
Business Cycle and Managers’ Interaction with iRates
Every day or according to a prescribed schedule, i-Rates pools fresh portion of sales progress for 365 business days & nights (a year ahead) that are being traded.
All computations are automatically initiated and fulfilled, so the system is always ready with updated pricing recommendations.
When the manager opens the i-Rates user dashboard, all collected information will be displayed in a convenient and intuitive aggregated form. The user dashboard of the i-Rates system comprises two major views, Calendar and Panel. Each view is designed to maximize the usability and comfort of the user during two business activities: (1) Deep study of sales for a particular business night (via Calendar) and (2) a broader inspection of sales of consecutive days in connection (via Panel).
The Panel view displays individual business nights as columns of an interactive table. The table is organized in several segments (groups of rows). Two that are most frequently used are Sales Stats and Control Panel. Other segments are related to market competitors’ data, as well as sales dynamics. Segments can easily be closed and re-opened, thus the user may concentrate only on the portion of information necessary for that particular activity. E.g. if the current goal is to adjust the system recommendations against the market position of competitors, only competitors and control panel will be made visible.
The Control Panel displays the i-Rates recommendations of particular decisions for each business night. Decisions include rack price, stay restrictions as well as estimation of current customer flow rating from very high demand for specific local events, down to very slow seasons.
The user can establish some strategic limits to i-Rates decisions (such as: do not downgrade the customer flow rating below level 3, or do not raise the price above $399, etc). These limits are optional and may reflect the additional useful market information, available for the Revenue Manager. i-Rates will take them into account for subsequent decisions.
And finally, the manger may accept or directly edit the system’s recommendations. These final parameters will be posted to distribution channels. The more i-Rates adapts to the hotel’s market position (by observing a longer history of sales), the less frequent will be the manual corrections of the system’s automatic decisions.
Calendar view is designed to meticulously understand and control the decisions for a particular business night. All nights are organized in the convenient calendar form, with the most important parameters shown in each night's cell. An additional panel is linked to one of the selected nights, with displayed detailed metrics and decisions. For a particular day of interest, the manager may adjust the customer flow rating and rack range, and then ask the system to re-calculate the decision. Normally, this may be done if the manager has some additional market information for this particular business night, or if existing sales statistics are sparse (early sales or low season). Different convenient graphs and reports are also available for viewing.
Both the Calendar view and the Tablet dialogue essentially follow the same business logic:
- Inspect the proposed system decision in the context of sales and other valuable metrics, pay attention to highlighted changes, then
- (Optionally) Inform the system about known sales-generating events, in the form of simple constrains to relative ratings. Establish price restrictions, if any
- (Optionally) Correct some decisions, if they contradict with your trading policy, or in case of poor statistical data available for that day
- Allow system to publish the decisions and display them in a convenient tabular format for your records
Computer Intelligence: What’s inside?
The algorithms of i-Rates decision engine are taking their roots in Machine Learning and Control theory. They are based on three equally important levels of system adaptation to particular market conditions, tuned and working simultaneously as a whole.
The first most basic level takes into account the relatively stable property information, such as capacity, comp set, distribution channels, default seasonal patterns and decisions, and historical sales (either at detailed level, if available, or final statistics for at least one year). This level is devoted to describe the overall market position of the hotel and its revenue and profit fluctuations. The information is aggregated into 10 to 30 most stable “scenarios”, which are individual for each property. This data is initially generated during the i-Rates installation, and is normally updated every 1-2 years, or sooner if desired.
The second intermediate level is responsible for automatic recognition of what kind of scenario, or mix of scenarios, is most likely relevant to observed sales for each traded night. Proprietary recognition algorithms, based on statistical neural networks, utilize the running sales and market information (in full dynamics). The current sales pattern will follow one of the scenarios, with infrequent switches between them. Automatic recognition of these changes in sales patterns constitutes the second, more sensible level of adaptation.
The third level is designed to react to local changes in collected data (so-called changes of system state, termed from control theory). The dynamic optimization of strategic decisions from current state in context of a running scenario leads to the most refined and reactive level of adaptation.
Such a hierarchical decision engine optimizes pricing decisions at all time scales, from broad market position of the hotel, down to quick adaptive reaction to the most recent data. This explains potential benefits of the revenue manager’s involvement into the system operation. The manager can assist the scenario recognition subsystem by providing additional information. For example, setting a particular business night rating below 5 (out of 10). This may happen, if initial sales indicate a high booking activity related to an event (e.g. popular town festival). But occasionally, event dates are changed. If this happens, the manager knows that the booking activity will slow down and the hotel will be expecting cancellations. Being properly informed, the i-Rates engine will not wait until the downgrade of the observed sales volumes becomes sufficient to automatically change the scenario. The manager can provide this information to i-Rates in a timely manner, making sure that the hotel doesn’t lose a penny. This interactive way of using the i-Rates will leverage the manager’s skills and knowledge to incrementally improve the revenue and profitability and obtain high ROI.
If the manager’s predictions don’t materialize – i-Rates will suggest moving to the most suitable scenario, based on the newly obtained information from the current sales flow.
This ensures yield optimization in the situations of minimal market knowledge, or when the system is run by a person who doesn’t possess skills of an experienced Revenue Manager, thus producing big results with little time and/or required expertise.