Data Science: The Paradigm Shift in Revenue Management
What is a paradigm shift in revenue management?
I am often asked to explain what a paradigm shift is, as it relates to hospitality demand forecasting in revenue management and, as a data scientist, I have difficulty explaining the meaning in a simple way.
Creating forecasts that are not (or minimally) based on historical data, but using forward-looking data, is a truly complex analysis exercise, especially in the tourism industry and in revenue management.
It’s important to know that, before Lybra’s Assistant RMS, no RMS provider has ever successfully included the data that we do in our algorithm, for many reasons, including the scarcity of future data and lack of scientific literature capable of providing adequate models.
The best way to explain this is to use a metaphor…
Instead of data, let’s use the spectrum of colors, from blue to yellow, as a metaphor to explain this concept. Imagine building the best algorithm in the world, which was designed to create new colors, by combining the two initial colors: blue and yellow. As powerful as the program is, the result (the new color) will always be one that is based upon the two colors; that means that the program’s suggested result could be any color on the spectrum from blue to yellow, including many gradients of green – but could never create any other colors that are outside of that spectrum.
This is how traditional RMS work; they are limited by their input data – primarily historical data, which has now been made obsolete due to the unprecedented market conditions we’re currently experiencing – not by the program’s capabilities itself.
Let’s look go back to our metaphor to better illustrate the limitations of the input data (or colors)… No matter how hard we tried, we couldn’t create the color brown with our initial input colors (blue and yellow); in order to create this new color, we would need to add a third input color into the algorithm: red.
Now, let’s bring that back to what we are doing at Lybra. We are creating RMS algorithms based on data from new and different sources, including flight searches, metasearches, OTA demand pressure and local events.
To bring it back to our metaphor, we are adding new colors to the input stage, which is used to create a forecast (hotel revenue management) that more accurately describes the true demand forecast for hotels – giving us the opportunity to create all of the colors in the world.
We started to develop Lybra’s Assistant RMS four years ago and we are glad to have the opportunity to work with a dedicated team within Zucchetti Group, one of the most important European companies in the IT sector. Creating forecasts based on future data sources is a paradigm shift in revenue management and, truly, the future of how hotels will price their rooms.
Some basic insights on revenue management with future demand data
Determining which variable is the most important for creating a revenue management forecast is more complicated than it first appears. For a start, is crucial to understand that “most important” is not the accurate word, in this case; it’s more correct to speak of the “relative importance,” which, in data science, means “attributing the right weight to the variables according to the context.”
So, back to the question at hand… which new variables are we analyzing in our advanced revenue management algorithm, and why are they the most “relatively important”?
Flight searches on destinations are an important factor, but the importance of this data tends to decrease the further you get away from the city that “hosts” the airport. (Learn more in our flight search module.)
It may seem redundant in some cases (when compared to flight searches) but, in reality, this data is important for hotels located in the province or far from the airport(s).
Events are another important variable; however, not all events have an impact on the destination and, therefore, on the occupancy of the hotel.
Understanding what the competition is doing crucial for hotels because, regardless of whether they follow revenue management or pricing strategies, as a whole, competitor’s rates determine the reference price. The reference price is one of the most important parameters that the customer evaluates before purchasing a product or service.
Reputation naturally plays an important role because it impacts the determination of the reference price for the customer and, on the other hand, it affects the pricing capacity of the hotel. Price and reputation are inextricably linked and play on different temporal levels. Reputation is an intangible asset, which can deteriorate even with a long-term discounting-based pricing policy, which undermines the brand’s value/reputation, in the minds of consumers.
The old paradigm in revenue management
All variables play an important role in determining the forecast, but unlike the analysis of historical data alone, the weight of each of these variables is relative, contextualized to the individual hotel.
In most cases, traditional forecasts take historical data and project it onto the future, through the analysis of pick-up trends. This methodology has been the mother of forecasting for 40 years; in other words, this methodology was the origins of revenue management as applied science. Much research has been done over the years on how to make forecasts using the time series stored in the PMS (property management system); however, this type of analysis is fundamentally inaccurate, because it does not consider many “relatively important” variables.
For example, if we only analyze last year’s pick-up trends, we only have a number but no information about why that pick-up came out last year. We know nothing about last year’s competitors’ prices or last year’s demand pressure. All of these are important variables that somehow conditioned last year’s pick-up but, as we mentioned earlier, it is information that traditional RMS, using historical, pick-up pricing models do not consider; for this reason, the accuracy of the forecast models is not optimal.
Based on these considerations, it is easy to understand the level of abstraction that traditional revenue management models – and the resulting RMS – have; it also shows us why, today, with new technologies and above all with the strong dynamism of the market, it is necessary for revenue management models to use forward-looking data, not historical.
The new paradigm in revenue management with future demand data
The revenue management optimization process – also known, as the use of complex algorithms with complex future data – has the primary purpose of finding the right measure of each individual variable, and the best combination of those variables, to increase the accuracy of the forecast and, therefore, of the room pricing.
However, given the complexity of the subject, it is unthinkable to manually change the weight of each variable for each hotel, and that is why machine learning algorithms are used. These new artificial intelligence models work independently to find the right balance in the measure of each of the market variables to optimize and scientifically support revenue management decisions.
This fundamental paradigm shift in revenue management is already happening because of Lybra’s Assistant RMS. What makes this new paradigm shift even more exciting is that we are working on a blank slate; as new research and/or practical studies are done by our development team and by universities, new information will be discovered, over time, that can be applied to further optimize revenue management pricing models.