Unravelling data-driven opportunities
in the hotel industry.

The need for data-driven decisions in businesses is increasing. All sorts of data are generated at a thrilling speed in almost all industries. Most companies have, at least to some extent, access to their own data. However, the challenge lies in translating all this data into useful business information. This is where the area of data science and (big) data analytics comes into play. This area is a dynamic one, rapidly growing and is evolving as a major component in many businesses. Data Science Lab. thinks that the hotel industry should be no exception in this evolution. The explosion of guest and travel data in the last couple of years, combined with a growing industry fuels the need to exploit the opportunities data brings to the industry. Especially considering COVID-19 which led to uncertain future demands. However, the extent to which these opportunities are exploited by different hotel operators varies greatly and unfortunately falls short most of the time. We believe that great data-driven opportunities for the hotel industry exist and can be exploited easily with the right expertise. Hotels that make use of these opportunities will improve guest satisfaction, operational efficiency and profits in an innovative way. By doing so will secure an important competitive advantage, especially when the hotel industry boots up again after the COVID-19 pandemic.

This article explores the data science opportunities for both independent hotels and hotel chains in different areas of the hotel business which include; guest satisfaction (via review analysis), food & beverage (F&B), revenue management and (dynamic) room pricing, and hotel property management. We highlight some specific use-cases within these departments which we think are low-hanging fruit and are quite easily exploitable. The use-cases described in this article are certainly not exhaustive, but very well reflect the extent to which opportunities are present within the hotel industry.

 

Review analysis – Natural Language Processing

The amount of user-generated-content (UGC) exploded since the worldwide adoption of the internet. Prior research showed that UGC is very important for driving demand for experience goods. This is especially the case in the hotel industry, where huge amounts of UGC in the form of online reviews on hotels and travel destinations have been, and increasingly are being, produced by consumers. Next to that, research showed that one of the most popular online activities is searching for travel-related information, and that consumers are highly being influenced by hotel reviews in making travel decisions. All this underlines the importance of exactly knowing what guests write about your hotel(s), and the underlying sentiment with which they write. Surprisingly though, not many hotels are tapping into the valuable information these reviews hold. At most, the reviews are being analysed by a simple text-analysis algorithm (such as the one from Booking.com), which is of course valuable, but does not reveal all valuable information which is included in these reviews.

We think that data science can be of great help here, by discovering patterns and extracting useful information from various written resources (in this case mostly reviews) by making use of natural language processing (NLP) techniques. Inferring the level of emotion, or sentiment, with which hotel guests are writing about hotels, seasonal differences in hotel reviews, specific aspects that hotel guests appreciate most (and least) of specific hotels (and how to improve these aspects) are just a fraction of the valuable insights these data science techniques could quickly generate from hotel reviews. By making use of data science techniques to analyse hotel reviews, the management of hotels would enable themselves to act more quickly and more specific to trends in guest‘s needs and wants, and by doing so quickly improving overall guest satisfaction. Furthermore, hotel chains can discover general trends about the likes and dislikes of specific hotels, and how these can be leveraged or changed positively. A text analysis on more than 100.000 guest reviews, showed that guests appreciate different aspects of a hotel very differently between seasons. For example, the location of a hotel was found to be of much more importance during high season, compared to low season. On the other hand, the level of service and facilities was deemed more important in low season compared to high season. Being aware of these kinds of guest insights, allow a hotel to react more appropriately to guest’s needs and increase guest satisfaction by doing so.

 

Revenue Management – forecasting and price optimization

Now, let’s explore one of the most important things that drives room revenue in the hotel industry: revenue management. The hotel industry has a long history with all sorts of room pricing strategies. Even though data-driven room pricing is a known practice for some hotels (hotels from hotel chains mostly), most hotels that apply revenue management and dynamic pricing today rely on old, complicated, and technical deficient legacy systems. Revenue managers making use of these systems often have to update room prices manually, in all different kinds of systems and web portals. This is a time inefficient and fault sensitive task, which also limits the speed and agility with which prices can be updated. Furthermore, most prices are updated based on simple metrics such as “the demand on the same day last year” and the “pick-up” (number of bookings for a specific date made in a specific time-window).

Although this approach works to some extent, modernization of current revenue management strategies into data-driven, flexible revenue management and dynamic pricing will boost room revenues, ADR (average daily rate) and RevPAR (Revenue Per Available Room). Data Science techniques can help to properly forecast demand and propose hotel room prices that maximizes ADR as well as occupancy. Accurate forecasts of demand can be made by leveraging multiple internal data sources (pick-up, demand last year, guest segments, hotel and room characteristics etc.) as well as multiple external data sources. Armed with an accurate forecast of demand, a pricing algorithm can be implemented which also automatically updates prices and ensures rate parity (making sure the same price is offered on all channels). This approach also frees up time for the revenue manager to concentrate on in-depth analyses, as the revenue manager has to worry less on all manual tasks and receives good pricing suggestions from the algorithm. Next to this, grouping guests into specific data-driven segments, each with their own pricing characteristics, can further optimise pricing strategies.

 

Take for example the following, simplified, scheme:

Image 1: The Data Science Lab approach to a data-driven revenue management strategy

 

It should not come as a surprise that the above situation is not the reality for most hotel chains, let alone independent hotels. We believe that the first step to implement a data-driven, flexible and self-updating optimized revenue strategy should not be that big of a step especially not with the right expertise.

 

F&B department – menu engineering and clustering guests

Another important playing field in the hotel industry is food and beverage (F&B). While the core business of most hotels is maximizing room sales, optimizing F&B sales can be an important additional stream of revenue. Due to POS (Point Of Sale) systems, which are present most of the times in F&B outlets, most F&B departments are quite “data rich”. Due to this data richness of F&B departments, data science techniques can help to quickly increase F&B revenue streams for hotels.

In-restaurant purchase behaviour of guests is stored in these POS systems. Analysing this data can help to create a data-driven menu that is more profitable and yields considerable savings in F&B procurement. These data-driven menus can be further specified to certain guests segments, addressing segment-specific behaviour. Applying these data-driven techniques results in a new menu that boosts profit, while at the same time more specifically answers different guests’ needs.

Next to menu engineering in general, identifying which (type of) customer is most likely to spend a lot in F&B outlets can yield significant results. Performing a simple, data driven, clustering of guests based on average F&B spend and other guest features can identify future guests that may be spending much on F&B. Identifying these guests upfront enables management to take specific actions in targeting these guests in various ways (and thus increase their F&B spending). Next to the direct increase in F&B revenue, it is highly possible that these specific guests feel their (F&B specific) needs better served which can lead them to return more frequently to the hotel’s F&B outlets now and in the future, ensuring a future F&B revenue stream as well.

Greatly reducing F&B waste, which often happens during (large) banqueting events, is another example where data science (and to start: simple analysis) can be of help. Analysing banqueting data and the associated waste of food per event, can help predict future F&B demand per banqueting event more specifically and precisely. With these predictions and insights, hotels can prepare the amount of F&B needed far more accurate, which in turn leads to less waste of F&B and other resources (for example, less staff costs for F&B preparations).

 

Facility/Real Estate management

Data science can also improve property management and maintenance efficiency, as well as increasing the effectiveness of refurbishment decisions. Properly maintaining a hotel-property is very important as the state of the hotel property is directly linked to its services and guest satisfaction, and high costs may be associated with the maintenance of hotel properties. Maintenance is often organised in a very responsive manner or completely set in stone (for example: refurbishments every 15 year). Neither of these methods give room for flexibility or innovative, cost-reducing solutions. If hotel management seek more efficient, smarter maintenance practices, then data-driven solutions can help.

Analytical models can be build using historical maintenance data of one property (or several properties), which can help predict when specific parts of the hotel require maintenance. Accurately predicting exactly when certain maintenance tasks need to be performed can help improve efficient maintenance scheduling and decrease costs while doing so. Using historical maintenance data to make predictions on the long term can also help building a proper multi-year maintenance plan. Next to this, undertaking preventive maintenance tasks, instead of merely reacting to occurred deficiencies as reported by hotel guests or staff, could cause hotel guests noticing less flaws and deficiencies, which in turn can improve overall guest satisfaction.

Data science can also help increase the effectiveness of refurbishment strategies by investigating the relationship between guest satisfaction and specific sorts of refurbishments. Analysis can show that specific refurbishment efforts directly increase guest satisfaction and incremental revenues (for example replacing carpet in rooms with laminate flooring) and that other maintenance efforts have no, or insignificant effect (for example replacing door knobs). Knowing this, supported by data-driven reports, helps management in deciding which refurbishments to do, and which not. Executing this data-driven refurbishment strategy cuts costs where possible and maximizes the return on investments made.

 

How to get there

So, this all sounds very cool and promising, but how to actually achieve this data-driven mindset and practical applications? At Data Science Lab we believe that innovation occurs one step at a time. So the first thing to do is identify a problem or business case to solve (for example: analyse reviews thoroughly and extract insights to improve guest satisfaction), and focus only on this specific case. The next thing to do is build a data lake, or, as an intermediate step, identify, clean and store useful data in a reproducible way.

Quickly building a first prototype is the next, and probably most important, step. A prototype solution (or showcase) of the business case in scope can quickly show results. These results can either be promising or not; both cases are valuable. If the results are promising, we keep developing the prototype until it’s a full grown data science solution. If the results are not promising, we start the process again with another use-case or, if this seems promising beforehand, another approach for the same use-case. This (agile) approach ensures continuity and results, while maintaining maximum flexibility. In short, this approach means that we focus on delivering value to the business quickly and build on it incrementally, while it also prevents that we linger for too long on non-rewarding use-cases or solutions.

 

Conclusion

The use-cases presented in this article are only a small portion of what is possible with today’s techniques. With the growing amount of data in the travel industry (hotels specifically), combined with ever demanding guests and fierce competition, we are convinced that data science can help sustain competitive advantage, especially after the COVID-19 pandemic. We believe that this does not have to be that hard. Our approach is focusing on one thing at a time, combined with the right data-expertise. This will quickly yield results: more revenue, less costs or a higher guest satisfaction.

If this article raised your interest we would like to get in touch with you to brainstorm and discuss, without any obligation, how we can realise your data-driven ambitions in the business together!