Big Data Demand Signals
How Machine Learning Identifies Demand Signals Across Multiple Disparate Data Demand Sources
By Ravneet Bhandari, Chief Executive Officer, LodgIQ
Big data, more than a buzzword, has by now become a conundrum that we, consumers and providers of information, try to crack and make sense of it. Essentially, we know that data is becoming larger with wider access to complex algorithms and connections. The onion metaphor - the peeling back of many layers - can be used to reflect the multifaceted aspects of machine learning technology. These swaths of data or rather layered strings of data sets turn these complex entities into a more accurate view of customer demand for the hotelier.
What is Machine Learning?
As a sub-field of computer science and artificial intelligence dealing with building systems that autonomously learn from data, machine learning allows companies to sift through remarkable amounts of information and make empowering recommendations.
Netflix utilizes it to make viewing suggestions, and Facebook uses it to populate your feed with trending video and content selections. The technology introduces mathematical modelling into the process of identifying patterns and making decisions. Then, it adapts and "learns" from all the data signals collected over time and optimizes its decision-making based on all previously inputted data.
The learning ability is what makes this technology powerful by becoming a faster and more efficient version of itself. In all, there are four types of machine learning styles, supervised, unsupervised, semi-supervised and reinforcement. The first three are all slightly different permutations of the same idea, the only major difference relates to whether there are specific desired outputs programmed in. Reinforcement learning maximizes performance though determining an ideal behavior, such as when a computer gets better at playing an opponent in an online chess match.
What's in it For the Hotelier?
For the hotel business, machine learning delivers demand and pricing clarity for hoteliers by constantly providing access to as many data points as possible to create the most accurate forecast. Essentially, the computer is able to calculate optimum demand generating forecasting in real-time, based on fluctuating market changes. Over time, a machine learning based revenue management system becomes smarter as it automatically adapts and responds to the constantly growing information database. The longer it "learns," the more efficient these systems become at predicting demand and forecasting the perfect price to charge.
This empowers the revenue mangers by allowing them to react quicker and rapidly detect meaningful demand signal changes.
New Streams of Demand
The more data points that can be examined, the more accurate a pricing recommendation becomes as machine learning is able to filter the noise and turn the small bits of hidden value into user knowledge. For the hospitality industry, this is exemplified through undetected demand streams, or data sets that were simply ignored in the past, such as the rise of peer-to-peer accommodation sites. This is a prime example of deep learning technology unveiling the data layers that impact a hotel's overall revenue.
Clarity in a Murky Market
A smart revenue management system analyzes numerous market variables when optimizing room rates, including current destination supply, room rates of direct and indirect competitors, the aforementioned historical room rates, flight patterns, meteorological patterns and destination relevant events, among many others. All these factors together craft a clearer demand picture than ever possible before the machine learning era.
Sweating the Small Stuff
Capturing an understanding of demand generators, such as local event or major music festivals, and forecasting their impact can be invaluable for dynamic pricing.
Let's take any weekend of the year. There may be an inordinate amount of small to mid-sized events, and they're all bubbling just below the revenue manager's radar. Because they're hidden within big data signals, revenue managers may fail to monetize until it's too late. The machine learning based RMS will pick up the demand signal changes created by 10 smaller events happening on the same weekend. Those 10 events together could equal the demand change for one major event, so revenue managers are automatically alerted to this now major demand change. Then they can immediately shift pricing strategy for those dates. Previously, revenue managers were much more reactive, but today they have the tools available to help with predicting demand shifts, and adjust pricing after the fact.
The Intersect of Photography and Machine Learning
Before machine learning, property photography was never considered a demand signal. We may instinctually know great photography can make a property look extremely enticing, but that was more empirical, not based on science.
Machine learning accesses "Deep Neural Networks" and "Image Classifiers" to identify which specific parts or elements in the photo are leading to more bookings. For example, when a consumer clicks on a particular advertisement it is due to an attractive element. Their attention is focused on key aspects all while filtering out irrelevant information. Machine learning technology essentially works the same way. It will understand the "desires" that play a stronger role in the booking process and therefore will highlight to the hotelier where he or she can boost revenue by changing picture placement, highlighting certain USP's or flagging up offers at various stages of the booking process.
Should the hotel have a specific photograph on a particular internet booking engine the technology can evaluate the importance of that unique data set and then determine optimal pricing. This offers the hoteliers powerful information on how to use that image or one highlighting the same content to use it across other booking channels to further drive customer engagement.
Guest Generated Reviews
Guest generated reviews and content are typically one place customers seek out information. People love to know what others think of a property before committing to a stay. It provides a certain aspect of comfort. There's a wealth of information here revenue managers can use, such as understanding the relation between review site hotel ranking, and how that translates into a commanding a specific price.
There is a direct correlation between the reviews, price, ratings and loyalty. Machine learning allows us to not only see more relevant information but it weighs its importance in the optimization process.
In addition, machine learning technology allows us to study sentiment more closely by reviewing the percentage rate and rank of the review along with the general tone of the review.
Different travelers weigh value differently. Leisure travelers are more influenced by price and emotional language, while loyalty is far more important for business travelers preferring more factual reviews.
Perceived Personalization Where it Matters
Perceived personalization is where machine learning in the short term will have the biggest effect on our daily lives.
Machine learning technology is aggregating data from millions of people to help personalize items for the consumer. These data sets can show a particular person's preference and shopping behavior and categorize them into "personalized" brackets. Specialized categorizations are used to help influence recommendations for the consumer based on similar items they have purchased previously or items others have purchased that are still relatable to their interests.
In customer service, machine learning is already being used to answer the frequently asked questions. This allows marketing and operations professionals across industries to leverage the power of deep learning to tailor better messaging that resonates with the consumer and therefore increases the likelihood of additional items sold, or increased customer satisfaction as "you (company) get me (consumer)". For hoteliers specifically this means the ability to create accurately priced packages appealing to a subset of a property's customer base.
Machine learning has been around for the last twenty-five years, however industries are just now realizing the benefits this type of technology can offer to their business. We see adaptive technology interwoven in all aspects of personal and professional lives. Whether we're watching Netflix, debating how to forecast our next pricing index or encouraging customer engagement, it has become an invaluable asset to understand the true demand drivers across seemingly unconnected nuggets of information. Adopting machine learning technology in your own business means immediately gaining significant advantage over competition.
Mr. Bhandari is the Founder and CEO of LodgIQ™; a start-up dedicated to providing advanced revenue optimization technologies to the travel industry. Mr. Bhandari was the first-ever Head of Revenue Management for Hyatt International, and subsequently for Caesars Entertainment, and is credited with creating and leading the integrated discipline of Revenue Strategy, Marketing and Technology for Trump Entertainment Resorts. He also served as an Executive Consultant for Starwood Capital, where he advised on, and managed various aspects of business strategy and portfolio optimization for Louvre Hotels. Most recently, he was the Chief Commercial Officer for Nor1 Inc. Mr. Bhandari can be contacted at 646-453-7699 or showme@LodgIQ.com Extended Bio...
HotelExecutive.com retains the copyright to the articles published in the Hotel Business Review. Articles cannot be republished without prior written consent by HotelExecutive.com.