What is common between buying an apartment and airline tickets, why your opinion is really important to a hotel and how to sell an “experience” to a user – read about this and not only in an interview with Pavel Velikhov, a lecturer at the Department of Data Analysis and Artificial Intelligence, who previously worked as Head of the Data Department Science.
How is Big Data being applied to the travel industry?
Tourism is a very vast area, there are giants like tripadvisor, booking.com, airbnb, and there are a lot of little players. Large companies have a huge number of tasks that machine learning helps to solve: from trying to predict conversion, ranking the output of analysis results and building a system of recommendations for users.
Machine learning is a method and technology that allows you to automatically find patterns in large amounts of data. For example, you can determine what characteristics of a hotel are for the satisfaction of users, or learn from the user’s travels to choose those directions for which the client will most likely buy a ticket.
Big companies have a lot of data: they have pictures, data about hotels, reviews. This data is often of very low quality. In this regard, they are working to verify information on the big data analytics services of the problem, and now all computer players are actively introducing vision systems. They allow you to recognize what is represented in pictures.
Also here are the tasks of determining the style and quality of photographs, from the hotel are sold better. There are scientific works on the detection of the quality of images, where the neural network selects the best photos for sale.
Another big challenge in the travel business, which is similar to that of credit scoring, is figuring out if a person will cancel a reservation. The fact is that from the moment a person makes a reservation to the moment of his check-in, it usually takes some time. This is similar to a situation when a person takes a mortgage to buy real estate and you need to predict the probability with which he will return this money to the bank. Before the trip, the client’s circumstances may change, and usually the aggregator or hotel already has some data about the client. There is also data about suppliers whose supplier has data whose history is being overridden. A significant task for a business, which allows you to significantly optimize the distribution of numbers, is to get in a timely manner what actions the client will take and minimize their losses, and on the basis of this data, you can build quite powerful predictive models.
But these are all the tasks of the big players. Smaller companies have a problem, because the conversion of users does not grow from the introduction of machine learning, the flow of customers is small and the significant effect of machine learning is difficult. There are a lot of such companies, for them advanced machine learning is not very useful yet, but classical analytical tools, optimization methods, and so on work well.
Define your client
Marketing in the tourism industry is a serious challenge, because clients are always different and from different cities, countries and even continents. Every business owner or manager constantly asks himself the question:
- How to expand your business?
- Where to open a new hotel, restaurant, etc.?
- What new services might interest my clients?
- Is advertising for my business effective?
The Big Data team at Kyivstar can give a clear answer to each of these questions. It doesn’t matter if a large company or a small local business asks about it. A mobile operator with almost 26 million subscribers offers to build a customer portrait based on big data analysis.
How it works
A portrait of a client requires a certain set of his impersonal characteristics. Namely:
- floor;
- age;
- income level;
- social status;
- marital status;
- the presence of children;
- place of residence;
- how often the client travels abroad
- whether he is an auto;
- what are his habits, interests.
A well-formed client portrait helps to make the offer more relevant.
