Machine learning in the travel industry – part one.

Inspiretec’s Head of Product, Luke Francis, gives us an insight into what Machine Learning is, how it works and how do we use it.

By Natasha Cook14th March 2019

What is Machine Learning?

Machine Learning (ML) is building programs that learn and improve from experience (patterns leading to an expected outcome) without being explicitly programmed to do so. The system is trained by being fed positive examples and data which leads to that outcome. Once the algorithm starts to recognise and identify patterns that create the same outcome, it can work on its own without a need for human input. The algorithm will only be able to operate on its own after feedback has been collected and the human’s function becomes to improve the accuracy, implement appropriate logic, and introduce better data.

Hard, logical rules are limited in their ability to solve real-world problems by scale. ML is flexible, adaptable and can operate at a scale or speed not achievable by a human. That having been said, this technology should be used to improve a process or a system - it’s not a solution.

There are different flavours of ML; reinforcement learning, statistical interference, probabilistic machine learning, unsupervised learning which all have different functions and goals.

How does it work?

ML is a revelation for making tasks easier to complete or for performing tasks at a much large scale whilst maintaining accuracy. ML can split large cohorts of people or users into smaller categories, which creates a much easier data set to handle.

ML recognises patterns from inputs of data – this process is called pattern recognition. As previously mentioned, the algorithm is trained to recognise positive outcomes, in travel that may be a successful booking. From these recognitions, the system can then recommend targeted and personalised offers to individual customers based on previous examples. The most common use of ML is exactly this – a recommender system.

Commercial applications of machine learning include; speech recognition, computer vision, bio-surveillance and robot control.

Challenges faced.

ML relies upon the data inputted, therefore, the quality of the data is crucial for the accuracy of the results that are presented. As there may be the issue where the recommendations provided would effectively be correct, but the data is inaccurate.

A lot of people assume that implementing ML will solve problems regardless of the circumstance. However, they are wrong. There must be a clear-cut purpose in place for ML to be useful and successful.

A key factor for using a ML based system can be the complexity of the system. By its nature, ML can be confusing, resource intensive and unpredictable. The challenge for the industry is to implement systems which do not require a user to be an expert, performs reliably and accurately, and where the cost does not outweigh the reward.

Another challenge is ensuring that the data is monitored so it is not perceived as ‘creepy.’ The key is to use data in a manner that always adds value or insight and ultimately benefits the person whose data is being used.

How do we use it?

ML is not implemented to solve issues, but to improve processes by allowing them to scale with accuracy. The main ways ML is used in travel are; price yielding systems, recommender systems, sentiment analysis, fraud detection and a few more. These can usually be put into two categories – detection and prediction.

Holistic, our travel CRM software, has an inbuilt recommender system used for personalisation. Personalisation can be anything from an email with a specific name on, rather than a generalised address e.g. sir or madam, to a “people like you also like this” level. Personalisation is key to enhancing customer experience and creating that intimate bond.

Holistic has been enhanced by ML making human interaction less necessary for small, simple tasks and enabling customers to have an optimised experience. Holistic recognises the “theme” – what the customer is looking to purchase, and the “stage” when the customer is likely to purchase. We use ML to attempt to improve the accuracy of these predictions. Then, reports and recommendations are made at each stage for specific users to cater to their specific needs. You can request a demo of Holistic here.

ML is present across the whole travel industry. It is used to enhance sales and to offer customers reliable or more accurate services. Translation apps perform astonishingly making travel easier for anyone. They have been able to develop in such a way due to the sheer amount of data available. Further, Google Duplex; a new, up and coming feature which provides telephone services. Google Duplex is extremely advanced with speech nuances which mimic human traits e.g. colloquial phrases like ‘um’ to indicate a humanised thinking process.

Key takeaway.

Remember – ML is used to improve a process or help scale a system, it will not magically create more revenue. A company wishing to embark on any such project should identify why it is needed.

With recommender systems, these are used to enhance a client’s relationship with a company, but this should not be done in isolation. Companies must create value in their products and their approach. User experience is crucial to creating a loyal customer and enhancing their overall experience.