Passenger Management and Experience: From Netflix to Railfix
‘Alexa, buy me a train ticket to Paris’ may soon be a valid way of purchasing a train ticket to Europe, according to Eurostar. Intelligent assistants like Alexa, Siri and Google Assistant are just one way in which machine learning (ML) is transforming customer services on a massive scale, and enterprising rail companies are already embracing the new possibilities it affords.
Simplifying and speeding things up
ML has all sorts of possibilities for improving the customer experience. Chatbots are one of the most prolific examples of this right now. A chatbot can field many of the common questions and easy tasks that were formally the realm of human operatives. With the quality of hold music seemingly something that no amount of technological intervention will ever improve, chatbots represent a fast and easy way to facilitate customer engagement. They have a role to play in improving the customer experience, leaving human agents free to handle more complicated customer problems.
Chatbots certainly had clunky origins. However, ML has enabled a new wave of highly sophisticated autonomous agents to emerge. These are capable of understanding human speech with its nuance and approximations, and respond in kind. Some chatbots are so good you might not notice you aren’t speaking to a real person. Google showcased this advancement this year with a demonstration of their virtual assistant booking a haircut for a human client over the phone, easily navigating a linguistically complex and unpredictable conversation with a completely natural-sounding delivery. Natural language processing (NLP) is the key capability which has led to these advances. NLP is the branch of ML which concerns translating human speech – accents, idioms and all – into something a computer can understand. We discussed the basic components of NLP already in this series in our article on safety analysis. Key to this is understanding the associations between words, the significance of their ordering in a sentence and their relationship to one another. ML allows us to impart this understanding to a computer, without having to teach it English – or any number of languages for that matter – from scratch. Understanding and responding to human speech in real time is something these systems are just now becoming capable of. For rail, this could mean new means of supplying customers with tickets and bespoke information that might soon be part of customer’s expectations.
Understanding your feeling and your needs
Our customers’ needs aren’t limited to acquiring information or making purchases – what else can ML do to make the customer experience something worth coming back for? NLP, in combination with sentiment analysis, is increasingly being used to understand what customers like and dislike. Sentiment analysis takes text or speech and seeks elements – including words, tone, volume and emojis – as indicators of what the user is feeling. This can allow companies to ‘put their ear to the rail’, as it were, of the whole of social media and work out where they are doing well or poorly. This can help companies evaluate the customer experience they provide and make improvements.
ML doesn’t just help us understand customers, it can also introduce entirely new services which add value. Customer service robots recently entered the UK rail network for the first time. Already deployed in Japan to help customers in busy Tokyo Stations, the Hitachi Emiew3 robot now also inhabits the St Pancras International departure lounge. Under the name of ‘Pepper’, the robot provides travel and destination information, but also entertains children and poses for selfies. Robots could also improve accessibility on the network. Combining elements of ML such as NLP and image processing, customers who need assistance can be aided on their journeys, shown through stations, and assisted with ticket purchasing. Facial recognition could even be incorporated into the bots to enable them to recognise people they have been notified will want assistance. This could offer a huge benefit at both busy and unmanned stations. The use of ML to deliver a personalised experience can be as simple as recognising and greeting returning customers and offering them a discount on their usual morning coffee order, to one day summoning their autonomous vehicle to the front of the station as they disembark their train.
ML integration into rail customer service can start now
The beauty of the adaptive nature of ML allows tools developed in one industry to be used in another with ease. Implementing a service chatbot to handle basic queries and transactions doesn’t need a bespoke team of computer scientists on payroll. Rudimentary sentiment analysis is possible with free online tools. Google also supplies a range of building blocks to developing bespoke tools through the Cloud Machine Learning Engine such as the Cloud Vision API, Cloud Translation API and Cloud Natural Language API. Customer service is an excellent area to test the waters of introducing ML tools to a company: there are low psychological barriers to their use and ready-made solutions are easy to acquire. Although there will be a learning curve and initial errors with introducing such tools to any field, in customer service it is possible to manage these through gradual introduction and beginning with areas where errors are acceptable to customers. TfL did exactly this when they released their ML customer service bot on Facebook in 2017. ‘Travelbot’ was released quietly, to give the developers time to test and tweak its performance. This meant that during the introduction phase, a high quality of customer service could still be retained. The bot uses NLP to process requests and send friendly responses, derived from the wealth of information in TfL’s data API.
Streamlining queries and transactions to make travelling as efficient and painless as possible is a growing customer expectation. ML can unquestionably help us deliver this. However, with all the advantages offered by ML, we must not forget that the role of humans is still critical. Some of us prefer to talk to humans on all matters, whether that’s queueing up to buy a ticket or asking a real person questions on the phone. More importantly, facilitating questions and transactions is but one aspect of good customer service. People who are in distress need human help, and unusual situations can often only be resolved by human intervention. The real value of customer service is empathy, and for the foreseeable future, this will remain a distinctly human quality. The Rail Industry Customer Service Awards are handed out every year to customer facing staff. The winners are not commended for their speed, cost or accuracy, but for their humanity. Last year’s winner coaxed an intoxicated and injured man from the tracks he had strayed onto to safety. He also befriended a man with learning difficulties, and because of this was able to alert emergency services to a concern he had for his welfare. Other nominations went to staff who ‘left train loads of passengers smiling with witty, brilliant announcements-. What ML will do for customer service is free up humans for the tasks like these which take a service beyond ‘fine’ and make it ‘fantastic’.