Heineken and Oracle create chatbot helper for polo events

Most people know Heineken for its beer, but as part of the brand’s annual Urban Polo event this year, it has dabbled in the world of chatbots to help people navigate through its combined music and sporting festivals in New Zealand.

The chatbot, which uses machine learning to continually improve its interaction with consumers, was built on Oracle’s artificial intelligence (AI) platform and is used to answer some of the more frequent questions asked by event attendees.

“Where is the bar?”, “When is the next match?”, “What are the latest scores?” and “What song is the DJ playing?” are just some examples of commonly asked questions given by Simon Wilson, owner of the Heineken Urban Polo events.

Wilson says part of the reason the event chose to trial the chatbot was to introduce a “new concept, a new idea” to a sport that largely hasn’t changed over the past 200 years.

By introducing technology to an “environment it doesn’t usually exist in”, Wilson says the event is “turning polo on its head”.

“It’s definitely adding some excitement for the spectators,” he adds.

The Oracle AI in the cloud platform for Urban Polo’s chatbot was built using natural language understanding (NLU) to interpret conversations and respond with a contextual answer. It uses machine learning to improve over time and extend the number of questions that can be answered, choosing more appropriate answers in the future if necessary.

People attending Urban Polo events can access the chatbot through Facebook Messenger, a decision Wilson says was made because social media is already often used by event attendees, whereas it can be difficult to get people to download a separate app for something that happens only once a year.

He also points out that there are rules governing how companies are allowed to communicate with customers, and consumers have to send the first message.

“You can’t just have them click on it and then we start hitting them with messages,” he says.

But this shouldn’t pose too much of a problem. Social media is becoming a huge part of sporting events as millennial customers seem more interested in posting about their experience on social platforms, often making the event secondary to their social profiles.

Chatbots are also increasingly adopted by firms to reduce costs by answering some of the questions that may otherwise be directed to a call centre, and to offer a better customer experience than sifting through a FAQ on an event website.

Research by Heineken’s partner firm Oracle has found that 80% of brands expect to be using chatbot technologies by the year 2020.

As far as Heineken Urban Polo is concerned, Wilson says: “It enables attendees to get all the information they want, when they want it – in a way that is very familiar to them – through the messaging tools they are already using.”

Wilson adds: “One of the things that breaks my heart is you leave the event at the end of the day and programmes are on the floor.”

Using the social platforms that people are already using, instead of a printed programme, is also a “better way of showcasing our sponsor brands”, he says.

Many attendees are already using Instagram, which is where many people are directed to the chatbot platform. Pictures are a huge part of the event, says Wilson.

“People get dressed up and spend months planning for it, so the photos are the biggest part for them.”

People asking where the official photos are kept could be directed to the event’s official website by the chatbot, one of the ways the AI became part of the overall experience of the event.

“People get dressed up and spend months planning for it, so the photos are the biggest part for them”

Heineken Urban Polo trialled a version of the technology in 2017, but found fewer people engaged with it than the organisation would have liked.

“But the people who were engaged and were inside the chatbots loved it,” says Wilson.

To make the chatbot more successful the next time around, Wilson says different ways were tried to engage people with the chatbot ahead of the event so they were familiar with it.

Wilson realised that the chatbot, which correctly interprets more than 90% of the questions put to it, is only as good as the data that goes in.

Many questions were asked repeatedly by consumers during the trial, and Wilson made sure these were already in the chatbot’s database for its launch in 2018.

“There were lots of questions we knew attendees were going to ask, and I made sure they were in the machine learning before they started,” he says.

Before the 2018 events, Oracle approached Heineken Urban Polo, and Wilson says the firm was excited at the opportunity to develop the AI platform, especially because when a technology is developed on the Oracle platform, you own it once it is finished, avoiding future expansion costs.

“We were excited about it – I had already been trialling some tech with a different provider,” says Wilson.

“All of these models like MailChimp, the bigger you get, the more successful you get, the more it will cost you.”

The machine learning aspect of the chatbot is a “key part” of making sure the selection of questions and answers available to the chatbot grows, eventually making it a low-maintenance investment, says Wilson.  

“It does work – we saw it working,” he adds. “The more data you can load into it, the better it is. Ideally, the machine learning is going to take care of that – and that’s the dream.”

“The more data you can load into it, the better it is. Ideally, the machine learning is going to take care of that – and that’s the dream”

Next year, not only will Heineken Urban Polo try new things with the chatbot, but it will also move into new regions, starting with Singapore in 2019.

Language will be a key factor when designing new features, as Wilson’s teams found people were using conversational language to talk to the chatbot, asking things such as: “What time does it finish, bro?”

Wilson says: “We need a bigger [language] selection to get better at that.”

As well as incorporating more slang into the chatbot’s database so it can understand questions better, Wilson says more is needed to help answer queries about the location of its four events, so that people’s questions can be answered more specifically.

This could be helped along if there are repeat customers who have attended the event and used the chatbot before, making it easier for the AI to answer these questions in a personalised way, based on what the customers were looking for in previous years.

Wilson says Heineken Urban Polo will try to get people using the chatbot before the event, so they are aware of it and know what to do with it before they go. A QR code for the chatbot platform included on tickets, beacon technology and push notifications were all cited as ways people could be directed to the bot.

Using the Facebook Messenger platform has some advantages, but it also has its flaws, says Wilson. “People are trained on Facebook to ‘like’ and ‘share’, but not to message,” he says, which can make it difficult to direct people to the chatbot.

“We’ll do better next year,” he says. “We want to simplify the entry into the chatbot. If we can really nail this, the experience is awesome.”