Building bots for Messenger – Tips from the Experts: Skyscanner
We’re kicking off a series highlighting some of the top experiences on the Messenger platform. These businesses and developers have approached Messenger in a unique and interesting way, and are seeing success during the early stages of the Messenger platform. First up is Skyscanner, the world’s travel search engine.
In 100 words or less describe what your bot does.
The Skyscanner Messenger bot helps people travel through simple conversational search. Users can converse with the bot, ask for destination inspiration or find the perfect flight for them. They are then able to click through to partners’ sites to complete their booking.
Who are you trying to reach?
We are trying to reach travelers at all stages of travel planning. Users who are talking to their friends to decide where to go for their next holiday, or who are just dreaming about destinations and want to know what it could cost to get there.
What’s the killer feature/interaction?
As we use a natural conversation to answer travel questions, we’re eliminating the need to use forms and date pickers to enquire about travel questions – this makes getting answers to users’ search queries that much more simple. The bot also has an inspirational search feature which allows users to ask Skyscanner to provide them with destination suggestions, based on travel options that have been popular with other users searching and booking on Skyscanner. The data that powers this comes from Skyscanner’s Browse Cache API (which currently powers the popular ‘Everywhere’ feature on site) and the Travel Insight product which provides trends and market data based on Skyscanner’s 50 million+ monthly users.
How did you go about designing your bot?
We set up a dedicated team with people from across the business, in readiness for the announcements at F8. We had a plan in place – to tackle what sort of user problems we want the bot to solve, what the architecture would roughly look like, and how conversations would flow through the bot. When the API docs were unveiled, the team started immediately on building a a proof of concept to map out the various flows.
Meanwhile, we set about refining the conversational experience and the look and feel of the bot. What could we do with the Carrousel? How do we make the flow through the conversation as natural as possible, while still guiding people towards the goal of the bot – answering questions about travel?
Once we had our “Earliest Lovable Product” together that could answer simple travel questions such as, “I know where I want to go and when, please show me the cheapest options to get there and back”, we cooperated closely with our partners at Facebook to ensure that we refined the conversation experience.
What have you learned since launching? Any surprises or successes you want to share?
One of the key things we have learned is that a bot is a great way for people to be inspired. People want to look for options and for possibilities, to see, for example, when they can fly cheaply to a nice beach destination around the end of June. The way users interacted with the bot has really showed us that incorporating inspiration was the right direction to take the bot in.
If you knew one thing when you started, what would you say that would be?
The main thing for us was to keep the conversation simple and natural. Rather than pose a big list of questions up-front, we wanted to guide people through the conversation to help them come to the point where we know enough about their travel question to be able to give a good answer. In order to do this, and learn and iterate quickly, we started off with the simplest conversational flow that we could think of, and as we extend and build on that foundation, this principle remains the same.
What will you build next?
As we touched on above, our goal with the bot is to be able to give clear and simple answers to people’s complex travel questions. This means that we’ll work hard to be able to answer real questions that you’d ask your friends (“what’s a good, cheap city trip I can take next weekend?”). We’ll keep expanding the array of questions that the bot can answer, and build out the inspiration we can get from it. Once it is possible to include bots in a group conversation, we want to be there to help friends and families to plan their travel.
Messenger is such an ubiquitous communications tool in people’s lives that it made perfect sense for us to be there – it’s where people have conversations about travel, and our goal as Skyscanner is to help people travel. Messenger opens up a whole new way in which we can help people plan and book travel, right there in the platform that people are already using to communicate!