The SnapTravel team uses machine learning and the Front API to create a bot and stay on top of all of the customer demand and inquiries within their Front inbox.
Finding travel accommodations can be exhausting — especially when you’re trying to strike a deal. SnapTravel is changing that by making the process of finding and booking a hotel as easy as sending a Facebook message.
Instead of scouring the web to book a hotel room, customers can send a Facebook message or SMS text to SnapTravel about what they’re looking for — something like, “I need a hotel room for 2 in Cabo.” Using Front to manage their messages and a bot to give automated responses tailored to each customer, the SnapTravel team provides suggestions for rooms at highly discounted rates. They manage 50,000 customer conversations every week, all with a customer service team of only 30 agents.
We spoke with CEO and Cofounder Hussein Fazal to learn how they use machine learning and the Front API to stay on top of all of the customer demand and inquiries within their Front inbox.
What inspired you to build SnapTravel?
Fazal: The current process of booking hotels is challenging and frustrating. We believe booking the perfect hotel room should be as simple as messaging a friend. At SnapTravel, we’ve taken the speed and convenience of booking with an OTA (online travel agency), plus the benefits and service of booking with a traditional travel agent, and added a sprinkle of messaging magic to create the absolute best way to book a hotel!
What made you decide to use Front to manage your customer messages?
Fazal: Front allows us to easily monitor and interact with thousands of conversations at once. When a customer begins a conversation with SnapTravel, they are interacting with a bot. At some point in the process, they may be directed to a human agent on our team. A smooth transition is critical, so the human agent on our team can get all the context from the customer’s previous interactions. With Front, we can make seamless and smart handoffs from bot to agent, and between agents. With internal commenting, reminders, and rules, Front ensures that we provide the customer with a delightful, conversational experience.
Walk us through what happens when someone chats SnapTravel.
Fazal: SnapTravel can be used over SMS text or Facebook Messenger. The user begins by interacting with the SnapTravel bot and is asked some basic questions (city, dates, budget) which allow us to present the right hotel deals automatically. If the user types something the bot cannot understand, or if the user explicitly asks to chat with an agent, we automatically hand off the conversation to a human agent via rules, the automated action triggers in Front. From there, our team can help customers find the perfect hotel by sending over personalized recommendations and answering any questions. All this is done without ever having to pick up the phone to call or having to download a custom app.
How do you use your Front plugin to help your bot learn over time?
Fazal: The Front plugin we’ve built allows our team to categorize messages with customer data, which fuels our machine learning model. The long term vision is to have the SnapTravel bot intelligently answer any customer inquiry, so our human agents have fewer conversations to handle manually.
In order to do this, we’re constantly using Front’s tags and a custom Chrome extension we built to organize customer messages. We tag data on every message, answering questions such as, ’What is this customer looking for?’ As we process more conversations and collect more data, our bot gets smarter and can reply to more customers automatically.
We tag users who are VIP’s (customers who make several large bookings with us) to ensure they receive priority service. We also tag users by marketing channel so we have an understanding of how that customer found our service.
What metrics do you track?
Fazal: Using Front, we track first response time, average response time, and accuracy of responses. When a user comes into SnapTravel, they are initially engaging with the bot. This means they’ll get a response in under a second! When the conversation is handed off to a human, we aim for an average response time under 2 minutes.
How do you maintain a response time of 2 minutes or less?
Fazal: We use Front’s message templates— saved, pre-drafted messages that our team can easily search for and send again with a few clicks. We have about 30 message templates prepared for our most common customer questions, and we categorize them (i.e. booking, upgrade request, cancellation policy...) to make it super easy for agents to get the right answer to the customer as fast as possible.