When Steph Curry Has Your Back: The SnapTravel Story

It’s not everyday that an NBA great funds your company, shows off your swag at the NBA Finals, tweets about you to his 13.7 million Twitter followers, and pulls you on stage at Disrupt SF, one of the most popular tech startup events of the year, but that’s exactly the dream that SnapTravel co-founder Hussein Fazal is living these days.

“Since we launched two years ago, we’ve booked over a million nights and now have well over $100 million in sales. We have grown 300% year over year and are projected to get to a billion in sales by the end of 2021, putting us in the category of one of the fastest growing travel companies in the world,” Fazal told me as I prepared to interview Curry and his investment partner, Bryant Barr, backstage at Disrupt SF.

I first met Fazal and his co-founder Henry Shi in July 2018 when they were visiting Silicon Valley to scout out venture capitalists investing in conversational commerce which I was writing about at the time. It had been a year since Mark Zuckerberg announced that SnapTravel had hit their $1 million sales milestone during the Facebook F8 keynote, and they were excited to share some impressive new numbers. Fazal, a monetization guru who co-founded AdParlor, Facebook’s largest ad platform, and Shi, a former Googler who had once developed for Bloomberg predictive analytics for sports teams, made quite a team. The two University of Waterloo graduates had been working on a fix for the user experience nightmare of online travel agencies which were plagued with pain points throughout the daunting number of web pages and fields users needed to navigate to simply to book a hotel room. To make the process more intuitive, Fazal and Shi moved the conversation to messaging platforms where people could chat with their AI-powered hotel booking assistant to book a room.

In the process of raising their Series A, Fazal just happened to be in the right hallway at the right time when he got pulled into a meeting with Bryant Barr, president of Steph Curry’s investment arm, SC30. Soon after the meeting due diligence ensued. Helped by the fact that SnapTravel had a major lead investor, Telstra, the venture arm of Australia’s largest telecom, which SC30 could lean on to vet the deal, and other respected investors like Peter Kern, vice chairman of Expedia, on board, made it an easy bet for Curry and Barr. Fazal attests it’s been a fairytale ever since, “I’ve been surprised at how involved Steph has been. He posts about us in his Instagram stories, involved us in the SC30 Partner Summit just held in Napa, and is genuinely interested in our business and technology,” Fazal told me as he described how the name recognition of Curry as the greatest basketball player of all time, beloved by millions around the globe, has opened doors where the startup was not previously known and has helped them build a trusted consumer brand.

I had a chance to catch up further with Fazal about their AI roadmap. What follows is an edited transcript of our conversation: 

Sounds like you’ve cornered the market since we last spoke. Who are your biggest competitors on Messenger and WhatsApp these days?

Fazal: Expedia and Booking.com are our main competitors on Facebook and they also put a lot of effort into voice. We’re also integrated with Alexa and Google Home but have found that where voice is a great way to start your search, it’s not necessarily the best way to complete a booking. Right now, you can say, “Hey Alexa, I need a hotel in San Francisco next week” and she'll come back and say, “The best deal we have is W San Francisco, do you want me to send these hotel deals to your phone?” If the deal is from SnapTravel, once you give permission to share your phone number, we'll send you a text to continue booking over SMS.

How does Alexa know to send hotel queries to SnapTravel vs Expedia?

Fazal: Yeah, that's a good question because most people are not going to ask for SnapTravel deals, they’re going to ask how much is the W Hotel in San Francisco. That’s just how people talk and it’s something that Amazon and Google are working on in terms of trying to figure out the best way to leverage things like “implicit invocation” and “canfulfillintent” in the code to help surface the right skill at the right time to drive conversion. It would be great if they also tracked the booking process until completion. So, for example, if the first search they offer is Expedia and the next search they offer is SnapTravel, they should note whoever is getting the booking, and that service should get the search the next time. 

What percent of your business is coming from Alexa and Google versus Messenger and WhatsApp?

Fazal: Messaging is 99% of our business, Voice is under 1%.

Can you pay Google to surface SnapTravel when somebody does a query?

Fazal: That would be great, but right now all you can do is build the best experience possible.

How dependent are you on Facebook?

Fazal: In addition to Messenger and WhatsApp, we’re also on Slack, Viber and SMS so we are pretty diversified in terms of messaging platforms. 

Your assistant still seems relatively transactional. Are you planning on humanizing its interactions to reflect more natural conversation? For example, at some point, will you be giving your assistant a persona or name?

Fazal: In a very early MVP of our product we used the name Emma. What we found is that the brand recall was very low. Customers couldn’t remember if our service was Emma, Amy, Jane, Lucy or any other common short name. It also made searching for us very difficult as we were competing with real people who have the name Emma. SnapTravel is much easier to recall and satisfaction has been pretty high - 94% of our bookings happen without any human having to get involved and 70% of our customers re-engage with us within six months - so I think we’ve achieved the right balance of chatbot, human agent and UI to achieve a conversational experience that our users find delightful.

Are you storing my data and how are you using it to inform the AI? So for example, if I booked my CES room with SnapTravel, would it then ask me at the end of my stay if I’d like to be offered deals to rebook next year?

Fazal: We collect data on everything that you do. When you make a booking, you're giving us all this information: your first name, your last name, email address, your phone number, your credit card, check in date, check out date. 

We use machine learning to analyze that data to present personalized recommendations. The question is how much of that data is relevant and how much can we leverage it. 

Right now we don't supplement our machine learning model with that extra layer of data, but in theory we could. We currently infer from your history, so we may not know that you go to CES every year, but we do know that you go to Las Vegas and have an automated system that will ping you a couple of months before your next anticipated trip..

We’re building the ability to use messaging as an operating system to drive bookings and conversion, while minimizing spam, and finding that there’s a fine line: How many times do I follow up with a customer, is it two times or three times? Do I use an emoji or do I not use an emoji? It’s a pretty intricate operating system learning how to leverage messaging to drive commerce without being annoying. What we are seeing is that when we send out messages that are targeted and personalized, they work extremely well but mass blasting a generic message has the opposite effect. So for example, if we ask “Do you want us to help you find the deal a few months before?” that would work extremely well. But if we mass blast everybody and say, “Summer is around the corner. Book your summer vacation home now,” well that’s just an ad. 

What advancements in machine learning and natural language processing are you hoping to see in 2020 and how will that improve the SnapTravel experience?

Fazal: We are quite excited about the continuous improvements in deep language models (like BERT and ELMo) as well as deep transformer models for text understanding and generation. These improvements mean that our customers can use free form natural language over messaging and the accuracy of our responses will continue to improve. 

What is your UA strategy? 

Fazal: We spend a lot of money as you would expect us to spend on the same places as other travel companies - Facebook, Google Display - we do a lot of performance based marketing. We also find organic social sharing drives a lot of traffic, “Click on the share button and send a message to 10 of your friends on Messenger that you just booked your trip!”  For conferences, we have a presence at the large travel ones like PhocusWright.

Airbnb recently acquired HotelTonight, do you now see them as a competitor or potential exit?

Fazal: Actually, they’re a potential partner. It would be great to offer deals on Airbnb. We've been chatting with them about working together. 

Is there an advantage being a Canadian company in the tech space and, in particular, headquartered out of Toronto?

Fazal: Yes. The biggest thing for us is the world class engineering, AI and machine learning talent, particularly access to disciples of Geoff Hinton, he godfather of AI who is based in town, works at Google and lectures at the University of Toronto.

You’ve raised $21.2 million, what are you going to be using it for?

Fazal: Fortunately, we have found very clear product market fit which has enabled us to grow at a rapid pace, but in order to continue at this growth rate, we need to innovate and maintain our leadership position. This may mean going deeper within travel (white-glove service offerings, flights, events, tours, activities) or it may mean expanding outside of travel. Regardless, at this scale, all of this is only possible with the right investment in technology. We are doubling down on machine learning models across the organization, improving our NLP engine for quicker and more accurate chat responses, strengthening our pricing and recommendations engine to drive sales, and taking a holistic view at building out a more personalized experience.