Back in 1996 when the Internet was still young, and users had to listen to the frequency shift key chirruping of modems singing to each other to access it, a Dutchman named Geert-Jan Bruinsma came across Hilton.com, which let you book hotel rooms online in the US, and decided to do something similar in the Netherlands; spinning up Bookings.nl to let hotels advertise vacant rooms in exchange for handing over a 5% commission.
The story had some twists and turns, but a few hard facts speak to how it turned out. Booking.com is big: record $27 billion in gross bookings in the first quarter of 2022 big; the highest quarterly figure in the company’s history. As part of Booking Holdings – listed as BKNG on the NASDAQ since 1999 and generating revenues of $11 billion last year, Booking.com is now available in 43 languages and offers over 28 million listings. Indeed, BKNG’s travel bookings of $76.6 billion in 2021 make it one of the biggest ecommerce platforms in the world.
That scale and its nascent “Connected Trip” plans to integrate more flights, attractions, car rental and other experiences into its offering and put them in a unified and data-powered way in front of its users (the company’s data science engines already handle over 12 billion decisions per day), have put unprecedented pressure on this early internet-native travel company’s traditional monolithic, on-premises infrastructure.
Booking.com CTO Rob Francis: Great engineering got us this far…
The man in charge of keeping the lights on whilst modernising Booking.com’s digital underpinnings is Chief Technology Officer (CTO) Rob Francis. Sitting down to speak with The Stack, the former director of corporate infrastructure at Amazon – where he led a migration of Amazon’s infrastructure from legacy data centres to AWS, before moving on to a CIO role at audio specialist Sonos – says it is “a pretty exciting time right now, because we are in the midst of reinventing our business; adding more verticals and experiences.”
As CTO, Francis is responsible for what Booking.com calls Central Technology.
“That’s a combination of things that are best done centrally. We also run things you find in a CISO organisation such as cyber and risk in addition to fraud prevention; the Chief Data Officer is also part of my team, so everything from Big Data to Machine Learning; core platforms that are part of the builder ecosystem – build, test, deploy, observability – support for languages and frameworks. We also started a new group a few years ago as part of our modernisation effort, called ‘Horizontal Business Services’”, he says, adding that under this rubric and as part of a shift away from a monolithic architecture “we’ve pulled out a couple of core capabilities that are business-agnostic: an order platform, and the identity platform for example…”
His priority is building “what we call the Connected Trip, which brings all those verticals and experiences together. The technology that powers all of that is particularly exciting for me. We’re known for our scale, but you can imagine adding all these other experiences brings a lot of additional complexity with it” he says, adding as one simple example: “If you’ve booked a taxi to pick you up after a flight, [we want] the contextual understanding if your flight is delayed to communicate to the pickup service or rebook a hotel.”
A careful greenfield rethink
So what does Booking.com’s current stack look like?
“We just celebrated our twenty-fifth anniversary as a company” Francis says: “A lot of really incredible engineering got us to this point from a scale and complexity perspective. Honestly,” he adds wryly, “in some ways, that great engineering and a culture of experimentation maybe took us further than we probably should have been able to get. We became world-class at managing our data centres, at MySQL, PERL; at a range of technologies that you would have chosen at the time – remember, there was no public cloud!.
“We’re on a journey” the Booking.com CTO points out: “I sometimes joke with my team that it’s hard not to make it sound like a TED talk: yep, we’re going from monolithic architecture to distributed services; we’re going from data centres to public cloud; we’re adopting new frameworks, new languages, new ways of doing things with Big Data and Machine Learning. So almost everything is on the table. It’s very greenfield; decades ago, we invested heavily in managing our technology, and we certainly got our money back on that investment. Now we’re doing it again, and that is particularly exciting given how much technology has evolved since.”
So where exactly is his team on that journey?
“I would say over the past two years we have really put in place the things you have to get right foundationally. There’s a common mistake companies make, jumping into the public cloud with a lift and shift mindset; that’s not something we’re looking to do. So we have put all of the controls in place from a cyber, from a cost perspective. What I’m really after is speed and control at the same time. To do that, you have to invest up front.
That prudent approach slowed further as the company tackled the sharp decline in business during Covid (as its Q1 results show, it has bounced back fast) and that careful pace has been challenging for his team, he admits: “One of our strengths as a company has always been our speed, an engineering culture, and our desire to go do things. It’s been a cultural shift for the company to slow down and get those foundational things in place. Because we have grown up as a build-first culture, the adoption of SaaS or IaaS is new for us. But now we are ready to spend more time on the business tech, as opposed to foundational enabling technologies: I’m confident with the engineering horsepower we have in the company, that we can pick up velocity quite quickly in the back-half of this year.”
That shift has included planning the migration of critical applications to AWS, the introduction of new big data and machine learning capabilities and more. What tips would he give for those on a similar journey? “I think sometimes one of the challenges when going through exercises like this is baselining where you are in the first place” he says, adding: “If you’ve reached a point where you have declared that you’re ‘going on a modernization journey’, you’re probably in a deeper hole than you thought. The biggest challenge of the first of couple years was just getting our arms around the situation and we probably did a few things before we were ready.
“For example, as I mentioned earlier a big part of our success was that we invested heavily in managing our data centres and the Perl and MySQL environment. When we got started building services in Java, we didn’t really have that feature parity, from the build-test-deploy side of things; the same is true with NodeJS. So now we’re playing a little bit of catch up in that builder experience environment as we work to make that a great experience for the engineering community that’s been so important to our success as a company.”
The Booking.com CTO also led a major architectural shift during his years at Amazon. Are there any key lessons from that experience he is bringing to his current modernisation efforts, we ask? He hadn’t particularly drawn a parallel, he says, thinking aloud about how to respond: “There are certain things Amazon does particularly well. Over the course of your career, you always take bits of everything from every place you’ve been… I would say that one of the things that I really came to appreciate at Amazon was the data-driven culture.
“This sounds slightly contrived but what I actually took from my time at Amazon is the key word in being data driven, is driven, it’s not data. Amazon also has that reputation for being very operationally strong. And they do drive their operations through data. In Amazon, they call them closed loop mechanisms; ways to be continuously validating that you’re still doing what you thought you were doing, rather than waiting until something has gone awry – and then trying to gather up all the data to figure out what happened.”
Talking of gathering data, a key priority he says is Booking.com’s “B.Fair” programme: “It’s crucial that we maintain a sterling reputation for prioritising our customers’ trust and online safety. That includes privacy and it includes high quality ethics around any decision making we make through Machine Learning. We’ve just kicked off an internal programme called B.Fair that is focussed on making sure we don’t have bias creep into ML models and we just joined a couple of external groups to ensure that we are well disciplined around this.”
For all his clear hands-on enthusiasm for the software and hardware stacks his team are working with, as well as the change management efforts, he admits that it’s “a bit of a surprise to me that I ended up in technology to be honest! I grew up in farmland in the middle of the US, and went to university not sure what I would study to be honest… But I sat in a hallway where people would check out cassette tapes to learn foreign languages. Two of the rooms had, I think there were Apple IIs, and 386 PCs, which Indiana University was experimenting with trying to teach languages on computers. One evening, someone was unable to make it into the computer lab so my boss told me to sit in the front and make sure no one stole any mouse pads. Just by sitting at the front of the room, people assumed I knew things. They would ask me questions, and I seemed to have a knack for it. My girlfriend at the time suggested I take a Computer Science course. That was some good advice.”
Booking.com is recruiting for hundreds of technology roles across Amsterdam, London, Tel Aviv and beyond. It’s a heavily engineering-led culture with some great perks. Check out the roles here.