Mike Hanrahan, CTO and co-founder of Jet.com. Image: David Solodukho
“The talent pool in Dublin is incredible,” said Mike Hanrahan, the Wexford native who co-founded e-commerce giant Jet.com.
Wal-Mart acquired Jet.com for $3bn, making it the largest US e-commerce company acquisition in history.
Jet.com is a gold sponsor for the upcoming Career Zoo event in The Convention Centre, Dublin on Saturday, 11 March.
Mike Hanrahan oversaw the building of the Jet.com platform, which, within 10 months of launch in 2014, had a revenue run-rate of $1bn.
The company has grown from 30 people in 2015 to 2,000 today. It has more than 8,000 suppliers, its own distribution centres and 20m stock items.
What roads led to Jet.com?
My background is in chartered accountancy. I worked at PwC in Waterford and then London in the mid-1990s, and that’s where I met Marc Lore, our CEO. Marc went back to the US and had a couple of successful start-ups, including a company called Diapers.com.
When Marc got back in touch with me, I was working in various banks, building systems, and he asked me would I like to work with him again.
We jumped at the opportunity and decided that a move to the US would be an exciting adventure. I always liked working with Marc. He is a super-smart guy, a very inspirational leader, and so we went to the US and I worked again with him at Quidsi.
I took a few years off because we had four kids under the age of five. Marc got in contact again and told me he was thinking of starting a big e-commerce company to take advantage of what we thought was a big disparity in e-commerce, relative to what you might see in Europe or Japan.
Very few US consumers did their weekly shop online; people were using Amazon to engage in e-commerce but were typically buying one item here and there, not thinking of the weekly grocery shop.
As a consequence, shipping costs for a single unit were very high – 16-20pc of order values went into fulfilment.
In the UK, there are very different price dynamics as people buy big baskets of stuff when shopping online and, as a result, fulfilment costs come down to 5pc. That’s a big disparity.
There was almost 10pc that we felt we could take out of the value chain in terms of cost. That was the genesis of the idea. We were obsessed with how we could build a website that could encourage people to buy bigger baskets and really help them to pull costs out of e-commerce so they could shop in a sustainable way.
What are the main points of your IT strategy?
From a technical perspective, if you speak to any CTO and ask them to describe their dream job, Marc gave me mine.
He sat me down, explained the amount of capital he was raising, when we were going to come out of the gate, how many visitors we would hit on day one and how he planned to grow the hell out of this thing. His plan was for it to become a $20bn-dollar-a-year business by 2020.
The technical requirement was fantastic. It was a complete greenfield set-up with no legacy systems, and we hadn’t written a line of code.
The challenge was to build a massively scalable e-commerce platform as good as anything Amazon or Wal-Mart had.
We had to make sure we were building it at scale and for the cloud, and thinking of all the engineering challenges.
We made the decision to do it all right at the start. Let’s not try and build a minimal viable product, let’s do it right.
There is nothing sweeter to the ears than when you hear that kind of technical requirement.
Marc himself is very cognisant of the power of technology and, as CEO, he was a fantastic champion of technology in the sense that he really gave us the latitude to take a lot of technological risks to get where we needed to be.
We ended up with 300pc the traffic we expected on day one, which was better than our best case and the systems scaled fantastically.
We didn’t have a single glitch from a technology standpoint and so it all came together quite nicely.
How complex is the infrastructure? Are you taking steps to simplify it?
We have an unusual platform in a lot of senses. It is an event-driven platform that has the properties we need to do what we want to, in terms of scaling and the cloud.
We use event sourcing as a system of record, which was unusual and scaled for us. It is extremely flexible and can be scalable if done right.
We have a very strong focus on microservices. We have more than 1,000 microservices in our production environment right now. Each piece of code is nicely encapsulated in small services, nice properties for scaling and decoupling the system.
We get fantastic utilisation out of our machines because they are cheaper to run.
We were native to the public cloud. We took a big bet on Microsoft Azure and that worked out perfectly for us.
Our systems run 100pc in the public cloud and to do that effectively really requires a deep understanding of the nuance of the underlying fabric.
It came with a lot of challenges but when we came out the other side, we were in a fanatic position with the foundations of a massively scalable business that would work.
Do you have a large in-house IT team?
Since the Wal-Mart takeover, I am responsible for some of the technologies in the greater Wal-Mart world as well. But within the context of Jet.com, I have around 200 people reporting to me right now but we are looking to double the team over the next 12 months – 180 people between Dublin and Hoboken. I also work with 300 people on the west coast in the Wal-Mart context.
Jet.com has been in Dublin for the last year. We have close to 20 people on site here and we’ve had fantastic success here so far.
Some of the thinking around Dublin was: ‘How do we have a team that is not based on the east coast of the US, who can help diversify our engineering and hiring, and help with doing some of the risky releases we do overnight that happen at 4am east coast time, which is 9am Dublin time?’.
Dublin really works from that perspective. We also wanted a team on the ground contributing to the larger platform, a team that would take ownership of significant projects at Jet.com.
The reason we did it this way was to attract the calibre of staff who want to feel they have a seat at the table of the greater Jet.com technology group.
We were very mindful from day one to make Dublin an equal partner to Hoboken, and that they would get the meaty projects and fun stuff to do, and not just view themselves as an ops team.
What do you think of the talent pool in Dublin?
The talent pool in Dublin is incredible. Every major tech company in the US has a presence in Dublin, from Microsoft to Google and Twitter. We typically attract talent from all over Europe as well as Ireland.
I split my time between the east and west coast of the US.
Jet.com is small enough in the greater Wal-Mart picture that we have plenty of flexibility, in terms of the numbers of people we hire. We have fantastic offices on Merrion Square to grow the team very quickly.
We are all systems go.
What other projects do you have lined up for the year, and what will they contribute to the business?
We are only getting started on what we are trying to achieve here.
Amazon is obviously massive and Jet.com is a minnow by comparison, so we have many engineering challenges that we are going to have to focus on in the coming years.
Jet distributes its data across a number of data centers and geographies so a strong understanding of distributed systems theory is important for our platform engineers.
We are doing a lot in the States around personalisation, taking huge data sets from machine learning and figuring out how we leverage for a better customer experience and apply that across different problem spaces. Machine learning will be applied to search along with personalisation and recommendations.
Another big engineering challenge will be around how we match products. It sounds simple on paper but is, in fact, one of the biggest problems in e-commerce. We accurately match 70m products and, at any one point in time, 25m of those are for sale on our site.
There will be a lot of engineering work and data science work trying to improve matching algorithms.
If you really want to leverage and get value out of data sets, you have to get really good at handling petabytes of data in ways that make sense.
