The Data Team Behind Instacart’s Hyper-growth
Backed by a $220M C-round, Instacart is looking to grow its cross-functional data science team by 200%
This portrait was commissioned by Instacart and produced by Job Portraits, which highlights job openings at Bay Area startups. For the interview below, Job Portraits spoke with Instacart Cofounder Brandon Leonardo, data scientists Eric Rynerson, Houtao Deng, Deepak Tirumalasetty, and Mathieu Ripert, as well as Logistics Team Lead Andrew Kane and Senior Technical Recruiter Tim Wong at their San Francisco office, near Bryant and Second. Check out their data science and analytics positions here.
What is Instacart and how are you making the world a better place?
Brandon: Instacart is a same-day grocery delivery service. We have an army of personal shoppers who shop for you at stores near you. So you can pick your local Whole Foods, or your local favorite, like Bi-Rite, and have groceries delivered to your door in one hour, two hours, or within a timeframe in the next seven days.
Why does this matter? I remember when we were first starting Instacart, people thought we were this service for really wealthy people who could throw away money on having a personal shopper. But we do it really inexpensively — $3.99 — so a lot of our customers are parents. I see my friends with kids and going to the grocery store can be a four-hour trip for them. We give them back that time to spend with their kids.
“We give [customers] back that time to spend with their kids.” — Cofounder Brandon Leonardo
Other people who are positively impacted by Instacart are blind users. They use our iOS app, which has voice over, to quickly find what they need and have someone pick it up and deliver it to them. We also help a lot of elderly people who can’t get around.
Then, on the flip side, we employ over 7,000 personal shoppers who are able to work truly flexible hours. A lot of them are students or using it as way to make some side income. So we see it as a good thing all the way around.
Did any of you connect with that mission when you were considering coming to work here?
Eric: One of the things that was really compelling was that customers just love Instacart. A lot of companies make a lot of money, or are also funded by Andreessen Horowitz, but they don’t have the loyalty of our customers. On Twitter people are always like, “Oh my God! It changed my life.” I thought it was cool to have the chance to work for a company that is making people really happy and providing an extremely valuable service.
Where does Data Science fit within the larger Instacart organization?
Andrew: At Instacart, we organize teams around business goals. All the teams are interdisciplinary. Right now all our data science is focused on logistics, but we plan to have data science in every engineering team.
Some companies like to have all the data scientists on one team by themselves. But we find if you put engineers and data scientists together on the same team, it makes it a lot easier to work together and allows you to move faster.
Let’s talk about the specific roles you are looking to fill.
Tim: We have four data scientists now and we’re looking to add nine more. There are a lot of interesting challenges that our team is facing right now, generally in three areas. First is a focus on operations, research, and logistics. For instance, we want to make sure we hit our delivery window, but as we scale the demand for our shoppers has gone up. Part of how we addressed that was by introducing variable pricing. Another example is batching, which is basically where we group similar orders based off the locations of our shoppers, the grocery store, and other customers. Our data science teams are looking at all that GPS data and helping optimize the supply chain logistics.
The second area could be described as growth and user behavior. We want to get to a point where we can predict, for instance, the likelihood a shopper will order something a second or third time, and the time between each order. Or we want to look at retention. What’s the negative value of either a late order or someone giving one star?
The third area focuses on our recommendations engine. Search is essential to what we’re doing here. That feeds into item stocking, and that feeds a little bit into user guidance as well. We want to be able to make helpful suggestions based on what someone previously purchased. So, if someone buys a Swiffer, for example, we might suggest they buy a Swiffer replacement head as well.
What kind of background or experience are you looking for in a new team member?
Eric: I would say we are looking for people with a strong statistical background, who also have some coding ability and experience. We’re looking for people who really want to push production code. Someone who wants to be able to build a feature themselves and get it in front of customers.
Mathieu: It’s very important to be able to code because we are at the stage where we aren’t just sitting around analyzing data all day long. We are also building and implementing models.
Deepak: I just want to add something. Technical abilities are definitely important, but it’s also important to highlight that we’re growing really fast, so the problems we deal with change every week or every month. It’s important for a candidate to be really comfortable with that. It’s been a hell of a ride for us for the last six months. It’s important to have the character to enjoy such a ride. Not everyone is comfortable with that.
Is there personality that might not be the right fit on this team?
Brandon: If you’re arrogant.
Eric: Or a perfectionist. It’s a terrible place to be a perfectionist.
Why did you decide you wanted to work here?
Deepak: I think I was looking at for two things: opportunity and challenge. Looking at the information Instacart is accumulating, I don’t think any other company in the world has such quality and granularity of data in terms of grocery shopping. There are a lot of opportunities to deliver business values from this data. That’s the first one. Second, in my last six or seven months of working at Instacart, I have worked to solve really challenging problems that nobody has ever faced before.
Mathieu: I studied mathematics and operations research, so I’ve always been in this world, but what I really like at Instacart is that the problems we’re trying to solve are very real. I used to work for banks, where I built financial models that I didn’t really understand. Here we are using complex methods and tools, but to build something useful.
“I really like at Instacart that the problems we’re trying to solve are very real.”
— Data Scientist Mathieu Ripert
Deepak: In terms of impact we can have on the company, our work can mean a good or bad experience for hundreds of thousands of customers. So it’s a very direct impact that we can sometimes see within minutes of making a change.
Houtao: I have three reasons I wanted to work here. One is the vision, which is clear and revolutionary. And not only the founders and the employees believe it, but also the most visionary people believe it — the top VCs in Silicon Valley. The second reason is there are tons of unique problems to solve. And the third is that data science is really a foundation of the company.
Can someone talk about a specific problem that you worked on recently?
Deepak: I can talk about that. We have this really complicated problem, when an order comes in, to match it with one of, let’s say, 15 local stores, then we assign that to one of 200 shoppers, and determine the window in which that order needs to be delivered. That leads to millions of combinations.
But whereas other logistics companies, like FedEx for example, have one night to solve that problem and create a route for the next day, we have this problem every minute. To utilize our shoppers well and maximize our customer satisfaction, with the kind of latency you’d want, has been really challenging. I haven’t worked on such a challenging problem since school. And this plays a huge part in the larger strategy of the company.
Who are your stakeholders? Who are you listening to?
Deepak: Any tool we build touches on most business functions, so I think it’s very important to talk and listen to everyone to make sure we build the right product. For example, Eric works in forecasting, which helps operations schedule the shoppers, which supports customer satisfaction.
Eric: Many of our challenges come together in how we manage the shopper hours we have at our disposal. A lot of our orders are same day or even within an hour or two, so it’s important to have the right amount of people on shift at a given time. But it’s a delicate balance, because if you have too many people on a shift then the shoppers aren’t making enough money or we’re not making enough money.
“It’s very important to talk and listen to everyone to make sure we build the right product.” — Data Scientist Deepak Tirumalassetty
That’s why being able to forecast demand is so important. But we don’t have weeks and weeks of orders we can look at and just do a linear trend. We’ve got to look at past patterns and make estimates of how much we’re going to grow. That’s dependent partly on actual growth and partly on what we could have done if all options had been available at all times.
Then there’s the question, how do I feed that into a tool that’s going to provide additional benefits to the users? There’s some downstream optimization and some interaction between user input and output from the tool. So I have to make sure whatever I’m providing sets that tool up to be successful. It’s an interesting cross functional problem that affects the core of the business.
Can you tell me about the culture at Instacart? How is it different from other companies?
Brandon: There are a couple of things about Instacart that I really like. One of them is, we’re pragmatic. We tend to focus on just getting stuff done and moving the needle. We also tend to be very metrics driven across the business, not just in data science. For instance, if we’re building something on the engineering team, we’ll start with the metrics, what is it we’re trying to fix. You’re not done until you’ve shown that you have moved the needle on that goal.
We also emphasize enjoying the people we work with. I think it’s rare that a company gets this big and there’s no one I can point to and say, “Oh I don’t really enjoying talking to them.” I like everyone.
We also have an extreme focus on the customer. Like Andrew said, we structured our teams around business problems; sometimes I like to say we’ve structured them around customer problems. Every team is focused on creating the best possible experience for the customer. I think that has played out really well because we don’t have to do a lot of marketing. We keep growing because people have a good experience and tell their friends.
“We work with a really smart group of people but no one carries themselves like, ‘I’m smarter than you.’” –Data Scientist Eric Rynerson
Deepak: I’d say everyone across the company is so approachable. If you have a question, everybody is eager to answer. Even the cofounders, they’re just sitting right next to you, and they are always open for questions.
Eric: We work with a really smart group of people but no one carries themselves like, “I’m smarter than you.” That’s very refreshing.
What kind of hours are you working?
Andrew: The hours here are very flexible. Some people get here at eight, some people get here at eleven — like our team. [Laughter.] You’ll also see a good number of people logged on at night, but it’s not necessarily to meet deadlines. People just really like working on these problems. They might be working on something that isn’t even related to their day job, like digging into a new data set. I think that’s really exciting to see.
“People don’t have to make a show of working or pretend to be busy here.” — Data Scientist Eric Rynerson
Eric: We have free lunch and dinner, and quite a few people stick around for dinner, but by an hour or two after dinner, there’s nobody here. They might be working from home, but I think it’s healthy that people don’t have to make a show of working or pretend to be busy. Nobody really cares if you worked 35 or 45 hours; they are looking at what you accomplished.
What questions would you ask if you were considering joining this team?
Andrew: There’s a lot of candidates who have asked me, “How is Instacart going to make money?”
Brandon: Really? No one ever asks me that. There are a lot of ways that we make money. One is we charge a delivery fee, which goes to us. Some retailers pay us as well; we have a very successful partnership with Whole Foods and several other partnerships are in pilots.
In addition to that, we think there’s a big opportunity to work with brands directly, whether through couponing or other forms of promotions on the site. There are companies with multi-billion-dollar market caps that just aggregate transactional information from retailers and sell it to brands. We have all that same data. That’s a huge untapped revenue source for us.
Tim: People often ask me where we are headed, too. I always say, “We are in hyper growth mode, but we want to make sure, as we expand, it’s a strategic move as well.”
Do you have the money in the bank to support hyper growth?
Brandon: We recently raised $220 million back in December, and we hadn’t used all the money we’d raised from the previous rounds. So the money is there and when we want to turn on the jet fuel, we will. That will fund geographical expansion and categories expansion. We’ll definitely be around for a while.
Join the team!
Interested in joining the Instacart data science team? Visit their career page or contact Senior Technical Recruiter Tim Wong, tim@instacart.com.
