The Journey of the First Data Scientist at Notion: An Interview with Erin
This is part of the "First Data Scientist at a Startup" series, where I interview the first data scientists at startups to learn about their experiences and challenges. The goal is to help founders understand when they might need to hire their first data scientist and how to set them up for success in this challenging role. Additionally, this series aims to help data scientists in making informed decisions when considering a first data scientist opportunity.
Intros
Erin was the first data scientist at Notion and she joined Notion from Asana where she was the 2nd data scientist. During her 2+ years at Notion, Erin expanded the data science team to over 10 members and was instrumental in optimizing Notion's data stack and building other crucial data functions.
Key Takeaways
Being the first data scientist at a fast-growing startup can be very fulfilling and provide a lot of learning opportunities. However, it can also be challenging, especially if your growth and the company’s are not in sync.
As the first data scientist, it is important to take on a proactive leadership role, such as advocating for the team and building processes that are not yet implemented. This investment can pay off in the long term.
Building clear wins for a completely new function is important to set the right foundation. Therefore, it is crucial to choose early projects that can demonstrate the value added by a data scientist.
The concept of starting completely from scratch may seem appealing, but it rarely happens that way. While the initial data scientists are not usually responsible for building data infrastructure from the ground up, involving a data scientist early on to establish a strong foundation of data models is crucial to prevent future technical debt.
Details
When you first joined Notion, how big was the company when you were hired?
I think it was around 50 people when I interviewed and then like somewhere between 60 and 70 when I actually joined. They were growing really, really fast at that point and I think it was almost three months between when I first applied and when I started.
So how did you find the job opportunity?
I think Notion had a job post online. I knew a Notion designer who I worked with briefly at Asana, so I asked him to help refer me. He connected me directly with Notion’s CEO and Head of Growth.
Can you tell me a little bit about why you decided to join Notion? Did you do any research about Notion?
I'd been at Asana for five and a half years, and after all that time, I left Asana thinking that I wanted to do something that was quite different from what I'd already done.
Maybe working in a different industry or a company of a different size, something that wasn't just a repetition of what I had done at Asana. But I took six months off after that job before I started looking for something new, and then COVID hit.
As a mega planner, I saw all of these layoffs happening and started panicking, thinking about how I would ever find a job. So when I did start my search, I included some companies that were very similar to Asana, thinking that they were growing in the pandemic because they were tied to remote work. Also, that I had experience in this area, so I probably had a better chance.
When I applied to Notion, I had heard of it because it would come up in competitive discussions within Asana, but I hadn't really used the product and didn't know much about it. It sounds flippant given what I know now about Notion’s cult following, but in my mind, it was like a safety school, my backup plan if something else didn't work out.
But then, as I started talking to the team, it became clear that even though they were in a similar space to Asana and it would mean joining a company of a similar size, it would still be a different enough experience because of the company culture. Even though the products have some overlap in terms of how people use them, their missions are very different. It felt like something I could be really excited about.
I also noticed that when I talked to friends about where I was interviewing, many of them were like, "Oh my God, Notion! I love Notion. Notion is such an incredible product." It was clear that there was something drawing people to it. So yeah, I was excited about being able to work somewhere where people were genuinely excited about the product.
Were you hesitant about being the first data scientist? I know you were the 2nd data scientist at Asana. Did you envision things were going to be different being the first vs the 2nd data scientist?
I wasn't too hesitant about being the first data scientist. It was something I was excited about in my career. I believed I needed to build a stronger foundation before taking on the role, so I planned to gain more experience first. However, I knew it was something I wanted. During the interview process, I mentioned to Ivan that I eventually wanted to be a first data scientist, but perhaps not at that moment. He responded by saying, "If that's what you want, why not just do it?"
I was more worried about the phase of growth that Notion was in. Towards the end of my time at Asana, we had pretty explosive employee growth, and growing a team that quickly always comes with some challenges while everything gets back into a new equilibrium. I knew Notion had also just entered a phase of rapid growth, both of the business and the team, and I was worried about having the capacity for that.
Do you find it hard to adjust from being a manager to an IC? Do you think the first data scientist should have management experience?
No. I actually enjoy individual contributor (IC) work and I think returning to IC work occasionally provides a good reminder of what's happening and keeps you in touch with how technology is changing rapidly. I had actually planned to stay in an IC role longer, but when a manager role became available, I didn't want someone else to lead the team. I wanted to be the one leading the team.
Looking back, I realize that I could have examined my ego more and done what I actually wanted in that moment, instead of doing what I thought would make me look good on my resume or give me the career I thought I should have.
But yes, I believe there's plenty of room for people to transition between IC and manager roles. If you're hesitant because you have management experience and don't want to lose that trajectory, there are still many opportunities to go back to an IC role and then become a manager again in the future. It may take a little longer, but it's possible. I think there's a shortage of good managers in the tech industry, so if you're skilled at management, there will be more opportunities in the future to utilize those skills.
Regarding becoming the first data scientist, I don’t think it's necessary to have management experience. It's good to have experience in driving the direction of projects and feeling comfortable setting strategy. However, you can gain that experience in an IC role if you are in a tech leadership position, without necessarily having people management responsibilities.
What do you think your CEO saw in you that led to your offer?
I don’t know for certain. I think early on, the draw was that I had experience being at a startup before, and even one that was in the same space (collaborative SaaS tools). Once he talked to my references from Asana, I think he also understood that I had experience taking charge of a lot of complexity and scope, in situations where there would never be enough time to do everything. Prioritization and being able to build something that’s good enough rather than perfect are huge at startups, because there are always so many things you could be spending time on, so I think he felt comforted knowing that I’d had to do that in my previous role. The last thing I heard from several people is that some of the other final candidates were very attached to growing out teams right away. I had experience leading a team and was open to doing that but also happy to do individual contributor work for as long as that made sense for the company. Every company will be different with the specifics of needing a manager or IC, but I think the more generalizable lesson is that most startups need some degree of flexibility because the boundaries around roles can shift so quickly.
What kind of data stack did they have? What was the status of data?
When I joined, the former Head of Growth had already built a solid foundational data stack, so I actually came into a relatively well-developed stack. He was comfortable enough with data to have set up a significant portion of our stack. Because he wasn’t a data scientist, he’d relied heavily on the “modern data stack” that’s a bit more self-serveable. We had logging happening in Amplitude, pipelines running in dbt, with data coming in from Fivetran. We had set up BigQuery and then Census to push data back out to some of our other tools, and Mode for reporting. So, many of the major components were already in place.
So I think it's pretty common that you come in and there's already some existing data infrastructure you're expected to work within or around, and you're not necessarily setting everything up from scratch unless you are like the founder of the company. Sometimes if the tool they chose before you is not something you really like, you might have to work within that constraint, at least to start. To be honest, I was a little bit surprised by the strong foundation of data stack Notion already had and the fact that I didn’t have to work on a lot of data infrastructure work.
How did you decide on building your pipelines? How do you think about your data model when you first joined?
In the beginning, we had some data sets created by the former Head of Growth. These data sets were specifically tailored for the needs of a growth marketer, so they weren’t always as versatile and applicable to other use cases as we might have wanted. I started identifying the elements that we would need to make the data sets more adaptable. One challenge was that it was difficult to argue for rebuilding our data sets, because even if they were sometimes difficult to use, they did exist.
On the other hand, when you have no existing data sets, it becomes easier to recognize the need for comprehensive ones. To decide what we should build next, I talked to a lot of folks across the different teams about their pain points, especially focusing on what questions they felt least equipped to answer with the current data. I ended up building out MRR (monthly recurring revenue) data as my first big project, which I would never recommend anyone do. 😂 It’s such messy data. In hindsight, I would definitely start with something that was a more fast, decisive win first.
What was your initial impression of the data at Notion when you first joined? On a scale of zero to 10, how would you rate it?
It was around 2 or 3. There was a lot of data available, but it was difficult to use.
Anything surprised you at Notion in the beginning?
I joined from another collaborative SaaS company so I thought things were going to be similar. However, it was more fluid and unstructured compared to Asana. This was surprising to me. It was actually good for me because I usually prefer a lot of structure and having a plan. But being in a place that often didn't have that showed me that sometimes taking a less structured path can lead to faster and more effective results, or even a more interesting solution. It was liberating to realize that there isn't just one correct way to approach a problem. Personally, I still prefer more structure, but it was nice to see that structure is not always necessary. It's only valuable if the resulting clarity helps you achieve impact faster.
How much guidance did you receive from your manager to improve the quality of your work when you first joined?
My first manager was the Head of Engineering. He was really good at keeping me informed about what was happening in the company, strategically focusing on what mattered for Notion at that time. I always felt well-informed during staff meetings and similar events. However, I didn't receive as much feedback on the data-related aspects. It can be challenging in the data field because you may not have peers as quickly as in larger teams.
So you often get feedback from people who don't fully understand your work. But in terms of what I needed the most, like the overall company direction and strategy, I felt like I was getting good feedback.
When you joined Notion, have you ever done anything that was outside of your job description as a data scientist? Something unrelated to data, for example?
I don't feel like there was a ton of work that I did fully outside of data because there was just so much to do with data, although a lot of it was data engineering rather than data science. At one point, we were looking for someone to take on a PM-type role for a new growth program and the team asked if I’d be interested in helping with that that, but it became clear before the team even kicked off that I’m not a great PM and that there was just too much need for me to be focusing on data anyway.
It is common in startups for individuals to take on additional roles or responsibilities outside of their core role. In this case, I believe that because they waited a bit too long to hire data personnel, there was a lot of catch-up to do, so it didn’t make sense to put my focus elsewhere.
When do you think Notion should have hired the first data scientist or first data person?
A year earlier would have been good. So it was like 30 people? They had the Head of Growth setting up the data systems, so that would have been a good signal that it was time to hire someone who had the background to set those foundations.
Can you share some mistakes that might have made?
Initially, the main issue was that the team was new, so some of the leaders didn't fully understand how to work with us. For example, there was some uncertainty about what to focus on when working with data, despite recognizing its importance. I should have taken more control over the data's direction, strategy, and goals. It would have been beneficial to maintain close contact with the leadership team. In my previous experience, I waited for permission in a top-down environment, but that approach doesn't work at Notion or many other places. I missed an opportunity to establish clear expectations for myself and the team from the start. We should have explained the importance of our work, how to collaborate with our team, and the strategic goals we were working towards. I now realize that I didn't step into that leadership role.
I also regret my choices in early projects. The main project I worked on upfront was building revenue data, along with various smaller projects in different areas of the business. My intention with the smaller projects was to demonstrate the value of data in a grassroots manner and create data enthusiasts in multiple teams. However, my one big project working on revenue was challenging due to messy data that could only be cleaned up through complex process changes for the sales and finance teams. Ultimately, only the revenue data project caught the attention of leadership, but it faced numerous setbacks. The other projects were too small to demonstrate the true impact of data for data to drive the key decision-making early on. In retrospect, I would have chosen one area to focus on initially, where I could make an immediate positive impact and gain support from at least one leader.
What do you think is the biggest lesson you learned from being the first data scientist?
It's important to give yourself permission to drive things forward based on your own opinion and expertise. Instead of assuming that the existing structure is the culture or waiting for someone else to do it, you need to take the initiative to put that structure in place.
If you’re the first person in a data science role, you’ll often end up with an individual contributor title but a significant load of leadership responsibilities early on. You’ll also be the only one moving forward a lot of individual technical work. People talk about prioritization all the time with data science. Still, it’s even more important when you’re having to so frequently jump between high-level strategic thinking and being in the weeds doing.
I wish I had been more comfortable stepping into my power and leadership even without a title, and pushing things forward. I saw people who joined shortly after me, in a similar role, taking ownership as founding members of a new function in a way that brought clarity to their respective areas. I wish I had done the same. I had the skills, but I was too afraid. Maybe it's because I'm a woman, but I sometimes struggle with feeling empowered to speak up.
What were perhaps the best and worse moment of you being the first data scientist at Notion?
Honestly, the best moment was returning from maternity leave and seeing the work that the team was doing, I felt very proud of that early team. It was amazing to witness the progress made in just a few months. You all were doing awesome things, and that was really cool to see.
But the most challenging period for me was the longer transition after I came back from maternity leave. I was used to being great at nearly anything if I really tried and I didn’t even feel good at most things, at work or at home. Keeping up became a struggle, particularly in a fast-growing startup where my capacity to learn and adapt couldn't match the company's rapid growth.
You mentioned that your proudest moment was when you returned from maternity leave and saw the great achievements made by the early data science team. So, what qualities are we looking for when hiring the first data scientist to join the team after you?
I would say someone with skills that complement yours, as data science is a broad field. For example, one of the first data scientists I hired was focused on marketing, while I'm focused on product. Also, I think our different personalities allowed us to advocate for the team in different ways and create visibility through effective communication. I felt like there was a lot of synergy there.
In terms of seniority, I found that at Asana, we started out hiring junior individuals and training them. There were things that were great about that system, but it did initially limit their involvement in strategic decision-making outside of the data science team and their visibility beyond our team. It meant in some cases that I had to spend extra time guiding them instead of letting them make decisions on their own. So, having someone who can quickly add value at a startup is crucial.
How did you convince the leadership to invest more in data? Were there any strategies or moments of realization that made them understand the importance of data?
I actually was lucky enough that I didn’t have to! There were several people who joined the company around the same time as I did who came from much larger companies with robust data teams, and they were super hungry for more data to inform their decisions. If you’re not in that position, showing value with early projects and then making it clear what incremental projects you could take on with more people can be really effective.
Do you have any kind of piece of advice or wisdom to share to anybody who are thinking about becoming the first design at a startup?
I believe that this experience is valuable and provides an opportunity for personal growth and learning. It can be challenging, so it's important to approach it with flexibility, humility, and a willingness to adapt. Whether that means expanding your role beyond data or narrowing it down from what you initially expected.
You can learn a lot and experience rapid growth compared to other environments, but you should also be prepared to face challenges and establish a healthy work-life balance to ensure your well-being.
