Your First 90 Days as a Data Scientist

DoorDash about five months ago. This is my first time starting at a new company as a Data Science Manager. DoorDash is moving fast, expectations are high, and the context is deep, making onboarding a challenge. However, it was one of the fastest growing periods of my career.

The first three months in any new job are the building phase – building connections, domain understanding, and data knowledge – and a smooth ride sets the foundation for later success. So, in this article, I will share what was most important in the first months and my checklist for any data science.

I. Build Links

Before anything else, let me start by building a connection. When I was in school, I pictured data scientists as people who spend all day looking down at the ground writing code and building models. However, as I got older, I realized that data scientists make a real impact by embedding themselves deeply in the business, using data to identify opportunities, and drive business decisions. This is especially true today with DS's robust headcount and automated AI for basic coding and workflow analysis.

Therefore, building connections and getting a seat at the table should be a priority during onboarding. This includes:

• Regular riding sessions with your manager and riding buddy. These are the people who best understand the scope of your future, expectations, and priorities. In my case, my boss was my riding buddy, and we met almost every day during the first two weeks. I always came with a prepared list of questions that I encountered during my riding.
• Set up conference calls with various partners.Here is the agenda I usually follow for those calls:
  • 1. Personal introductions
  • 2. Area of ​​focus and priorities
  • 3. How can my team best support them
  • 4. Any riding advice or “things to know”
  • I especially like the last question as you always give great details. Five years ago, when I boarded Brex, I asked the same question and summarized the answers in sections here. The best I have found at this time”Don't be afraid to ask dumb questions. Play the new hire card as much as you can in the first three months.“
• To our valued colleagues, set up a weekly/bi-weekly 1:1 and get yourself added to recurring project meetings. You may not include much at first, but just listening and gathering context and questions helps.
• If you ride like a boss like me, you should start talking to your direct reports early. During onboarding, I aim to learn three things from my direct reports: 1. Their jobs and challenges, 2. What they expect from me as a manager, 3. Their career goals. The first one helps me to get up in the area. The latter two are important for establishing trust and working relationships early on.

II. Create a Domain Status

Data scientists succeed when they understand the business well enough to influence decisions – not just analyze results. Therefore, one of the most important aspects of onboarding is building your domain knowledge. Common techniques include talking to people, reading documents, searching Slack, and asking more questions.

I usually start with interviews to see the important business context and projects. Then I dig up related documents in Google Drive or Confluence, and read Slack messages in project channels. I also put together questions after reading the docs, and ask them in 1:1s.

However, one challenge I have encountered is digging down the rabbit hole of documentation. Each document leads to additional documents with many unusual metrics, acronyms, and projects. This is especially challenging as a manager – if each team member has 3 projects, 5 people means 15 projects to be completed. At one point, my browser's “To Read” tab group had over 30 tabs open.

Fortunately, AI tools are here to the rescue. While reading all the documents one by one helps to gain a detailed understanding, AI tools are good for providing a holistic view and connecting the dots. For example,

• At DoorDash, Glean has access to internal documents and Slack. I often chat with Glean, asking questions like “How is GOV calculated?”, “Provide an overview of project X, including the goal, timeline, findings, and conclusion.” It links to the sources of the document, so I can still dive deeper quickly if needed.
• Another tool I tried is NotebookLM. I shared documents with it on a certain topic, asked it to create summaries and mind maps for me to collect my thoughts in a more organized way. It can also create podcasts, which are sometimes more digestible than reading documents.
• Other AI tools like ChatGPT can also link to internal documents and serve the same purpose.

III. Build Data Knowledge

Building data knowledge is as important as building domain knowledge for data scientists. As a top manager, I hold myself to a simple standard: I must be able to do data work that is efficient enough to provide effective, reliable guidance to my team.

Here's what helped me get up to speed:

1. Set up the technology stack in the first week: I recommend setting up a technology stack and developer environment early. Why? Access issues, permissions, and weird environmental issues always take longer than expected. The sooner you get everything set up, the sooner you can start playing with data.
2. Make the most of AI-assisted data tools: Every technology company is integrating AI into its data processing workflow. For example, at DoorDash, we have Cursor connected to Snowflake with internal data information and context to generate SQL queries and analysis based on our data. Although the generated queries are not yet 100% accurate, the tables, joins, and previous queries indicate that they serve as excellent starting points. It will not replace your technical judgment, but it greatly reduces the time to have an initial understanding.
3. Understand key metrics and their relationships: Data literacy means not only being able to access and query data, but understanding the business through the lens of data. I usually start with weekly business updates to get core metrics and their trend. This is also a good way to contextualize metrics and get an idea of ​​what “normal” looks like. I found this incredibly helpful when analyzing bowel movements and test results later.
4. Get your hands dirty : Nothing forces your understanding of data more than doing a specific task. A good boarding plan usually includes a small project to start. Even as a manager, I did some IC work during my tenure, including estimating the opportunity for the planning cycle, designing and analyzing multiple experiments, and identifying and predicting movement metrics. These projects accelerated my learning more than just reading.

IV. Start Small and Contribute Early

While boarding is all about learning, I highly recommend starting small and contributing early. Early contributions show ownership and build trust — often faster than waiting for a “perfect” project. Here are some practical ways:

• Update boarding documents: As you go through the onboarding document, you'll encounter random technical issues, notice broken links, or find outdated instructions. Not only a win for yourself, but developing a boarding document is a great way to show that you are a team player and want to improve boarding for future hires.
• Create documents:No company has perfect documentation – from my experience and talking to my friends, many data teams face the challenge of outdated or missing documentation. Since you're on board and not busy with projects at the moment, it's the perfect time to help fill those gaps. For example, I created a project directory so that my team could link past and ongoing projects with priorities and clear points of communication. I also developed a set of metrics heuristics, summarizing the causal relationships between the different metrics we learned from previous tests and analyses. Note that all these documents become an important element for AI agents, improving the quality and consistency of the results produced by AI.
• Suggest process improvements: Every data team works differently, with pros and cons. Joining a new team means you bring a fresh perspective on team processes and may see opportunities to improve efficiency. Thoughtful suggestions based on your past experiences are very important.

In my opinion, successful onboarding aims to innovatecross alignment, business agility, and data insight.

Here is my riding checklist:

1. Week 1–2: Basics
   – Meet great business partners
   – Make yourself add to core meetings for various activities
   – Understand the team's focus and priorities at a high level
   – Set technology stack, access, and permissions
   – Write your first line of code
   – Read the texts and ask questions
2. Week 2–6: Get your hands dirty
   – Dive deep into the OKR group and the most used data tables
   – Deep dive into your area of ​​focus (many documents and questions)
   – Complete the first project to the end
   – Make contributions in advance: Review outdated information, create one document, or suggest one process improvement, etc.
3. Week 6–12: Identity
   – Be able to speak in various meetings and give your informed opinion on data
   – Build trust as the “go to” person for your domain

Onboarding looks different across companies, roles, and senior levels. But the principles don't change. If you're starting a new role soon, I hope this checklist helps you step up with more clarity and confidence.