There's a world of opportunity for data scientists beyond the tech giants everyone's heard of. Smaller companies and startups can offer practical experiences that serve as stepping stones in your career. Even seasoned data scientists might find the appeal in taking on bigger responsibilities in smaller organizations. But working at a startup as a data scientist comes with its own set of challenges and considerations. I've worked at both big and small companies, and here are some frameworks to help you navigate the startup world.
Startups are not the Same
When we think "small" in biology, bacteria often come to mind. Yet even these microscopic organisms are incredibly diverse. Take E. coli, the most well-studied bacterium: different strains can vary by up to 25% of their DNA! This level of variation is akin to comparing humans with zebrafish, which we'd consider as completely different species.
Startups are also like completely different species. When I chat about big company experiences with other data scientists, we often find a lot of commonalities. But every time I talk about working at startups with a fellow startup adventurer, I find that our experiences can be worlds apart. Such diversity in startup environments makes it quite hard to build expectations about what it's like working at a startup.
There are a few main factors that can shape a startup's experience. Let's look at two of them.
Stage
A 10-person startup is vastly different from a 100-person startup (or a 1,000-person "startup"). In the former, you might be the lone data scientist wearing multiple hats, from analytics to machine learning to data engineering. In the latter, you could be part of a specialized team with more defined roles and processes. The size impacts everything from resource availability to the breadth of your responsibilities and the potential for immediate impact.
And number of people is not the only way to look at startup stage. Another common divider is "before product market fit (PMF)" versus "after product market fit". This is an important distinction for data scientists, because a common responsibility of data scientists is to analyze product data and propose growth strategies. But in the early dawn of a startup, before a product is ready for market, it's often too premature to make any concrete recommendations. You can't do data science without data, and you can't have data without a product.
That doesn't mean super early stage startups are never suitable for data scientists. If you identify as a "generalist data scientist" who enjoys wearing many hats and building from the ground up, then early stage startups can be an energizing environment.
When a startup has blessfully hit PMF, it enters the "growth" phase. This can be a very interesting time for data scientists as the company needs to simultaneously maintain product quality and expand user base. Monitoring and observability become a priority, which is a challenge well-suited for data scientists. In addition, oftentime product roadmap begins to consider more algorithm-driven user experiences to create market differentiation. This can also be an opportunity for data scientists to develop model-based solutions.
Here is a breakdown:
-
Very early stage startups (on the order of low 10s of people / seed stage or series A):
- Company Priority: Build Product
- Pros for Data Scientists:
- Big responsibility; wear many hats
- Hands-on experience building from the ground up
- Cons for Data Scientists:
- Can be the first data scientist in the company (reduced opportunity to learn from DS peers)
- Requires tremendous amount of self sufficiency
- Less chance to do advanced analytics (unless the company's product is about analytics)
-
Mid stage startups (on the order of low 100s of people / Series B+):
- Company Priority: Growth
- Pros for Data Scientists:
- Product analytics becomes crucial
- Opportunities to work on data-driven and model-driven product features
- Potential to be a part of a nascent and growing data science team
- Existence of some tools and infrastructure
- Cons for Data Scientists:
- Increased pressure to deliver measurable results
- Balancing between quick wins and long-term data strategy
- Fast growth can mean fast change; potential for role ambiguity as the organization scales
-
Late stage startups (1000s of people / pre-IPO):
- Company Priority: Scale and Profitability
- Pros for Data Scientists:
- Well-established data infrastructure and processes
- Opportunities for specialized roles (e.g., ML Engineer vs Data Analyst)
- Exposure to large-scale data challenges
- More resources for advanced projects and research
- Company name recognition in resume
- Cons for Data Scientists:
- Less flexibility and more bureaucracy
- Harder to have company-wide impact as an individual
- Potential for siloed teams and reduced cross-functional exposure
- Increased competition for promotions and career advancement
Industry
I come from an academic background (in case opening with E. coli genetics didn't already give that away). So I didn't appreciate how different (software) companies can be depending on the industry. Specifically, I'm referring to the divider of B2C (business-to-consumer) versus B2B (business-to-business) companies.
In large organizations, individuals might not feel this distinction (is this Central Limit Theorem?). At a big enough scale, there are all the typical data science works you'd expect: analytics, predictive modeling, data engineering, and more. But in startups, the types of data science works that are prioritized depends heavily on the business. As a result, in B2C vs B2B startups, data scientists can have extremely different day-to-day.
A lot of these differences are rooted in user volume. In consumer business, you don't get "interesting" until you have a sizable amount of users (say, 1M). The lifetime value (LTV) of a user of a consumer software is typically quite low, so you need a lot of them. This makes large-scale analytics on user behaviors and personalization an existential challenge for the young company.
In enterprise business, the situation is quite different. It often takes longer to sell to an enterprise customer, but the average contract value is much higher. So instead of dealing with "scale" early on in a consumer business, typically you deal with "complexity" in an enterprise business as winning each deal might require you to customize the product to the customer's needs.
Here is a breakdown:
-
B2C (Consumer) Startups
- Data Volume: High & real time
- Data Type: Typically internal product data such as from logs
- Key Challenge: Large-scale analytics and personalization
- Focus: User behavior, engagement metrics, retention strategies
- Typical Projects:
- A/B testing for product features
- Churn prediction models
- Recommendation systems
- Skills Needed: Scalable data processing, consumer psychology
- Impact Measurement: Often more direct (e.g., increased user engagement)
- Stakeholder Interaction: Primarily internal (product managers, marketers)
-
B2B (Enterprise) Startups
- Data Volume: Lower
- Data Type: Typically customer data through integrations
- Key Challenge: Dealing with complexity and customization
- Focus: Customer-specific solutions, long-term value metrics
- Typical Projects:
- Data engineering to unify different data types
- Customized analytics and models for clients
- Skills Needed: Domain expertise, stakeholder management
- Impact Measurement: May be long-term (e.g., improved client retention)
- Stakeholder Interaction: Both internal and external (including clients)
It should be noted that the separation between enterprise software and consumer software has started to blur in the last few years. More and more "work softwares" gain traction by directly selling to the individual workers, much like the acquisition mode for a consumer software (for example, Notion, Figma, and to some degree even GitHub). Nonetheless, knowing the difference between B2C and B2B is still a helpful framework when you consider startups.
De-risking with Research and Asking Questions
We can't think about working at a startup without considering the risk. You've probably heard of how 90% of startups fail overall. And even if you are lucky enough to work at the 10% that succeed, more often than not, the level of success is not going to be the life-changing outcome we read about in the news.
You should only consider working at a startup if it aligns with your career goals at this point. And while there are no sure-fire ways to choose a startup that will eventually go big, doing research and asking questions is a must to ensure you will have a happy and productive time at the company. You should almost evaluate the company as if you are an investor (because you are! With your time, labor, and eventually the equity you will receive).
Here are key areas to focus on in your research.
Understand the Company
- Founders' Background: Who started the company? What's their experience?
- Look for founders with industry experience, professional networks, and prior startup experience
- Studies have shown that repeat founders are more likely to succeed
- Funding: How much funding has the company raised? Who are the investors?
- Well-funded startups with reputable investors may offer more stability
- Understanding the funding stage can give insights into the company's growth phase
- Company Culture: What values does the company prioritize?
- Look for alignment with your own work style and values
- It's okay to check platforms like Glassdoor for employee and interview reviews, but take them with a grain of salt because, as a data scientist, you know about self-selection bias
Analyze the Market
- Market Size: Is the target market growing? How big is the opportunity?
- A growing market offers more potential for the startup's success; rising tide lifts all boats
- Consider both current market size and future projections
- Competition: How crowded is the space? Who are the major players?
- Understand how the startup differentiates itself from competitors
- Look for markets with room for innovation or disruption
- Industry Trends: What are the current and emerging trends in the industry?
- Startups aligned with positive industry trends may have better prospects
- Being able to articulate industry trends during interviews will boost your profile
Evaluate the Product
- Problem-Solution Fit: Does the product solve a real, significant problem?
- Look for products addressing clear pain points in the market
- Consider the urgency and frequency of the problem being solved
- Differentiation: What makes this product unique?
- Understand the startup's unique selling proposition
- Look for innovative features or approaches that set it apart from competitors
- Traction: Does the product have any early adopters or success stories?
- Early traction can be a good indicator of product-market fit
- Look for case studies, testimonials, or public user metrics
Ask Questions During Interviews
When conducting interviews, I always leave 10-15 minutes for the candidate to ask any questions. You should always prepare a list of thoughtful questions to ask. How the interviewers respond can provide valuable insights not only into your specific queries but also into the company's culture, including their approach to transparency, integrity, and trust-based communication.
Here are some questions you might consider:
- "What do you see as the most exciting opportunities for the company in the near future?"
- "What has been the most surprising learning for the company in the past year?"
- "What are some recent wins and setbacks the company recently had? How does the team typically deal with them?"
- "What process do you use to build the product roadmap?"
- "How does the company approach financial planning and resource allocation, especially in terms of runway and future funding strategies?"
Watch for Red Flags
The Anna Karenina principle, inspired by Tolstoy's novel, suggests that success requires a number of factors to all go right, while failure can occur from any single deficiency. In the startup world, this translates to the idea that while there's no guaranteed formula for success, there are common pitfalls that can lead to failure. So you should be diligent in your research to identify potential red flags.
-
High turnover rate: If the company has a history of employees leaving quickly, it could indicate problems with culture, management, or overall stability.
-
Unwillingness to discuss finances: While startups may be cautious about sharing detailed financial information, a complete refusal to discuss runway or funding plans could be concerning.
-
Unsustainable working style: While you shouldn't expect the same level of work-life balance at a startup compared to a big company, an overly demanding working style is not sustainable. Working every weekend and into the AM's everyday should not be a norm. Neglecting family or personal life should not be glorified.
-
Lack of aligned company vision: If different interviewers give conflicting information about the company's goals or direction, it might suggest a lack of alignment or clear leadership. Startups need to move fast, so if everyone is running in different directions, this is a bad sign for progress.
-
Overemphasis on perks: If the company talks more about their ping-pong tables and free snacks than on substantive aspects of work and innovation, it might indicate misplaced priorities. Think about it, if people are joining the company for the perks, then are they going to be there to grind out the product?
Seeing one of two of these red flags doesn't necessarily mean you should reject the opportunity outright. There could be very reasonable explanations (such as a major deadline crunch, or a difficult decision to pivot). You should gather as much data as possible, and make informed decisions that best align with your career goals.
Finally
Working at a startup can be a transformative career move, but it's important to know what you are getting into. I hope this post has been helpful for you!
