Navigating Data Science in Startups: Insights from Industry Experts
Written on
Chapter 1: The Unique Journey of Data Scientists in Startups
The world of data science in startups is renowned for its thrilling and unpredictable nature. One moment, you might be analyzing data on spreadsheets, and the next, you could be conducting customer interviews or fine-tuning CI/CD pipelines. The diverse responsibilities that come with being a data scientist in a startup can lead to continuous learning and growth.
To delve into this topic, I consulted my esteemed colleague, Minkyung Kang (MK), who established the data science team at Aquicore. As the second data scientist in our organization, I was curious about our shared and differing experiences. Together, we identified three key insights from our journeys, alongside three essential questions for anyone considering a role as an early data scientist in a startup.
- Embrace a Culture of Continuous Learning
Startups often operate with limited resources, a stark contrast to established companies. As a result, the breadth of data science roles—from analytics to engineering—falls to you, the sole "data expert." This can be either a source of frustration or empowerment, depending on your outlook.
With no dedicated product manager to guide your roadmap, you can enhance your business acumen by collaborating directly with stakeholders to shape user experiences. When engineers are occupied, you might find yourself responsible for setting up databases, containerizing models, and deploying them.
Engaging in every phase of a project—from brainstorming ideas to launching your model—provides a comprehensive understanding of the business and the potential of data science. This all-encompassing experience can transform you into a full-stack data scientist, a highly sought-after skill set.
However, this learning journey is not without its challenges. If you lack extensive experience, you'll frequently encounter questions without clear answers. For instance, should you choose a SQL or NoSQL database for a new data source? What happens if your dashboard isn't utilized by internal teams? Are customers more concerned about false positives or false negatives when alerted to anomalies?
While you may find mentors for business or engineering queries, core data science challenges—like selecting appropriate model features or addressing biased training data—might leave you isolated, especially if you're the first data scientist in the organization. Even with a second data scientist on board, it can be difficult to establish scalable best practices without guidance from someone with prior experience.
This often leads to a cycle of trial and error. While it can be humbling to confront your knowledge gaps, if you're open to learning from missteps, you will rapidly develop as an analyst, engineer, and critical thinker.
- Cultivate Business Acumen
A significant advantage of being an early data scientist in a startup is the opportunity to work closely with senior leadership. Your tasks are driven by pressing business needs, ensuring that your work holds real significance. Regular interactions with the CEO, product heads, or sales VPs help you grasp the broader context in which your company operates.
This engagement fosters a profound understanding of how to conduct impactful data science. You won't have to second-guess the usefulness of your models or dashboards—your stakeholders are likely just a desk away! If your outputs fall short, you'll quickly grasp why.
As you iterate on feedback, you naturally adopt a solutions-oriented mindset, focusing on "What problem needs solving?" instead of "What can I achieve with this tool?" This approach is invaluable for developing products that truly benefit the company.
However, the time spent honing your business insights may come at the cost of deep expertise in specific data science domains. If your startup isn't centered on image recognition, for instance, your expertise in that area may remain limited. Similarly, if the product isn't related to text summarization, you might not significantly enhance your natural language processing skills. This trend is especially pronounced in startups, where business needs can shift rapidly.
Instead of becoming a specialist in time series analysis, you'll likely excel in communicating technical concepts to non-technical stakeholders. You may also find yourself mastering fundamental statistics as you repeatedly clarify how your model reaches its conclusions.
- Drive Your Vision
Startups are characterized by a strong bias toward action. Typically, they aren't profitable until well-established, leading to a constant drain on investor funds. As expenses mount—salaries, office space, materials—there's immense pressure to achieve profitability or secure further investment.
Given the potential of data science to drive success—from automating reports to creating innovative product features—startups actively seek your insights to ensure their survival and growth.
As the sole data scientist, you'll be regarded as the authority in the field. This dual role requires you to generate ideas rooted in business context while also executing the technical aspects. While this can be daunting, it also empowers you to shape the vision for data science within the organization.
Do you believe that scraping Twitter for customer sentiment could enhance marketing strategies? Go ahead and implement it! Is an internal OCR app to automate receipt transcription worth pursuing? Build it and see! Are you confident that a customer churn predictor could assist the Customer Success team? Create it and evaluate its impact!
In contrast to larger organizations with rigid protocols, your influence in a startup is primarily limited by how quickly you can conceive and implement innovative ideas. It's akin to the Wild West—if you possess creative ideas and the initiative to see them through, you'll thrive.
Chapter 2: Critical Questions for Aspiring Data Scientists
As you consider a role as an early data scientist in a startup, reflect on these essential questions to ensure a fulfilling and productive experience.
- What is the Company's Vision for the Role?
The foremost question to ask is about the company's objectives for data science. Are they aiming to empower business leadership with data-driven insights, or are they in search of a software engineer with machine learning expertise? The answer will shape your experience significantly.
The nature of your role will vary based on whether you are in an analytics-heavy or engineering-focused position, influencing the data sources you'll access (e.g., Excel files vs. databases), the deliverables you'll provide (e.g., reports vs. software), and the colleagues you'll collaborate with (e.g., business executives vs. engineers).
Moreover, it's crucial to ascertain whether the company has a clear understanding of how data science can benefit them, or if they expect you to carve out the role. If they are merely hiring data scientists because it's the trend, be cautious—joining such an organization may lead to frustration, regardless of your expertise and assertiveness.
- Is the Data Ready?
Companies vary widely in their data readiness for a data scientist's use. It's not uncommon for firms to recruit data scientists even when their data isn't adequately prepared. The levels of data readiness can be categorized as follows:
- Ideas but No Data
In this scenario, a nascent startup requires a data scientist to initiate their data journey—identifying what data to collect and how to utilize it. Your role may involve running proofs of concept as you explore which ideas hold merit. This can be rewarding if the company's vision resonates with you, but be prepared to sharpen your data engineering skills.
- Some Data but Not Structured
In this situation, the company possesses data but could benefit from a data engineer. Data may be scattered across Excel files, Google Drive, or Salesforce, which can hinder your analysis and reporting efforts. If your deliverables require robust software, unreliable data access could significantly impede your work. You might need to either handle data engineering tasks yourself or advocate for hiring data engineers.
- Easily Accessible Data
The ideal scenario is when a company has readily consumable data via a database, data warehouse, or APIs. If they have been collecting data for years and have clear ideas for how a data scientist can extract value, consider this a major advantage when evaluating potential employers.
Keep in mind that a company may fit into multiple categories based on various data types—core product data might be well-structured, while customer metadata could be disorganized.
- Is There a Data Science Champion?
Successful data science efforts require a supportive ecosystem: well-defined business questions, accessible data, and software engineering support to integrate your work. Coordinating these resources can be challenging without assistance, particularly as the first or second data scientist.
A "data champion" in leadership is essential for advocating for your work and ensuring its visibility. Without this support, your contributions may go unnoticed, and you could find yourself struggling to demonstrate the value you provide.
Having a champion doesn't negate the need for regular communication and education about your work. Yet, even the most compelling arguments may falter without an influential ally who can facilitate necessary changes—such as hiring data engineers or defending the time required for research. Lacking this support can set you up for failure, making it exceedingly difficult to carry out your responsibilities.
Conclusions
Joining a startup is a significant decision! The rapid pace may not suit everyone, but for those eager to make immediate impacts—especially early in their careers—startups offer unparalleled opportunities for growth. They foster continuous learning, business acumen, and the implementation of innovative ideas. You'll develop a critical understanding of the business landscape and the kinds of problems that data science can effectively address.
However, it's easy to find yourself in a company unprepared for a data scientist. Ensure the organization has a clear vision for your role that aligns with your aspirations, or you might end up in an entirely different position than you envisioned. Be ready to adapt to whatever data formats you encounter, even if that requires developing your data engineering skills. Lastly, ensure you have at least one data champion to help navigate the complexities of organizational change.
Good luck!
Best,
Matt and MK
The first video titled "How I Became a Data Scientist at Google (with a Low GPA)" offers personal insights and strategies for breaking into data science, emphasizing resilience and adaptability.
The second video, "Data Science at a Tech Startup," provides an overview of the unique challenges and rewards of working in a startup environment as a data scientist.