Lessons Learned from a Decade in Data Science

Over the past ten years, working in analytical roles at various companies—from a small fintech startup in Europe to high-growth pre-IPO scale-ups like Rippling and big tech firms such as Uber and Meta—has provided a wealth of insights. Each company had a unique data culture and view on data, and each role presented its own challenges and hard-learned lessons. Here are ten key ideas from this decade of experience, applicable to any company regardless of stage, product, or business model.

  1. Tell a Story with Data
    • Understanding your audience is crucial. In research-focused organizations or when presenting to technical stakeholders (e.g., engineering), a detailed, academic-style analysis might be appropriate. Conversely, non-technical business teams or executives require a focus on key insights and their implications for business decisions. The Pyramid Principle, developed by McKinsey consultant Barbara Minto, is a well-known approach for this type of insights-led, top-down communication. The Pyramid Principle helps busy executives absorb your message quickly because it uses vertical relationships between the key points: Top level: The summary point you need to communicate. Second level: The key points supporting the top-level point. Third level: Data that supports second-level points.
  2. Business Acumen Differentiates Good from Great Data Scientists
    • In companies with high standards, technical skills are a given among senior data scientists. What sets individuals apart is their ability to drive maximum impact for stakeholders by understanding business priorities, scoping relevant analytics solutions, and communicating insights effectively. Developing business acumen involves paying attention to strategic priorities, connecting them to team projects, and ensuring data insights are actionable.
  3. Be an Objective Truth Seeker
    • Unlike many roles that reward goal achievement, data scientists have the luxury of prioritizing truth over convenience. Stakeholders might pressure for data that supports their narratives, but long-term success comes from promoting data-driven truths, even if they are uncomfortable.
  4. Combine Data with Primary Research
    • While quantitative analysis is essential, it often misses emerging signals. Complementing data with primary research, such as reviewing closed-lost deal notes or talking to customers, can reveal underlying issues not captured by structured data alone.
  5. Skepticism with Too-Good-To-Be-True Data
    • A sudden positive change in metrics often signals data issues or one-off effects rather than sustainable improvements. Healthy skepticism and thorough validation are necessary to avoid misleading conclusions.
  6. Be Open to Changing Your Mind
    • Regularly changing opinions based on new data is natural in analytical work. Although it may cause friction, revising recommendations in light of new evidence demonstrates intellectual rigor and adaptability, which are crucial for sound decision-making.
  7. Pragmatism Over Perfection
    • The ideal scientific approach is often impractical in business contexts that demand timely answers. Data scientists must balance rigor with pragmatism to support business needs effectively.
  8. Avoid Burning Out Data Scientists with Ad-Hoc Requests
    • Hiring full-stack data scientists to handle ad-hoc tasks like dashboard building and data pulls leads to burnout. Solutions include implementing AI chatbots, training teams on basic SQL, and using self-serve BI tools to reduce reliance on data scientists for routine tasks.
  9. Not Everything Needs a Fancy Dashboard
    • While dashboards in BI tools are essential for critical decisions, simpler solutions like Google Sheets can suffice for many needs, enabling quicker and more flexible analysis without overburdening data science teams.
  10. Standardized Metrics Across the Company Are Unrealistic
  • Perfect metric standardization is often unachievable, especially in fast-growing startups. While discrepancies in metric definitions can be uncomfortable, they typically do not significantly impact overall narratives or recommendations. Prioritizing rigorous handling of critical reports is more important than enforcing perfect governance across all data.

Final Thoughts

Some of these points may initially seem challenging, such as pushing back against cherry-picked narratives or adopting a more pragmatic approach over perfection. However, embracing these practices will ultimately help establish oneself as a true thought partner and a valuable asset to any organization.

Related Posts
Guide to Creating a Working Sales Plan
Public Sector Solutions

Creating a sales plan is a pivotal step in reaching your revenue objectives. To ensure its longevity and adaptability to Read more

Web Pages That Helped With My Google Data Engineer Exam
Salesforce Certifications

Google Data Engineer Exam It seems like every day more resources appear to help you study for the Google Data Read more

What is Advanced Reporting in Salesforce?

Cross Filters, Summary Formulas, and More: Advanced Reporting in Salesforce Salesforce comes with report types out-of-the-box for all standard objects Read more

How Travel Companies Are Using Big Data and Analytics
Salesforce hospitality and analytics

In today’s hyper-competitive business world, travel and hospitality consumers have more choices than ever before. With hundreds of hotel chains Read more