Building a world-class data org | Jessica Lachs (VP of Analytics and Data Science at DoorDash)

What is the best structure for a data team?

When it comes to building a world-class data team, the structure is crucial. Jessica Lachs, VP of Analytics and Data Science at DoorDash, strongly advocates for a centralized data team model over an embedded structure.

A centralized model brings multiple benefits:

• Consistent Talent Quality: Centralized teams maintain a high and uniform talent bar. They use a consistent rubric for evaluating candidates, ensuring uniform technical and soft skill levels.

• Growth Opportunities: It offers better career growth options, allowing data scientists to move between different pods, providing varied experiences and reducing boredom.

• Methodological Consistency: Central methodologies and metrics ensure that everyone speaks the same language. This prevents discrepancies like different definitions for key metrics such as sales or retention.

• Scalability: Centralized teams can identify and scale successful methodologies across the organization. Teams avoid duplicating efforts like building the same churn prediction model multiple times.

• Unified Culture: A central team creates a strong, cohesive culture, fostering mutual support, learning, and shared goals.

Lachs emphasizes that while embedded teams might feel more integrated into their respective business units, a centralized model better aligns with overall business goals and maintains high standards across the board.

How do you balance proactive and reactive work in a data team?

Balancing proactive and reactive tasks in a data team requires intentional strategies. Jessica Lachs from DoorDash shares valuable insights on this.

To manage exploratory work while addressing immediate data requests, here are key strategies:

• Set Clear Goals for Exploratory Work: Allocate specific time for self-directed projects. This prevents exploratory work from being overshadowed by immediate tasks.

• Utilize Hackathons: Organize hackathons where the team focuses solely on uncovering new insights. These events stimulate creativity and often lead to impactful discoveries.

• Prioritize and Communicate: Always share the trade-offs when a new task arises. This helps in making informed decisions and ensures that everyone understands the prioritization.

• Create a Supportive Culture: Encourage a culture where curiosity and deep dives are valued. This approach inspires the team to look beyond immediate requests and focus on long-term business impacts.

These strategies ensure that while immediate needs are met, time is also carved out for important exploratory work. This balance is crucial to drive both short-term results and long-term innovation.

By implementing thoughtful prioritization and encouraging a culture of data-driven decision-making, teams can efficiently handle both proactive and reactive tasks.

What are the key qualities to look for when hiring data scientists?

Finding the right data scientists is essential for building a world-class data team. Jessica Lachs, VP of Analytics and Data Science at DoorDash, outlines critical qualities that you should prioritize during the hiring process:

Curiosity

Curiosity drives self-motivation and the desire to explore deeper into data. Data scientists who are naturally curious will not just stop at answering a question; they will dive deeper to find more insights.

• Testing for Curiosity: Include something slightly off in your interview case and observe if the candidate notices and probes further.

Problem-Solving Skills

Great data scientists need to be adept at breaking down complex problems and finding actionable solutions, especially in data-driven decision making.

• Testing for Problem-Solving: Use real-world problems from your company's history in the interview to see how they approach and solve issues on the fly. Assess how they handle new information and pivot when needed.

Technical Proficiency

Although softer skills are important, a strong technical foundation in skills like SQL, Python, and machine learning is non-negotiable.

• Testing for Technical Skills: Conduct a coding exercise as part of the interview process to evaluate their technical abilities effectively.

Additional Tips for Effective Hiring

• Consistency in Evaluation: Maintain a standard rubric across interviews to ensure fairness and uniformity.
• Soft Skills: Look at how candidates react to being wrong or taking in new information. Can they make a decision without having full information?

By focusing on these key attributes, you can ensure that you are hiring data scientists who not only fit the technical bill but also bring a wealth of curiosity and problem-solving prowess to your team. This approach aligns well with the qualities sought in other impactful roles, as described in Gravity Climate's hiring practices.

These strategies ensure you build a robust and dynamic data team capable of driving significant business impact.

How do you define effective metrics for a data team?

Defining effective metrics is essential for driving long-term business outcomes. Jessica Lachs, VP of Analytics and Data Science at DoorDash, emphasizes the importance of simple, actionable metrics.

Key Points to Consider

• Simplicity Over Perfection: Opt for metrics that are easy to understand. A composite score might seem comprehensive but could become meaningless if no one understands it.

  • Example of a good metric: Percentage of merchants with photos.
  • Example of a bad metric: A weighted composite score for "merchant health."

• Short-Term Metrics with Long-Term Impact: Choose short-term, measurable metrics that drive desired long-term outcomes.

  • Example: Instead of setting a goal on retention directly, focus on immediate actions like improved onboarding experiences that can enhance long-term retention.

• Consistency and Actionability: Metrics should translate easily into actions that teams can take.

  • Ensure the metric is understandable and actionable so that teams can make decisions quickly.

• Focus on Edge Cases and Fail States: Metrics should also cover fail states to capture rare but impactful issues.

  • Example: Tracking "never delivered" orders to ensure even rare failures get attention.

These principles align with effective data-driven decision-making. When metrics are simple, actionable, and cover all scenarios, including edge cases, they can more effectively drive desired business outcomes.

How do you manage a global data organization?

Managing a global data organization brings its own set of complexities and nuances. However, the core principles of effective management remain strikingly similar across different regions.

Here are the main points to consider:

• Common Challenges:

  • Currency variations: Different currencies can complicate financial metrics and data comparisons.
  • Language barriers: Miscommunications can arise from language differences, making it crucial to ensure clear translations and understandings.
  • Regulatory distinctions: Regulations may vary widely between countries, requiring teams to adapt their strategies accordingly.

• Universal Core Principles:

  • Consistency in Methodologies: Regardless of location, using standardized methodologies helps maintain coherence and reliability.
  • Unified Culture: Cultivating a robust, unified team culture across all regions fosters mutual support and shared goals, ensuring everyone moves in the same direction.
  • Common Currency for Metrics: Translating different levers of business like price and quality into a common currency facilitates easier decision-making across regions.

• Similarities:

  • Problem Sets: While the specifics might differ, many of the fundamental problems remain consistent, allowing lessons learned in one market to be applied in another.

• Differences:

  • Cultural Nuances: Each country has unique cultural aspects that can influence how teams interact and data is interpreted.
  • Market Dynamic: The competitive landscape and consumer behavior may differ, necessitating tailored approaches for each market.

Overall, emphasizing data-driven decision-making ensures consistent quality and performance, while also adapting to the unique challenges each region presents. Integrating insights from various markets can lead to a more comprehensive and resilient data organization. For more insights into cutting through complex layers in organizations, explore this article on reducing bureaucracy.

How can AI improve the productivity of a data team?

Leveraging AI tools can significantly enhance the productivity of data teams by automating repetitive tasks and providing intelligent assistance. Here are some key ways AI can drive efficiency:

• Automated SQL Query Assistance: AI-powered chatbots can help non-technical users craft SQL queries. This reduces the workload on data scientists and empowers more employees to retrieve data independently.

• Automating Repetitive Tasks: AI can automate tasks like data cleaning, preliminary analysis, and regular reporting. This frees up the team to focus on more complex, high-value tasks.

• Enhanced Data Retrieval: AI tools can streamline the process of finding and interpreting data, enabling faster and more accurate decision-making.

• Predictive Analytics: By utilizing machine learning algorithms, AI can predict trends and anomalies, allowing teams to proactively address potential issues.

• Improved User Support: Implementing AI-driven office hours can support team members in real-time, providing quick solutions and reducing downtime.

A practical example is DoorDash's "Ask Data AI," an internal chatbot that assists with SQL queries. Such tools not only make teams more efficient but also foster a culture of self-service, enabling more data-driven decision-making across the organization. For insights into effective AI integration, check out OpenAI's latest advancements.