Economists at Instacart

Tech companies are increasingly choosing to hire PhD students in Economics.

I read an article about the Economics Team at Instacart, and here are my takeaways.

Projects they worked on

The economists at Instacart, part of their Econ Team, work on a variety of challenging projects that blend economics with machine learning.

  1. Optimized Bidding for Ad Auctions:
    • Economic Understanding: The team applied their knowledge of ads incrementality, auction formats, and auction pressure to develop a comprehensive understanding of the ads marketplace.
    • Technical Solution: Developed an algorithm for optimized bidding, tailored for a high-volume, latency-sensitive environment, offering flexible and effective bidding options for advertising partners.
    • Impact and Collaboration: Worked closely with the Ads team, achieving strong results and providing significant value to advertising partners through this initiative.
  2. Contextual Bandit Algorithms for Incentive Targeting:
    • Economic Analysis: Utilized econometric methods to estimate heterogeneous treatment effects and optimize trade-offs between different business metrics.
    • Model Development: Created both batch inference and real-time inference models, integrating them into the system for effective upsell placement.
    • Production Integration: Ensured the models were production-ready, aligning them with Instacart's operational environment and achieving practical utility in a real-world setting.
  3. Causal Inference and Experimental Design:
    • Long-term Outcome Measurement: Used surrogate models to measure long-run outcomes, demonstrating the capacity to understand and predict extended impacts of various initiatives.
    • Natural Experiments: Leveraged natural experiments for experiment design, including treatments inspired by regression discontinuities, showcasing innovative approaches to problem-solving.
    • Non-randomized Treatment Effects: Estimated causal effects of treatments like Instacart+ membership and developed customer segments for enhanced model inputs and experimental analysis, contributing significantly to understanding customer behavior and preferences.

Their Value as Economists:

  1. Holistic Problem-Solving Approach: Merging economic theory with machine learning to tackle complex, multidimensional problems, providing end-to-end solutions.
  2. Efficient and Effective Solution Development: Rapid development and deployment of solutions, facilitated by the dual expertise in economics and technology, showcasing a streamlined process for handling complex, data-intensive tasks.
  3. Insight into Economic Principles and Causal Inference: Profound grasp of economic theories and causal relationships, driving data-centric decision-making and refining experiment designs for improved outcomes.
  4. Versatility and Adaptability: Cross-functional capabilities, enabling work across diverse product areas and problem types, adding a unique, interdisciplinary perspective to traditional tech roles, and catalyzing innovation and growth.