Senior Data Scientist, Fetch

• 5+ years of experience in data science, machine learning, or applied analytics, with demonstrated ownership of complex analytical problems in product-driven environments
• Experience owning an audience, measurement, or data product from an ambiguous question through delivery and a business decision
• A track record of working regularly with product, engineering, and commercial partners, with the ability to clearly explain both the customer need and your own contribution
• Experience building useful datasets from fragmented, incomplete, or biased sources, and showing how you assessed their validity
• Strong population inference skills, including generalizing from an observed sample to a broader population and checking who’s missing
• Experience building propensity or audience models and validating them against real-world usefulness (population coverage, calibration, and measured outcomes), not just offline accuracy, then using the results to improve the next iteration
• Hands-on causal inference experience, including designing an experiment or observational study you can defend (the comparison, assumptions, diagnostics, and limits) and separating attribution from true incrementality
• Strong SQL, Python, applied statistics, and predictive modeling skills, plus a habit of building reusable, reproducible assets
• The ability to explain scientific findings clearly to nontechnical and customer audiences
• Sound judgment about when to move independently and when to seek review because assumptions or scope have outrun the evidence

Desirable

• Independently own a defined audience or measurement product, from problem framing through delivery
• Translate advertiser goals and purchase behavior into audience definitions, labels, and predictive models
• Build and validate statistical expansion beyond Fetch users, testing quality, calibration, and reach on held-out populations while keeping observed purchases distinct from inferred propensity
• Combine receipt, partner, panel, survey, and contextual data into reliable datasets, accounting for linkage uncertainty, missingness, and coverage, and applying weighting or calibration where warranted
• Design credible studies linking ad exposure to receipt-observed purchases, using experiments when feasible and well-justified observational methods when they aren’t
• Check receipt coverage, exposure matching, selection bias, and uncertainty, and keep attribution and incrementality distinct
• Turn customer problems into scoped products with clear success criteria and launch recommendations, balancing audience quality, reach, advertiser value, user value, revenue, and cost
• Feed campaign outcomes back into audiences and models, using holdouts and fresh data to confirm improvements are real
• Partner with Sales and GTM to shape pilots and explain the product’s value, presenting methods and limitations without overpromising precision or lift
• Build reproducible datasets, analyses, and validation checks, partnering with Product, Engineering, and ML engineers to move prototypes into production
• Document your methods, share reusable code, and pass on what you learn to the team