Social Scientist, Performance (Durham, NC)

Hive Science LLC; Durham (NC); Hybrid; USD 120,000 - 180,000; Full time

Job summary

Hive Science LLC seeks a research-focused scientist to design and scale experimentation for ad tech products. You’ll run randomized trials, quasi-experiments, and incrementality studies, and build repeatable frameworks for fast testing in live campaigns.

You’ll analyze real-world campaign data (impressions, CTR, ROAS), validate predictive models, and communicate uncertainty clearly to commercial teams. Hybrid in Durham, US.

Qualifications

• Advanced quantitative training at PhD level or equivalent.
• Proven ability to design and analyze experiments in real-world settings.
• Experience with ad tech, marketing measurement or media data.
• Strong coding in Python and/or R, with large datasets.

Responsibilities

• Design rigorous experiments (randomized, quasi-experimental, incrementality).
• Build repeatable experimentation frameworks for product/partner testing.
• Analyze real-world campaign data (impressions, CTR, conversions, ROAS).
• Validate predictive models and calibrate uncertainty.
• Communicate assumptions and limitations to commercial audiences.

Skills

Experimental design

Causal inference

Advertising data

Python

R

Claude Code

Model validation

Communication to audiences

Education

PhD or equivalent

Tools

Python

R

Claude Code

Job description

About the role:

Design and scale experimentation

• Design rigorous experiments for Hive products, including holdouts, back-tests, and A/B, geo and incrementality tests with customers and partners.
• Build repeatable experimentation frameworks, so each new product or partner can be tested quickly and consistently.
• Maintain and extend Hive's validation standards, covering out-of-sample performance, calibration, stability and uncertainty.
• Work with Hive's scientists to keep results scientifically meaningful as well as statistically significant.
• Help set the scientific bar for when a product is ready to ship.

Work with real-world campaign and media data

• Analyze campaign performance data such as impressions, reach, frequency, CTR, conversions, CPA and ROAS.
• Design tests that account for how ad platforms actually work: auction dynamics, delivery optimization, pacing, audience overlap, and non-random delivery that can confound naive comparisons.
• Work with large, messy real-world datasets, and know which questions the data can and can't answer.

Make it commercially useful

• Design experiments that answer the questions customers, partners and procurement teams actually ask.
• Turn results into case studies, benchmarks and proof points for commercial teams.
• Make pragmatic tradeoffs among rigor, cost, speed and what's feasible in live campaigns.
• Work with Product to prioritize the experimental work that most strengthens Hive's differentiation.
• Communicate assumptions, uncertainty and limitations clearly to technical and non-technical audiences.

Use AI as a force multiplier

• Use Claude Code and other AI tools daily to accelerate coding, analysis, diagnostics and documentation.
• Independently validate AI-generated code, assumptions and interpretations, and help set team standards for reproducible AI-assisted science.

Requirements:

• Exceptional experimental design skills. You've designed and run many experiments, in the field and in live commercial settings, and know how to get clean answers from messy real-world conditions.
• Deep expertise in causal inference, including randomized experiments, quasi-experimental methods and incrementality measurement.
• Hands-on experience with advertising and campaign data: impressions, CTR, conversions, CPA, ROAS, attribution and lift measurement.
• A working understanding of how ad platforms and delivery algorithms behave (auctions, optimization, pacing, targeting) and how they bias experimental results.
• Commercial experience in ad tech, marketing measurement, media, consumer intelligence, market research or consumer-facing products.
• Strong experience validating predictive or modeled data: calibration, generalization, uncertainty and bias.
• Advanced quantitative training (PhD or equivalent experience) in a field such as social or quantitative psychology, computational social science, statistics, economics, quantitative marketing or a related discipline.
• An instinct for the difference between a model that predicts well and one whose outputs mean what they claim to.
• Experience with geographic and multilevel data is a plus.
• Strong coding in Python and/or R, with comfort working directly with large datasets.
• Fluency with Claude Code, including an understanding of AI failure modes and how to verify output.
• The ability to explain complex methods simply and persuasively to commercial audiences.

Backgrounds that could be a strong fit:

We care more about experimental rigor, commercial judgment and creativity than about a specific discipline. Strong candidates might come from marketing science, ad measurement, experimentation teams at ad platforms or tech companies, media agencies, behavioral science, quantitative social science, or economics/econometrics.

Additional Information:

• Location: Hybrid, 3 days a week in our Durham office. You must live within commuting distance.
• Employment type: Full-time.
• Work authorization: You must have an existing right to work in the US. We're not able to offer visa sponsorship for this role.