Senior Data Scientist - Bayesian Experimentation & Causal Inference

Position: Senior Staff Data Scientist - Bayesian Experimentation & Causal Inference

New York, New York, United States;

San Francisco, California, United States;

Seattle, Washington, United States

Headway’s mission is a big one – to build a new mental health care system everyone can access. We’ve built technology that helps people find great therapists with the first software-enabled national network of providers accepting insurance.

1 in 4 people in the US have a treatable mental health condition, but the majority of providers don’t accept insurance, making therapy too expensive for most people. Headway is building a new mental healthcare system that everyone can access by making it easy for therapists to accept insurance and scale their practice.

Headway was founded in 2019. Since then, we’ve grown into a diverse, national network of over 60,000 mental healthcare providers across all 50 states who run their practice on our software and have served over 1 million patients. We’re a Series D company with over $325m in funding from a16z (Andreessen Horowitz), Accel, GV (formerly Google Ventures), Spark Capital, Thrive Capital, Forerunner Ventures and Health Care Service Corporation.

We want your time here to be the most meaningful experience of your career.

Join us, and help change mental healthcare for the better.

Join us to build the truth engine behind better mental health outcomes.

As a Senior Staff Data Scientist, Bayesian Experimentation and Causal Inference

, you will be the company-wide owner of how Headway learns from data, especially when the stakes are high and the signal is noisy. You will report directly to the Head of Data and serve as a core leader for standards, frameworks, and decision quality across Product, Growth, Ops, and Finance.

Your work will set the default methods for how we answer questions like:

• Did this actually cause the outcome we care about?
• How sure are we, and what should we do given that uncertainty?
• What evidence is strong enough to change strategy, policies, or spend?

A major objective of this role is to build and institutionalize a clear map of “what we know” about patients, providers, and payers

, with explicit confidence levels that tie directly to business action. Think of it as a shared language that prevents the organization from treating a correlation like a law of physics, while still moving fast.

What you will do

• Own causal inference and experimentation standards across Headway. Define the canonical approaches, guardrails, documentation, and review mechanisms for experiments and quasi-experiments, including when and how to use Bayesian methods.
• Build the confidence ladder for company knowledge. Create a clear, shared framework that maps findings to levels of confidence (for example 1–10), where lower levels reflect correlation and early directional evidence, mid levels reflect increasingly credible causal inference, and the highest levels reflect stable, repeatable, decision-grade truths. Operationalize it so it shows up in artifacts teams actually use: PRDs, launch reviews, growth planning, quarterly business reviews, and postmortems.
• Design the learning strategy for our hardest questions. Lead the approach for ambiguous, high impact domains like provider activation and retention, payer economics and policies, patient conversion and engagement, and marketplace dynamics. Recommend the right combination of randomized experiments, stepped rollouts, geo tests, natural experiments, and observational designs.
• Raise the organization’s statistical maturity. Introduce and standardize Bayesian experimentation practices where it improves speed and decision quality (priors, posterior interpretation, sequential decision rules, credible intervals, expected value framing). Build training, playbooks, and reusable tooling.
• Be the escalation point for difficult measurement problems. Tackle issues like interference and spillovers, network effects, selection bias, noncompliance, measurement error, multiple comparisons, seasonality, and Simpson’s paradox showing up in real life and causing confusion.
• Partner with Data Platform and Engineering to make rigor scalable. Ensure experimentation and inference are supported by instrumentation, logging, metric definitions,…