About - Eric McKnight

I am a business-focused data scientist with experience spanning analytics engineering, reporting, and applied data science. My work centers on using causal inference and modeling to support high-impact business decisions, particularly in environments where experimentation is constrained and data is imperfect.

Photo of Eric McKnight

My career path has been intentionally non-linear. I began in finance and risk-focused roles, transitioned through underwriting and analytics, and eventually moved into data-focused engineering and modeling work. Along the way, I developed a strong foundation in both the technical and business sides of data—learning how systems are built, how metrics are used, and how decisions are actually made.

Today at Intuit, I focus exclusively on data science initiatives, applying techniques such as propensity modeling, causal inference, and Bayesian methods to real-world partner and growth problems. I’m particularly interested in bridging the gap between rigorous statistical methodology and practical business application—ensuring that models don’t just perform well, but meaningfully influence decisions.

Outside of project work, I enjoy digging into methodological edge cases, improving data pipelines, and thinking about how analytical rigor translates into operational impact.

Resume

My resume reflects a career progression from analytics and data engineering into applied data science, with a current focus on causal inference and decision-driven modeling. While my formal title has not yet changed, my recent work has been exclusively data science–focused.

Selected Experience Highlights

  • Led causal inference analyses to quantify the impact of partner outreach initiatives, translating model results into staffing and investment recommendations.
  • Developed propensity models to assess partner health and identify high-potential and at-risk partners, supporting tiering and engagement strategy.
  • Built and maintained end-to-end data science pipelines using Python, R, SQL, BigQuery, and Databricks.

Download PDF Resume

Projects

Below are selected projects that represent my recent data science work, with a focus on causal inference, propensity modeling, and decision-driven analytics. All projects are described at a high level to respect data privacy and confidentiality.

Estimating the Impact of Partner Outreach

Context

A service partner team conducted outreach meetings with scale partners to drive activation and improve downstream performance. Because outreach was not randomly assigned, traditional experimentation was not feasible.

Approach

I applied causal inference techniques to estimate the effect of initial outreach meetings on the likelihood of a partner generating a new booking within 30 days. The analysis accounted for selection bias and baseline partner characteristics.

Outcome

The estimated lift was translated into an expected dollar impact, enabling leadership to evaluate staffing needs and investment levels required to scale the outreach program effectively.

Methods & Tools

Causal inference (PSM, IPTW, AIPW), Bayesian methods, Python, R, SQL, BigQuery

Partner Look-Alike and Health Modeling

Context

Partners are segmented into managed and scale tiers, with managed partners receiving dedicated support. Identifying high-potential scale partners and managed partners at risk of decline is critical for effective resource allocation.

Approach

I developed a propensity model to estimate how closely a partner’s behavior aligns with that of high-performing managed partners. The model produces a continuous score that serves as a proxy for partner health.

Outcome

The propensity score is used to inform partner tiering decisions and guide Partner Development Manager discussions, helping teams prioritize outreach and proactively address partner risk.

Methods & Tools

Propensity modeling, classification, feature engineering, Python, SQL, BigQuery, Databricks

Next Best Action Modeling for a Hypothetical Sales Team

Context

To explore next best action (NBA) modeling in a setting where production data is not publicly available, I designed and implemented an end-to-end NBA framework using synthetically generated sales and customer interaction data.

Approach

I generated a realistic synthetic dataset representing customer behaviors, engagement history, and sales outcomes, then built a modeling pipeline to predict the most effective action for a sales team to take at a given point in time. The project emphasizes problem formulation, feature engineering, and decision-oriented model outputs.

Outcome

The resulting model demonstrates how NBA frameworks can be applied to guide sales prioritization and engagement strategy. While based on synthetic data, the structure and methodology are directly transferable to real-world use cases.

Methods & Tools

Classification modeling, feature engineering, synthetic data generation, Python, scikit-learn, XGBoost

View project on GitHub