Conjoint Research @ Wharton

Sped up the statistical models behind a Wharton consumer-research project (Python, scikit-learn), cutting simulation runtime and storage roughly 20x while keeping the results just as statistically valid.

DomainConsumer behavior modeling, pricing sensitivity, simulation pipelines

StackPython, scikit-learn, NumPy, SciPy

Impact~20x reduction in runtime/storage with maintained significance

~20x faster: optimized simulation and estimation pipelines cut runtime and storage roughly 20x while preserving statistical significance.

Implementation Notes

Benchmarked existing and proposed models with statistical tests to ensure efficiency gains did not degrade inferential quality.

Implemented simulation and utility-estimation workflows to support part-worth computation on larger datasets.

Built analysis tooling for willingness-to-pay and tradeoff interpretation to make outcomes easier to use downstream.

Integrated quickly into an in-flight research codebase and delivered iterative improvements collaboratively.