I use data to answer business questions.

My analysis is mostly in Python, on retail sales, store-level and public commercial-district data. The studies here use group comparisons (ANOVA, Tukey HSD), propensity-score matching, fixed-effects panel regression, bootstrap confidence intervals, and logistic regression, random forest and gradient boosting models. The method follows the structure of the data and the question being asked, for example matching where groups differ in size and household fixed effects where the same households are observed week after week.

I report estimates with confidence intervals or tests, and I describe observational results as adjusted associations rather than causal effects. Where the data cannot separate two explanations, I keep both and state what additional data would settle it.

Each case study has its own limitations section covering what the design cannot identify and which data or analysis would address it. Revenue Concentration and Seoul Closure Risk were team projects; each page states which part is mine. Campaign Response Heterogeneity grew out of a team project on the same dataset; its Scope note explains the details.