Portfolio

&Collar (D2C Men's Dress Shirts)  ·  2026

Built a full statistical analysis pipeline and interactive Power BI dashboard around a simulated 30-day A/B test for a D2C apparel brand. The core analytical question: when a site change lifts conversion rate but drops average order value, is it actually worth deploying? This project demonstrates that trusting a single metric blindly can be deceiving.

• Variant converts 14% more visitors but each buyer spends 17% less — the AOV drop outweighs the CVR gain.
• Deploying the Variant as-is would risk reducing annual revenue by ~$100K at current traffic levels.
• Breakeven AOV is $67.57 — a $3.87 lift through upsells or bundles makes the Variant the winner on both metrics.

Nomatic (Premium Travel Gear)  ·  2026

Built a demand forecasting and inventory planning pipeline addressing a challenge multi-channel brands face: D2C sales are smooth and seasonal, but B2B wholesale orders are erratic — most days have zero volume, then a retailer suddenly orders 1,500 units. A single forecast model can't handle both channels, so this analysis uses different strategies for each and combines them into a unified inventory recommendation with explicit risk trade-offs.

• D2C forecast achieves 13.3% MAPE; B2B orders cluster at quarter-ends (z = 4.44, p < 0.00001).
• B2B spike sizes follow a log-normal distribution (KS test p = 0.52), enabling probabilistic modeling.
• Combined 6-month inventory recommendation: ~42,900 units (base) to ~48,500 units (safety stock).

Swire Coca-Cola  ·  2026

Applied the full RGM analytical toolkit to 3 years of beverage distribution data (7.8M transactions, 11 product lines, 158 weeks). The analysis covers five pillars: revenue decomposition into volume/price/mix effects, promotional ROI measurement, XGBoost-driven volume driver identification, Monte Carlo pricing simulation across 7 scenarios, and multi-method demand forecasting.

• 2023–2024 revenue declined -0.9%, driven by volume loss despite stable pricing.
• Only 1 of 9 products (Core Power) showed positive promotional ROI.
• Best forecasting method: Prophet (avg MAPE 16.6%); XGBoost volume driver model achieves R² = 0.61.

Crumbl Cookies  ·  May 2026

Engineered a multi-source data pipeline integrating web-scraped menu history from CrumblCookieFlavors.com, a flavor catalog, and Reddit sentiment analysis to analyze Crumbl's flavor rotation patterns, category distribution, and popularity signals. Built a predictive model for flavor return likelihood using machine learning.

• Identified rotation patterns and category distribution trends across Crumbl's flavor history.
• Random Forest model predicts flavor return likelihood based on rotation history and popularity signals.

1st Place Winner  ·  October 2025

Collaborated with an interdisciplinary team to secure 1st Place in the 2025 Wasatch Cup, the region's premier healthcare case competition hosted by the BYU Healthcare Leadership Association. Our winning solution proposed a mobile health clinic to expand critical access to underserved populations in Southern Utah. I spearheaded the data validation phase, modeling financial outcomes to prove the clinic model was both scalable and financially viable.

• Built time-series forecasting models to project patient demand and clinic utilization over a 3-year horizon.
• Successfully differentiated our strategy from top competing teams across the state, winning the grand prize.