Problem: 100% inspection of every production lot is too costly and slow, but accepting lots without any inspection risks letting defective batches reach customers.

Work: Designed a single sampling plan (n=121, c=4) against AQL=1.5%/LQL=6.5% targets using the binomial distribution, derived its theoretical OC curve, then applied the plan to 3,240 real production lots by drawing actual inspection samples from each lot's true defect rate.

Tools: R (base) · Binomial distribution theory · OC curve analysis

Outcome: Empirical acceptance rates across 3,240 lots tracked the theoretical OC curve closely, validating the plan; accepted 2,331 lots (71.9%) with a mean Average Outgoing Quality of 1.15% defective.

Statistical Quality Control · University of Zimbabwe · 2026

Problem: A retail network needed to know, for every product at every store, which items were about to stock out and how much to reorder.

Work: Used SQL to compute real demand statistics (mean and standard deviation) from 731 days of daily sales history per store/product, then built a live Excel reorder-point model with statistical safety stock, EOQ, and a risk dashboard.

Tools: SQL (SQLite), Excel

Outcome: Flagged 62 of 100 store/product combinations as below their statistical reorder point, surfacing $2.3M in suggested reorder value and pinpointing which categories and stores needed attention most.

Problem: Leadership lacked a single up-to-date view of sales performance, forcing manual cross-checking across multiple spreadsheets to answer basic questions like which regions or products were underperforming.

Work: Consolidated raw sales data into a clean model and built an interactive Power BI dashboard with drill-downs by region, product category, and time period, including KPI cards for revenue, growth, and target attainment.

Tools: Power BI, Excel, DAX

Outcome: Gave stakeholders a self-service view of sales performance, cutting the time spent answering ad-hoc reporting requests and surfacing underperforming segments earlier.

Problem: Multiple financial time series exhibit non-stationarity and potential long-run equilibrium relationships — naive regression on such series produces spurious results, requiring cointegration and equilibrium modeling techniques.

Work: Built a rigorous 7-D analytical handbook applying ADF & KPSS unit-root tests, Johansen cointegration, and a Vector Error Correction Model (VECM) to 10 years of daily AAPL, MSFT, and GLD prices. Concluded with a live pairs-trading blueprint using ECT Z-score signals.

Tools: Python · statsmodels · yfinance · Pandas · Matplotlib · Seaborn · SciPy

Outcome: Confirmed one cointegrating vector across AAPL–MSFT–GLD; identified structural breaks during COVID-19 (2020) and the Fed rate-hike cycle (2022); designed a VECM-GARCH extension roadmap for production deployment.

HASTS 211 · University of Zimbabwe · 2026