Real workplace · observed data
Case 01 · Diagnosis · Retail operations
A replenishment dashboard was 80% wrong for a year. Two weeks of data remediation fixed it.
A daily fill list flagged about 250 items; roughly 200 were false positives. The dashboard logic was
correct. The bin capacities, on-hand counts and department codes feeding it were not. BPMN models in
three states, fishbone and five-whys, four data views, and six requirements traced from cause to
control.
~250 → ~50daily list items
~80% → minimalfalse positives
1 yr → 2 wkslifespan to fix
Canadian Tire, Toronto — my own workplace. Figures are
observed values; where a KPI was not formally re-measured, the case study says so.
Read case study 01 →
Academic · fictional client · nothing built
Case 02 · Specification · Healthcare
Writing a specification precise enough that someone could build from it — then auditing it
until it held.
Online new patient registration for a clinic: current- and future-state process models, a use case
with five alternate flows, 13 business rules, a 14-element data catalogue, three specified screens
with a clickable recreation, and end-to-end traceability. Then a QA pass that found 13 defects in our
own spec.
1 + 5use case & alternate flows
13 · 14rules & data elements
13spec defects found
Academic team project, George Brown College, May 2026. Sole
author of UC001 within a team of seven. The clinic is fictional and the stakeholder was played by
faculty. Nothing was built, deployed or measured, and no outcome is claimed.
Read case study 02 →
