Back to the lessonPreparing the printout…

Prompt Engineering · Module 4 — Prompting for Data-Driven Problem Solving

Period 10 · Diagnose with Data — Case 2

[video] Lecture video — Period 10: Diagnose with Data — Case 2 — /api/lesson-media/videos/4d3653ff-a3ca-4f63-bf8e-bb93c553e599-1794892501132-7f61ab0a6101d8a5.mp4

Flashcards · Diagnose with Data

Task brief — A short statement of audience, input, decision, outputs and success criteria, written before the first analysis prompt.

Audience: the manager. Input: the 200-row CSV. Decision: what to investigate first.

Net sales per transaction — A group's summed net sales divided by its own transaction count, calculated at full precision and rounded only for display.

East: 182,004.38 / 39 = 4,666.78.

Reconcile — Check that parts add back to the whole and that your figures match the file: regional counts to 200, regional sums to 892,252.11, categories to the region.

East categories: 12 + 4 + 10 + 13 = 39.

Descriptive analysis — Reporting the patterns that are present in the supplied data, without claiming why they exist.

Candidate investigation point — A small or odd segment worth a closer look, not evidence of a failure or a cause.

East Electronics: 4 records.

Coverage — Which stores, entities and time periods a data extract includes. Unequal coverage can make counts incomparable.

Hypothesis — A possible explanation that has not been tested with the evidence you hold.

East may have fewer active stores or a shorter reporting window.

Working diagnosis — A provisional interpretation linking observed evidence to a question; it names checked findings, labelled hypotheses and the data still missing.

Claim register — A table that labels each statement as verified finding, verified comparison, working hypothesis, missing input or decision to defer, with what supports it.

Prompt Engineering · Period 10 · Diagnose with Data — Case 2 · printed on November 20, 2026