Skip to course content
Free SQL course

SQL for Data Analysis and AI

Module 13 Activity

Scenario

This is the deliverable the whole course has been building toward. You are handing an analysis to a sceptical reviewer who will not be able to ask you anything. Everything they need to trust it - or to reject it - must be in what you send.

Task

Work through the five stages in order. Do not start writing the memo before stage 3 is done.

  1. Frame eight questions. Each one must fix a population, a grain, and a measure. Use the eight in

Unit 13.01 or substitute your own to the same standard.

  1. Build one analysis-ready dataset as staged CTEs, with a grain check after every join. Answer all

eight questions from that single dataset.

  1. Reconcile every figure you intend to report: parts against wholes, at least one independent source,

and a magnitude check.

  1. Write the evidence memo - the headline figures, the population each covers, the checks you ran, and

the limitations.

  1. Review your own work as a sceptic, using the checklist below, and fix what it surfaces.

Deliverable

Three files:

Check your work

Your dataset must hold 1,000 rows. Anything else means a join changed the grain.

QuestionAnswer
Orders placed / completed1,000 / 988
Booked revenue₹27,01,463
Completed revenue₹26,85,905
Average completed order value₹2,718.53
Customers who never ordered3,812
Items sold3,400
Payments received minus revenue owed₹0.00
Customer country coverage93.8%

Marking rubric

CriterionWhat earns it
Questions are answerableEach names population, grain, and measure
Dataset is correctGrain checked after every join; 1,000 rows
Figures reconcileParts sum to wholes; one independent source agrees
Limitations are statedMissing countries, pending orders, and the line-item discrepancy all appear
Memo is self-containedA reader can verify every number without asking a question
Normalisation is documentedThe city rule is stated, applied once, and its gaps acknowledged

The sceptic's checklist

Read your own memo and ask, for each number: what population is this over? What would make it wrong? Which check would have caught that? If any figure has no answer to the third question, it is not yet evidence - it is just output. Three limitations belong in every honest version of this memo: the 300 customers with no country, the 12 pending orders, and the 999 orders whose line items do not reconcile to their order total.