Frame
Clarify the decision, define metrics and build hypotheses.
Career Lab · Module 20
The Analyst Case.
This is the course compressed into one realistic assignment: an ambiguous stakeholder question, imperfect data, multiple possible explanations, and a decision that needs evidence.
Core idea
The capstone is not graded on how much SQL you write. It is graded on how defensible your decision-making process is.
01
Context
QuickCart growth stalled. Leadership wants one answer.
“Growth slowed this quarter. Tell us what changed, what matters most, and what we should do next.”
There is no instruction telling you which table, chart or method to use. You decide what the business question means, which metrics matter and what evidence is sufficient.
AI is allowed, but your submission must make validation visible and you must be able to explain every important query, metric and conclusion.
02
Capstone workflow
Use the full analyst operating system.
Analyze
Model the data, query it, explore drivers and quantify impact.
Validate
Check integrity, uncertainty, alternative explanations and AI output.
Communicate
Deliver a dashboard, memo, recommendation and next-step plan.
Rule
A polished output cannot compensate for a weak definition, broken denominator or unsupported recommendation.
03
Challenge
Complete the end-to-end case.
Exercise · 20–30 minutes
Your portfolio project.
- What is the decision behind “growth slowed”?
- Which 3–5 metrics define the problem?
- How will you decompose growth into drivers?
- Which data-quality risks could invalidate the analysis?
- Where will you use AI and how will you validate it?
- What recommendation can you defend with the available evidence?
04
AI assist
Use AI with an audit trail.
Capstone AI policy
AI is allowed for:
- drafting SQL/Python,
- brainstorming hypotheses,
- checking edge cases,
- critiquing visualizations,
- editing communication.
You remain responsible for:
- metric definitions,
- source selection,
- validation,
- statistical/causal interpretation,
- final recommendations.
For every material AI-assisted output:
record what context you provided, what you checked,
and what changed after validation.05
Validation
The capstone submission checklist.
01
Is the business question explicitly framed?
02
Are important metrics defined and reproducible?
03
Does the analysis preserve grain and pass QA checks?
04
Are uncertainty and alternative explanations addressed?
05
Can every recommendation be traced to evidence?
06
Is AI use documented and independently validated?
06
Final deliverables
Submit work that looks like an analyst’s real output.
SQL/Python with definitions, QA and reproducible logic.
A focused decision-facing view with drill-downs.
One page: change, evidence, confidence, recommendation.
Problem → approach → insight → decision → reflection.
Course finish
You are done when another analyst can reproduce the work, a stakeholder can understand the decision, and you can defend what the evidence does — and does not — support.