The owner of an online shop sees three different revenue numbers in three spreadsheets and cannot tell why sales dipped last month. You build one dashboard on the order data that explains what changed by product, channel and region.
A data analyst answers business questions with data. You pull and clean it, find the pattern, and present a clear recommendation. The value is the right question and an honest answer, not the number of charts.
Goal: Go from zero to answering business questions with spreadsheets, SQL and simple charts, and use AI tools to speed the work without trusting them blindly.
Before you start
Beginner · 7 weeks
Week 1: Data vocabulary
Rows, columns, tables · Types: numbers, text, dates · Mean, median, percent change · What a metric is
Practice: Describe a small dataset in plain words and compute five summary numbers by hand.
Practice: Describe a small dataset in plain words and compute five summary numbers by hand.
Common pitfalls
Think first, then reveal
Mean or median?
Think first, then reveal
What does a JOIN do and what can go wrong?
Think first, then reveal
Why verify AI-written SQL?
Think first, then reveal
What makes a chart misleading?
Ask what decision the dashboard supports and agree on definitions such as "revenue" and "active customer".
Turning a request like "why are sales down" into specific questions and written metric definitions.
Why this pick
Three revenue numbers usually mean three unwritten definitions, and the dashboard fails until they are agreed.
What the role must know
Beginner mistake
Core
Analysis
Communication
Think first, then reveal
Mean or median?
Think first, then reveal
What does a JOIN do and what can go wrong?
Think first, then reveal
Why verify AI-written SQL?
Think first, then reveal
What makes a chart misleading?
Think first, then reveal
Why not stop an A/B test as soon as it looks significant?
Think first, then reveal
What is Simpson's paradox?
Think first, then reveal
How do you handle a metric that changed after a tracking release?
Think first, then reveal
Where do natural-language data assistants fail?
Think first, then reveal
How do you tell an executive an uncertain result?
Think first, then reveal
When is a quasi-experiment acceptable?
Think first, then reveal
How do you stop metric sprawl?
Think first, then reveal
How do you govern AI-generated analysis?
Build one, measure it and put it on your CV. Each has tools, setup commands and milestones.
CV, portfolio and interview loop checklist →