Wing
The Data and AI Wing
Consulting and finance interviews increasingly open a dataset instead of a case book: a table to query, a sheet to fix, a chart to read, a test result to judge. This wing teaches the whole toolkit in six parts, in the order the work uses it: Excel, SQL, Python, visualization, statistics, and the AI tools now sitting on every analyst’s desk. The first briefing is free.
Open the data problem bank General practice
The six parts
Data 1
Excel for Analysts
The spreadsheet is where banking and consulting analysis actually happens. This part builds modeling-grade habits: clean layout, lookups and pivots that never break, charts that read in seconds, and the keyboard discipline that makes you fast without making you sloppy.
- Spreadsheet layout as a model: inputs, calculations, and outputs kept apart
- Lookups and pivots: XLOOKUP and INDEX MATCH thinking, pivot logic, and when each one lies
- The arithmetic analysts do daily: growth, mix, weighted averages, and unit checks
- Keyboard discipline and the checks that catch a broken sheet before a partner does
Next step: Open SQL for Business Questions and learn to pull the numbers the sheet starts from.
Data 2
SQL for Business Questions
Most business data lives in tables you cannot open in a spreadsheet. This part teaches SQL the way interviews and the job use it: start from the business question, write the query that answers it, and read the result with suspicion.
- The core query: SELECT, WHERE, GROUP BY, and ORDER BY as one sentence
- Joins without fear: what each join keeps, what it drops, and the row count checks
- Window functions for running totals, ranks, and each group's latest row
- Translating a vague business question into a query with a defensible definition
Next step: Open Python for Analysis and do the same work where the data is too big or too messy for SQL alone.
Data 3
Python for Analysis
Python is the second tool on every analyst's desk, used when the data is large, messy, or repetitive. This part teaches the pandas way of thinking in plain language, so a non-programmer can follow every step and check every result.
- Tables in Python: rows, columns, filtering, and grouping as familiar operations
- Cleaning data honestly: missing values, duplicates, and types that are lying
- Merging tables and predicting the row count before you run it
- Plotting from a table and reading your own output like a skeptic
Next step: Open Data Visualization and learn to turn the result into the exhibit a decision gets made on.
Data 4
Data Visualization
Every analysis ends as an exhibit someone glances at for ten seconds. This part teaches choosing the chart that answers the question, keeping axes honest, and writing titles that carry the conclusion instead of naming the data.
- Chart choice by question: comparison, trend, composition, and distribution
- Honest axes, baselines, and the tricks that make small differences look enormous
- Exhibit craft for decks: one message per page, titles as conclusions, annotation
- Reading exhibits like an interviewer: mix effects, overlapping percentages, and the so-what
Next step: Open Statistics for Decisions and learn how much of what the chart shows you should believe.
Data 5
Statistics for Decisions
Statistics in business is not a math exam. It is the discipline of knowing how much a number can be trusted: what an average hides, what a sample can and cannot say, and how to read a test result without superstition.
- Averages and spread: mean against median, and the outlier that moves one and not the other
- Sampling: what a sample mean estimates, and how sample size buys confidence
- Significance in plain words: what a p-value is, and the three ways it gets abused
- Reading an A/B test: base rates, absolute against relative lifts, and when to ship
Next step: Open AI Tools for Analysts and put an assistant to work inside the habits you have built.
Data 6
AI Tools for Analysts
AI assistants now draft queries, explain code, and write documentation at analyst speed. This part teaches using them as junior analysts you supervise: fast, useful, and never trusted with a number you have not checked yourself.
- Drafting SQL and Python with an assistant, then reading what it actually wrote
- Verification habits: reconcile totals, sample rows, and rebuild one figure by hand
- Where AI is silently wrong with numbers: invented figures, plausible joins, stale definitions
- Documentation and explanation: making the tool write the first draft of the audit trail
Next step: Open the data problem bank and put all six parts under graded conditions.
The problem bank
Eighteen graded problems, six at each of Easy, Medium, and Hard, spanning SQL, spreadsheet work, chart reading, Python, and estimation with data. Every problem is graded on the server, every solution is worked in full, and your solves build a data readiness signal that stays separate from your case readiness, because firms score them separately too. Three problems, one per tier, are free.
Browse the problem bank