A clean dataset teaches the wrong reflex
If the data is tidy, the technique is the whole task. Real analysis is mostly deciding what the data does not say, and a project that never has a missing value never asks for that.
For people who teach analysts
A capstone your students cannot fake.
Every course ends with a dataset and a known answer, so every student finishes with a certificate and no idea whether they are any good. This is fifty minutes in a company that is not clean, where the dashboard is wrong and nobody has done the analysis for them.
No sign-up, and nothing is recorded. It is the same environment your students would sit.
The problem with a tidy final project
The answer is already online
A fixed dataset has a fixed answer, and by the second cohort it is in a public repository. Here the company is generated, so there is nothing to look up.
A certificate says they finished, not that they can
Which is why employers discount them. A student who has never been measured against a real problem walks into their first interview without knowing where they stand.
What a student actually does
The same thing a candidate does in a real hiring process, because it is the same product.
-
They meet a company, not a CSV
Twelve colleagues who each see one part of it and disagree with each other. Six to twelve tables of its own CRM, billing and marketing spend, with tens of thousands of rows.
-
They work in SQL, Python or a pivot table
You choose which, per cohort. It all runs in the browser — nothing for a student to install, and nothing that fails the night before a deadline.
-
They find what the tables are hiding
A third of the deals have no source recorded. A date column holds when somebody updated a record, not when anything happened. Take it at face value and you reach a confident wrong answer — which is the lesson.
-
They commit, and say how sure they are
Nine decisions, each with a confidence and the result behind it. Being confidently wrong costs more than being tentatively wrong, which is a thing worth learning before a job rather than during one.
What they get that a certificate cannot give them
A report about how they think
- Whether they did the analysis before committing, on each of nine decisions — not just whether they were right.
- Whether they knew when they were sure. Confidently wrong and unevidenced is marked separately from wrong, because they are different problems.
- Every query they ran, in order, including the ones that went nowhere.
- A shorter version they can share or print, carrying the six measures and what would move each of them — and none of the questions or answers.
Somewhere they actually stand
- Scored the same way as candidates sitting it for real jobs, so it is a rehearsal rather than an exercise.
- Specific enough to act on: a student who never checks before committing has a different problem from one who checks and misreads.
Where this actually is
A small number of courses first, free, while the parts that are specific to teaching get built. Being straight about which is which:
Works today
- The whole assessment — the company, the console, the questions, the report.
- Five analyst seats, so a marketing course and a finance course set different work.
- SQL, Python or a pivot table, chosen per cohort.
- A generated company per cohort, so there is no answer key to circulate.
- A shareable summary per student, as a link or a PDF, with the questions and answers left out of it.
Being built with the first courses
- A bank of companies reserved for students, so somebody who sat yours can never meet the same one in a real hiring process.
- A certificate carrying your brand rather than ours.
- Cohort results as a spreadsheet, rather than one report at a time.
If one of those three is the reason you would say yes, tell me and it goes first. That is what a pilot is for.
Join the pilot
Tell me what you teach and how big a cohort is. If it fits, you get it free while it is being built out, and a say in what gets built.