Data practice: a claim a stakeholder can trust, with limits.

Who this track is for

Students who like finding the sentence hiding in a table, including beginners who are willing to start small. You do not need a programming identity. You need care.

Modules in a data flavor

Research defines the decision the data must serve. Analysis is the center. Consumer needs asks who is affected by the metric. Recommendation is what to do given the caveat. Next steps is what data you still lack.

Quality bar

A dictionary of fields. Cleaning notes. A simple baseline before a clever visual. A sensitivity or a segment split. A limits paragraph that would still be true if a journalist read it.

Ethics

No scraping private data. No deanonymizing. No fake surveys. Classroom or public sets only unless a written cohort data pack exists.

Resume language

Name the question, the tool, and the caveat. Do not claim “data scientist” as a job title from this track.

Cleaning is part of the analysis

Write down what you dropped and why. A beginner who documents cleaning is more employable later than a beginner who produces a mysterious chart. GVI marks the dictionary of fields and the limits paragraph as seriously as the visual.

Do not average things that should not be averaged. Do not hide small samples in a big percentage. If n is tiny, say so and recommend a next measurement instead of a sweeping strategy.

The claim sentence

“Among the public rows we have, X is associated with Y, which suggests Z, unless the missing segment behaves differently.” That shape is a data-track recommendation. “The data shows we should disrupt the industry” is not a shape; it is a wish.

Limits paragraphs that keep you employable later

Say what the table cannot tell you. Say who is missing. Say whether a percentage sits on a tiny n. Say what measurement you would run next if you had lawful access. That paragraph is the difference between a student intern and a chart decorator. GVI marks it.

Do not dress a thin table in a complicated visual. Beginners hide in complexity. A simple comparison with a clear caveat is stronger. If a mentor cannot restate your claim after one look, the visual failed even if it looks expensive.

Ethics again, because data tracks get tempted: no private scraping, no fake surveys, no deanonymizing classmates. If the only way to make the project feel “real” is to cross a line, shrink the question. Reality includes law.