A reviewer gives your portfolio about eight seconds before deciding whether to scroll. Most candidate portfolios lose in those eight seconds — not because the analysis is weak, but because the first screen shows a notebook instead of a conclusion.
Lead with the answer
Every project page should open with a single sentence a non-analyst could repeat in a meeting:
Churn concentrates in accounts that skip onboarding call two — moving that call earlier recovers an estimated 4.1% of annual revenue.
Then the chart. Then the method. Then, far below, the notebook.
Three projects, one narrative
You need exactly three, and they should not be three of the same thing:
- A cleaning project. Deliberately ugly source data. Show the decisions: what you dropped, what you imputed, and why.
- A modelling or trend project. Something with a forecast or a comparison that carries risk if you're wrong.
- A dashboard. Built for a named audience with a stated decision it supports.
Together they say: I can get data into shape, reason about it, and hand it to somebody else.
Tooling that reads as professional
- Pandas for the transformation layer, with named steps rather than one 40-line chain
- Matplotlib or Seaborn for analysis figures, labelled and captioned
- Tableau for the shareable dashboard
- Excel for the model a stakeholder will actually open and edit
The things reviewers quietly check
- Does the repo have a README that states the question?
- Are the numbers reproducible from raw data?
- Is there any statement of uncertainty at all?
- Would this survive somebody asking "how do you know?"
What to cut
Cut the Titanic dataset. Cut the notebook with forty cells of df.head(). Cut every chart that doesn't support the headline sentence. A tight portfolio of three sharp projects beats nine mediocre ones, and it takes less of your weekend to maintain.