The Analytics Portfolio That Actually Gets Interviews

A hiring manager gives a portfolio about five minutes before deciding whether to read properly. Most portfolios lose in the first thirty seconds, and almost always for the same reason: they show that you can operate tools, when the question being asked is whether you can think.

What is actually being screened for#

1. Can you scope a problem? Titanic and Iris were scoped for you. That is why they carry no signal.
2. Can you make a decision under a trade-off and defend it?
3. Can you communicate to someone who is not technical?
4. Would your code survive contact with another person?

Nothing on that list is "knows pandas".

The structure that works#

Three projects. Each with a README that answers five questions in this order:

README.mdmarkdown
# Project title

**The question.** One sentence a business person would recognise as theirs.

**The answer.** What you found. Up front, not buried at the end.

**How.** Data, method, and the three decisions that mattered — with what
you traded away for each.

**What I would do differently.** The single strongest signal of seniority
on the whole page.

**Run it.** Two commands, working.

The "what I would do differently" section does more work than anything else. Someone who can articulate the weaknesses of their own work is someone who will catch problems before they ship.

Choose problems with a real decision in them#

Bad: "Analysis of Netflix titles." Nobody decides anything with it.

Better: "Which of our three delivery regions should get the next depot?" — even with public data, because it forces you to define a metric, handle ambiguity, and commit to a recommendation.

Presentation matters more than you want it to#

• A README with a screenshot beats a notebook with no context.
• A two-minute Loom walkthrough beats a longer written explanation.
• A live deployed dashboard beats a screenshot of one.
• Working code beats elegant code that errors on clone.

Clone your own repository into a fresh folder and run it. A shocking share of portfolio projects fail at this step, and it ends the review immediately.

Three is the number#

Three strong projects, each with a real question and a defended recommendation. Ten thin ones read as inability to judge which work is worth showing — which is itself the thing being assessed.