Your Portfolio Has a Reading Problem

Meet Dani. Dani spent six weekends on a churn analysis: 41 notebook cells, a cleaning step that would make a tax inspector weep with joy, a model, and one genuinely useful finding at the end. Dani linked it in thirty applications.

Thirty applications later, Dani had heard nothing. Silence isn’t feedback. It just feels like it.

(Dani is a composite: a few portfolios we’ve reviewed, blended so nobody gets recognised at a family dinner. You may still recognise Dani in the mirror. We did.)

What was wrong with the project?

That’s the question Dani kept asking, and it’s the wrong one. The analysis was sound. The code ran. The finding was real. If you had sat Dani down and asked for a walkthrough, you’d have been impressed by minute four.

Nobody sat Dani down. That’s the part portfolio advice tends to skip: a portfolio isn’t marked like an exam. Nobody is obliged to finish it. It’s read like a shop window, by someone with a full inbox, a coffee going cold and a stack of other windows to look into.

You’ve probably heard the folklore that recruiters spend six seconds on a CV. The exact number is disputed, and we’re not going to pretend we know it. The behaviour is easier to believe, because you do it too: skim the top, then decide whether to keep going.

Here’s the test we ran on Dani’s project

Open it like a stranger would. No context, no scrolling yet. What’s on the first screen?

A title: churn_analysis_final_v3. Eleven import lines. A CSV being loaded. A df.head() showing five rows of customer IDs. Which raises the uncomfortable question: if that was all you saw, what would you think this person can do?

“Import pandas,” mostly.

So where was the good bit? We kept scrolling. Past the cleaning. Past the exploration. Past three charts with default titles and one with no title at all. And then, finally, there it was: the sentence that belonged on the first screen.

Cell 37.

The work wasn’t bad. It was buried. It had something worth saying and had put it where nobody would hear it, after thirty-six cells of throat-clearing.

Flip it

The fix isn’t more work. It’s the same work in a different order. We build analysis from the bottom up: imports, data, cleaning, exploring, answer. People read it from the top down. So lead with the thing they came for, and let the method follow for anyone who wants it.

Now, what this doesn’t mean. It doesn’t mean hiding the code. The people who hire analysts do care how you got there. They just won’t find out unless the first screen gives them a reason to.

The three lines that do the heavy lifting

Here’s the change we’d make first, because it’s the smallest: put three lines at the top of every project, before a single import.

On LucidPlexus, this is exactly what the snap button on a guided-project workbench cell is for. Snap the cell that holds the answer, the table or the chart, and post that, rather than the thirty-six cells that got you there. Then link the full project underneath for the people who want the whole walk.

As for Dani: same analysis, three new lines at the top, one chart moved up, a title a human could say out loud. Nothing else changed. Whether that’s enough is the next thirty applications’ job to tell us.

One question before you open your own portfolio

If someone read only the first screen of your best project, and nothing else, what would they think you do for a living?

And would they be right?