Why I stopped putting percentage bars on my CV

For a long time my portfolio had a row of progress bars. Python 85%. TensorFlow 60%. Deep Learning 65%. They looked precise. They looked like data.

They were not data.

The question I could not answer

A friend reviewing my site asked a simple question: 85% of what?

Not 85% of what a beginner knows, because I was well past that. Not 85% of the language specification, because nobody has read all of it. Not 85% of what the person interviewing me knows, because I had never met them.

I had picked a number that felt modest but confident, and drawn a bar to match it.

What a number implies

Writing 85% makes three claims at once:

1. There is a fixed thing called "all of Python"
2. I have measured my distance from it
3. That distance is meaningfully different from 84% or 86%

All three are false. The first is the worst, because expertise is not a container you fill. Five years in, I know less of "all of Python" than I thought I did at two years, because I have since learned how much there is.

The more precise the number, the more it promises. A bar at 85% promises a measurement. There was no measurement.

What I use instead

Grouped lists of what I actually reach for:

Natural language   spaCy · NLTK · Transformers · NER · text classification
Vision             OpenCV · CNNs · YOLO · object tracking
Infrastructure     Docker · AWS · Git

No numbers. If someone wants to know how deep any of it goes, that is what the projects are for, and a repository is a far better answer than a bar chart.

The bars were doing something for me, though, and it is worth naming: they made a short list look substantial. Twelve bars fill more space than twelve words. That is not a good enough reason to keep them.