Why Python, and when Excel still wins
The problem
Someone tells you to learn Python because your spreadsheet is slow. That is not a reason on its own, and switching tools for the wrong reason costs you a month and leaves you slower than when you started.
The useful question is not "which tool is better". It is "has this particular job crossed a line".
What you will have at the end
A test you can apply in ten seconds to any task on your desk, and an honest sense of what Python will and will not fix.
The three things that move a job across the line
Repetition. Anything you do every month is the strongest signal. A monthly report takes the same four hours in Excel every single month. In code it takes four hours once and three minutes thereafter. The first month you lose; by month three you are ahead and you stay ahead.
Reproducibility. If anyone might ask how a number was produced — a donor, an auditor, a colleague taking over — the answer has to be a file someone can re-run, not a memory of which cells you sorted. This is the reason that matters most in programme work and the one people think about last.
Size. Past roughly a hundred thousand rows Excel stops being pleasant, and past a million it stops being possible. This is the reason everyone cites first, and in practice it is the least common of the three.
If none of those apply, Excel is the right answer and you should use it without feeling bad about it.
When Excel is genuinely better
- A one-off. Something you will do once and never again. Writing a script is slower.
- Anything you need to see while you think. Exploring an unfamiliar table by eye is faster in a grid than in code.
- Work you hand to someone who does not code. A workbook they can open beats a script they cannot run.
- Small, visual, final. A table of twelve numbers for a slide.
What Python will not fix
It will not make bad data good. A survey with no validation produces nonsense faster in Python than in Excel, and prints it with more confidence.
It will not make you faster in week one. Expect to be slower for a fortnight.
And it will not stop you making mistakes — it will only make your mistakes repeatable, which is genuinely an improvement, because a repeatable mistake can be found and fixed.
Where it breaks
| What you tell yourself | What is actually happening |
|---|---|
| "Excel keeps crashing, I need Python" | Usually volatile formulas or whole-column references, not file size. Fix those first. |
| "I'll rewrite the whole workflow in Python" | Rewriting everything at once is how these projects die. Move one step. |
| "The numbers changed and I don't know why" | A manual step nobody wrote down. This is the reproducibility case, and it is the real reason to move. |
Your turn
Take the last report you produced by hand and count two things: how many steps you repeated from the previous month, and how many of those steps you could describe precisely enough for a colleague to reproduce.
If the second number is smaller than the first, that gap is your reason.
Next
Install Python without the headache — one recommended path rather than six.