Akiri Start free →

How subjects are learned

How to actually learn data science

Data science is three disciplines in a trenchcoat: statistics that must be worked like mathematics, investigation that must be practised on messy real data, and code that must actually run. Study plans that treat it as any single one of the three produce the familiar failure — fluent notebooks, fragile understanding.

The method

Work the statistics until the method is yours, investigate real datasets, and build the code that does both.

Problem sets for the mathematics, open-ended questions against real data, and code that runs.

From the engine These aren't study tips — they are the planning values Akiri's pedagogy engine uses for every data science plan it builds: the session rhythm below, with about 60% of study time spent doing rather than being told.
ExplainDoDoDrill

Planned as a blend: quantitative method 45% · investigation 30% · building working code 25%.

Where it goes wrong

A prerequisite from years ago silently blocking the new topic.

Usually the statistics. A placement that finds the real gap — often two levels below the course you enrolled in — saves the semester.

Tutorials work but real datasets are chaos.

That chaos is the actual discipline. Plans that end each week against a real, messy dataset build the skill the clean tutorial cannot.

What mastery actually looks like

Not a grade — observable abilities. A plan should be able to say which of these it has evidence for:

Keeping it

Mixed problem sets for the methods; returning to your own past analyses for the craft.

Common questions

Should I learn the maths or the coding first?

Neither first — interleaved. The engine plans data science as roughly 45% quantitative method, 30% investigation, 25% building, in the same weeks.

Why can I follow tutorials but not analyse my own data?

Tutorials hand you the question and clean data — the two hardest parts. Practice must include open-ended questions against messy datasets.

Do I need to memorise formulas?

Less than you think; you need method choice — knowing which technique answers which question. That comes from mixed problem sets, not flashcards of formulas.

Engine values reviewed 2026-09-01. Related: mathematics, accounting, statistics.

Akiri turns this into your actual week: tell it your goal and your date, and it measures where you're starting from, builds the plan backwards from the deadline with this discipline's own rhythm, and rebuilds it as you go — in English, French, Spanish, Hindi or Chinese.

Build my data science plan — free