What data science is, in plain words

Data science is answering questions with recorded facts instead of opinions. That is the whole idea. Everything else — Python, charts, machine learning — is only tooling built on top of it.

A school keeps a register of who attended class. A shop keeps receipts. A hospital keeps patient files. A mobile network keeps call records. Each of those is data: facts that were written down. A data scientist takes that pile of facts and turns it into an answer someone can act on.

The four kinds of question

Almost every data project is one of these, and they get harder in this order:

• Descriptive — what happened? "How many students passed last term?"
• Diagnostic — why did it happen? "Why did the pass rate drop in Term 2?"
• Predictive — what will happen? "Which students are at risk of failing?"
• Prescriptive — what should we do? "Which 30 students should get extra tutoring first?"

Notice that the first two are the majority of real work, and they need no machine learning at all. A beginner who can answer descriptive and diagnostic questions well is already useful to an employer.

What the day actually looks like

Newcomers imagine advanced mathematics. The honest split of a working day is closer to:

• Getting the data — asking for it, exporting it, joining two files that disagree with each other.
• Cleaning it — fixing dates typed five different ways, removing duplicates, deciding what to do about blanks. This is usually the largest part.
• Exploring it — counting, averaging, grouping, charting until you understand its shape.
• Explaining it — one chart and three sentences that a manager can act on.
• Modelling it — the machine-learning part. Real, valuable, and much smaller than people expect.

What it is not

Data science is not proof. Data records what was measured, and measurement is never complete. If your attendance register only records students who showed up to sign it, your data cannot tell you why the others stayed home. Knowing what your data cannot answer is a senior skill, and you can start practising it today.

The three job titles you will hear

• Data analyst — mostly describes and diagnoses. SQL, spreadsheets, dashboards.
• Data scientist — adds prediction and statistics. Python, models.
• Data engineer — builds the pipes that move data so the other two have something to work with.

The titles overlap everywhere and nobody agrees on the boundaries. Start with the analyst skills; they are the foundation of all three.

Try it yourself

Before you touch a computer: write down one question about your school, workplace or neighbourhood that you believe recorded facts could answer. Then write down what data you would need, and who holds it. Keep the note — it becomes your first project at the end of this track.