Data analyst from scratch: a route

What an analyst does

Not draw charts for their own sake. Answer business questions with numbers: why sales fell, which advertising pays for itself, why people stop using the service.

Ordinary work looks like this: a question arrives from a manager, you work out what they actually want to know, find the relevant data, calculate, check whether the numbers are lying to you, and explain the result in plain language.

That last point is underrated. A calculation nobody understood is useless.

What to learn, in order

1. Spreadsheets. Yes, the ordinary spreadsheet. Pivot tables, formulas, working with large exports. It remains the primary tool at a great many companies, and it's the easiest place to start understanding data.

2. Database queries. The main skill. Selecting, filtering, grouping, joining, window functions. Without it you don't work as an analyst: the data lives in a database and you have to get it out yourself.

3. Statistics at a practical level. Means and medians and how each misleads differently, proportions, distributions, what significance means. Not a university course but an understanding of when a number means something and when it's noise.

4. A visualisation tool. One: any reporting and dashboard system. The task isn't to draw beautifully but to present so the conclusion is obvious.

5. Python for analysis. Libraries for tabular data and charts. Needed when there's a lot of data or the calculation is complex: what takes a day in a spreadsheet takes ten lines here.

How long it takes

Four to eight months of regular study. Faster than development if you're comfortable with numbers, longer if maths was never your friend.

Adjacent experience accelerates it considerably: reporting, finance, procurement, marketing. Such people already understand the domain and only need the tools.

Building a portfolio

Easier than for a developer: you don't need a working service, you need an investigation.

Take open data and answer a meaningful question. Not "looked at the statistics" but "found that accident rates in the city are higher at junctions without traffic lights, here's the evidence".

Present it as a story: the question, where the data came from, what you did, what emerged, what it contradicts.

Attach the code and the dashboard. So it can be checked.

Two or three such investigations are enough for first applications.

What to expect at entry

A first analyst job is rarely exciting: exports, reports, checking other people's numbers. That's a normal stage, followed by harder work.

And an honest warning: a substantial part of the job is establishing where data came from and whether it can be trusted. There's little romance in it, but it's exactly where a good analyst is distinguished from someone who calculates from whatever they were handed.

What strengthens an application

Experience with a real client. Even a small one: calculating something for a friend who runs a business, helping a non-profit make sense of their data.

That counts for more than practice investigations, because there were live requirements, incomplete data, and the need to explain the result to someone who doesn't work with numbers.