Learn · Data analyst

📈 Data analyst

Turn raw data into a decision someone acts on.

Analytics interviews test three things in a fixed order: SQL, then reasoning about a business question, then whether you can present a finding without burying it.

The path, in order

1. SQL until it is boring

Joins, aggregation, grouping, subqueries and window functions. SQL is the single highest-return skill here — it appears in nearly every analyst screen and is the most common reason candidates fail one.

2. Spreadsheets and a BI tool

Pivot tables and lookups, then one of Power BI, Tableau or Looker. Which tool matters far less than being able to build a dashboard somebody would actually use twice.

3. Statistics you will really use

Averages versus medians and when each misleads, distributions, sampling, and enough about significance to avoid claiming a result you do not have.

4. One end-to-end analysis

Take a public dataset, ask a specific question, answer it, and write the finding in a paragraph a manager could act on. This single artefact is worth more in interviews than any course certificate.

5. Communication

Lead with the answer, then the evidence. Analysts who bury the conclusion on slide nine do not get promoted.

Where people lose months

⚠Learning Python before SQL. Analyst screens are overwhelmingly SQL, and candidates regularly arrive with pandas fluency and fail on a join.
⚠Dashboards nobody asked for. A chart-heavy portfolio piece with no question behind it signals tool familiarity but not judgement, which is the thing being assessed.
⚠Reporting numbers without a recommendation. 'Revenue fell 8% in the north region' is an observation; the job is to say what you think should be done about it.

Common questions

SQL or Python first?

SQL, without much hesitation. It appears in nearly every analyst interview, is faster to reach a working level in, and is the most common single reason candidates get filtered out at the technical screen.

Do I need a statistics degree?

No. You need to use a small amount of statistics correctly and know when a number is misleading. Over-claiming significance you have not established damages credibility far more than admitting uncertainty.

What makes a good portfolio project?

One specific question, a real dataset, an honest answer, and a short written conclusion a manager could act on. Depth on one analysis beats five dashboards.

Put it to work

School-level foundations for this subject live on our sister site Syllab.

Other paths: Software developer · Accounting & CA · Commerce & economics · English & communication · All paths