Data Analyst Learning Path: Skills, Tools and Steps

What a data analyst actually does

Job titles in this field are used loosely, which leads people to prepare for the wrong role. The distinctions matter because the skills and the interviews differ.

• ✓In practice an analyst spends most of the day writing queries, cleaning data, building or maintaining reports, and explaining findings to people who are not analysts.
• ✓The hardest part is rarely the query. It is working out what question the requester actually needs answered, which is often not the one they asked.

Step 1: spreadsheets (3-4 weeks)

Do not skip this because it feels basic. Spreadsheets remain the most widely used analysis tool in the world, and fluency in them is assumed in interviews for analyst roles.

• ✓Lookup and reference functions, conditional aggregation, and text functions for cleaning messy fields.
• ✓Pivot tables until they are second nature - they are how most business questions get answered quickly.
• ✓Charts, and the judgement to choose the right one rather than the most impressive one.
• ✓Data cleaning: duplicates, inconsistent categories, dates stored as text, and trailing spaces that break every join you attempt later.

Step 2: SQL, deeply (6-8 weeks)

This is the skill that gets you hired. Analyst interviews are largely SQL interviews, and the difference between someone who can write a SELECT and someone who can answer a real business question in SQL is enormous.

• ✓Foundations - SELECT, WHERE, ORDER BY, LIMIT, and the difference between WHERE and HAVING.
• ✓Aggregation - GROUP BY, the aggregate functions, and multi-level grouping.
• ✓Joins - inner, left, right and full, plus self-joins. Know exactly what a left join does to unmatched rows, because that is where wrong numbers come from.
• ✓Subqueries and common table expressions, which make complex logic readable.
• ✓Window functions - ROW_NUMBER, RANK, LAG, LEAD and running totals. These separate intermediate analysts from beginners and appear constantly in interviews.
• ✓Date handling, string manipulation, CASE expressions and NULL behaviour. NULL logic causes more silent errors than any other feature.
• ✓Practise on realistic multi-table datasets, not single-table toy examples, since joins are where the real difficulty lives.

Step 3: statistics you will actually use (4 weeks)

You do not need a mathematics degree. You need enough statistics to avoid confidently reporting nonsense - which is the most damaging thing an analyst can do.

• ✓Descriptive statistics: mean, median, mode, variance, standard deviation, percentiles, and when the median is the honest choice.
• ✓Distributions and skew, and why an average is misleading for skewed data such as income or session length.
• ✓Correlation, and the discipline never to present it as causation.
• ✓Sampling and bias: how a sample can mislead, and why survivor and selection bias appear in almost every business dataset.
• ✓Hypothesis testing and p-values at a working level, including what a p-value does not mean.
• ✓A/B testing basics - what you are comparing, why sample size matters, and why stopping a test early inflates false positives.

Step 4: visualisation and BI tools (3-4 weeks)

• ✓Learn one BI tool well - Power BI, Tableau or Looker Studio. Which one matters far less than being genuinely fluent in it.
• ✓Data modelling inside the tool: relationships, calculated fields and measures. This is where most self-taught users stop and where the real capability begins.
• ✓Chart selection by intent: comparison, composition, distribution, relationship, trend over time.
• ✓Design discipline - clear labelling, honest axes, no unnecessary decoration. A truncated y-axis exaggerating a trend is a competence question as much as an ethics one.
• ✓Build dashboards that answer a question rather than displaying everything available. 'What decision does this support?' is the test.

Step 5: Python for analysis (6-8 weeks)

Python is what you reach for when spreadsheets and SQL run out - repeated cleaning, bigger datasets, statistical work, and anything you need to reproduce reliably.

• ✓Core Python: variables, control flow, functions, lists and dictionaries. You need enough to write scripts, not to build applications.
• ✓pandas for loading, cleaning, filtering, grouping, merging and reshaping data. Merging is SQL joins again, which is why SQL comes first.
• ✓A plotting library for exploratory charts, where speed matters more than polish.
• ✓Notebooks for exploration, and plain scripts for anything that runs repeatedly.
• ✓Reproducibility habits: version your work, avoid manual steps, and never edit the source data in place.

Step 6: the portfolio that gets interviews

Most data portfolios fail the same way: a well-built dashboard on a famous public dataset with no question and no conclusion. Hiring managers see hundreds of those. What they are looking for is evidence that you can turn a vague business question into an answer somebody could act on.

• ✓Start with a question that has a stake in it. 'Which customer segment drives repeat purchases, and what should marketing do differently?' beats 'an analysis of the Titanic dataset'.
• ✓Use messy data and show the cleaning. The cleaning is the work, and hiding it hides your competence.
• ✓State the answer plainly at the top, then the supporting analysis, then the limitations. Leading with the conclusion is exactly what the job requires.
• ✓Name the limitations honestly - what the data cannot tell you is a senior instinct and it is rare in juniors.
• ✓Three projects is plenty: one SQL-heavy business analysis, one dashboard built for a specific decision, and one Python analysis with statistical reasoning.
• ✓Write each one up as if for a manager, not for a marker. Short, conclusion-first, no jargon.

Interviews and the first job

• ✓Expect a SQL test, often live. Practise writing joins, aggregations and window functions while talking through your reasoning.
• ✓Expect a case question: given this business situation, what would you look at? Structure the answer - clarify the goal, name the metrics, state the data you would need, then the analysis.
• ✓Expect to explain a project. Be ready for 'what would you do differently' and 'how confident are you in that number'.
• ✓Expect basic statistics questions, especially about correlation, sampling and A/B tests.
• ✓Domain knowledge is a genuine advantage. An analyst who understands the industry asks better questions, so aim your applications where you already know something.