How to become a Data Analyst

A complete, honest roadmap for the skills data analysts actually use, from spreadsheets and SQL through statistics, data cleaning, visualization, business understanding, storytelling, a programming language, and AI-assisted analysis. It runs top to bottom, foundational to advanced, so you always know what comes next. Free to read, no signup required.

1. 01

   Step 01Foundations: how analysts think

   Data analysis is a way of thinking before it’s a set of tools. Start with what the job actually is: turning messy data into decisions someone acts on.

   • The analysis lifecycle: Question, gather, clean, analyse, visualise, communicate. Almost every project follows this loop.
   • Asking the right question: The most valuable skill: translating a vague business ask into a precise, answerable question. A perfect answer to the wrong question is worthless.
   • Analytical reasoning: Thinking in comparisons, trends, and segments, and staying skeptical of a number until you understand how it was produced.
   • Types of analysis: Descriptive (what happened), diagnostic (why), predictive (what might), and prescriptive (what to do), and knowing which one a question needs.

   Tools change; the reasoning doesn’t. Analysts who can frame a question and interrogate a number stay valuable regardless of which software is fashionable.

   Analysis lifecycleFraming questionsAnalytical reasoningDescriptive vs diagnosticSkepticism
2. 02

   Step 02Spreadsheets: Excel & Google Sheets

   Spreadsheets are still the most-used analysis tool in the world. Deep spreadsheet skill is a genuine differentiator, not a beginner’s crutch.

   • Formulas & functions: Lookups, conditional logic, aggregation, and text and date functions: the workhorses of everyday analysis.
   • Pivot tables: Summarising and slicing data quickly, often the fastest path from raw rows to an answer.
   • Cleaning in spreadsheets: Deduplicating, splitting and combining columns, and fixing inconsistent data by hand when that’s the right tool.
   • Charts: Building clear charts directly, and knowing a spreadsheet’s limits so you graduate to SQL and BI tools at the right time.

   FormulasLookupsPivot tablesConditional logicData cleaningCharts
3. 03

   Step 03SQL: the core analyst skill

   SQL is the single most important technical skill for a data analyst. Almost all real business data lives in databases, and SQL is how you get it.

   • Querying fundamentals: SELECT, WHERE, ORDER BY, and filtering, to retrieve exactly the rows you need.
   • Joins: Combining data across tables. This is the concept that makes real analysis possible, since business data is spread across many tables.
   • Aggregation & grouping: GROUP BY, aggregate functions, and HAVING to summarise data into metrics.
   • Advanced querying: Subqueries, CTEs, and window functions for running totals, rankings, and period-over-period comparisons.

   If you learn one technical skill deeply, make it SQL. It’s the most requested skill in data-analyst roles and it transfers across every database and BI tool.

   SELECT & filteringJoinsGROUP BY & aggregatesSubqueries & CTEsWindow functions
4. 04

   Step 04Statistics & analytical reasoning

   Statistics is what separates real analysis from pulling numbers. You don’t need a maths degree. You need the practical concepts that stop you from drawing wrong conclusions.

   • Descriptive statistics: Mean, median, spread, and distributions, and why the median often tells the truer story than the average.
   • Correlation vs causation: The single most important idea in analysis: two things moving together doesn’t mean one causes the other.
   • Sampling & bias: How data is collected shapes what it can tell you: survivorship bias, selection bias, and unrepresentative samples.
   • Significance, carefully: The idea of statistical significance and confidence, enough to avoid over-reading noise as a real signal.

   Descriptive statsDistributionsCorrelation vs causationSampling & biasSignificance
5. 05

   Step 05Data cleaning & preparation

   Analysts spend a large share of their time cleaning data, and the quality of that work sets a ceiling on every conclusion that follows.

   • Messy real-world data: Inconsistent formats, duplicates, typos, and mixed types: what real data always looks like before it’s useful.
   • Missing values: Spotting missing data and deciding how to handle it (drop, fill, flag) without distorting the result.
   • Types & consistency: Getting dates, numbers, and categories into consistent, correct types so analysis behaves.
   • Validation: Sanity-checking totals and ranges to catch errors before they become wrong conclusions.

   “Garbage in, garbage out” is the analyst’s law. A beautiful dashboard built on dirty data is worse than no dashboard, because it’s confidently wrong.

   Messy dataDeduplicationMissing valuesType consistencyValidation
6. 06

   Step 06Data visualization

   A chart is how most people will ever experience your analysis. Choosing and designing visuals well is a core, durable skill: the tools change, the principles don’t.

   • Choosing the right chart: Matching chart type to the question (comparison, trend, distribution, composition, relationship) instead of defaulting to a pie.
   • Design principles: Clarity over decoration, honest axes, meaningful colour, and removing everything that doesn’t help the reader.
   • Dashboards: Assembling visuals into a coherent view that answers a set of questions at a glance, with BI tools like Tableau, Power BI, or Looker.
   • Interactivity: Filters and drill-downs that let stakeholders explore without needing you for every question.

   Chart selectionDesign principlesDashboardsTableau / Power BI / LookerInteractivity
7. 07

   Step 07Business & domain understanding

   The best analysts understand the business, not merely the data. Context is what turns a query result into a recommendation someone acts on.

   • Metrics & KPIs: What the key business metrics are, how they’re defined, and how they connect: revenue, retention, conversion, and the like.
   • Unit economics: The basic economics of the business: what drives cost and value, and how a metric moves the bottom line.
   • Domain knowledge: Learning the specific field you work in (e-commerce, finance, healthcare) so your analysis reflects how it actually operates.

   Metrics & KPIsUnit economicsFunnelsDomain knowledge
8. 08

   Step 08Storytelling & communication

   An insight nobody understands or acts on is wasted. Communicating findings clearly and persuasively to non-technical people is what makes an analyst influential.

   • Narrative structure: Leading with the “so what”, then supporting it, rather than walking through every query you ran.
   • Presenting to stakeholders: Tailoring depth and language to the audience, and answering the decision they actually face.
   • Written communication: Crisp summaries, clear recommendations, and captions that make a chart’s point explicit.
   • Honesty about uncertainty: Being clear about caveats and confidence so decisions rest on an honest picture.

   The gap between a good analyst and a great one is usually communication, not technical depth. The insight only matters if it changes a decision.

   NarrativeStakeholder communicationExecutive summariesData storytellingUncertainty
9. 09

   Step 09A programming language: Python or R

   Beyond spreadsheets and SQL, a programming language opens up bigger data, automation, and repeatable analysis. Python (with dataframes) is the common default; R is strong in stats-heavy fields.

   • Language basics: Variables, control flow, and functions, enough to manipulate data programmatically.
   • Dataframes: Loading, cleaning, transforming, and aggregating data in a dataframe library, the analyst’s power tool beyond the spreadsheet.
   • Automation & reproducibility: Turning a one-off analysis into a repeatable script or notebook others can rerun and trust.
   • Charting in code: Producing visualisations programmatically for analyses too large or repetitive for a BI tool.

   Python or RDataframesData wranglingNotebooksReproducibility
10. 10

    Step 10Experimentation & A/B testing basics

    Analysts are increasingly asked to help run and read experiments. Understanding the basics keeps a team from shipping changes on the strength of noise.

    • Hypotheses: Framing a clear, testable hypothesis before looking at results, rather than fishing for a story afterward.
    • Control vs treatment: Why comparison groups matter and how randomisation isolates the effect of a change.
    • Reading results honestly: Significance, sample size, and the traps (peeking, multiple comparisons) that produce false wins.

    HypothesesControl vs treatmentA/B testingSample sizeReading results
11. 11

    Step 11How data is stored: modelling & warehousing

    A little understanding of where your data comes from makes you faster and more self-sufficient, and it’s the bridge toward analytics-engineering work if you want it.

    • Databases & warehouses: The difference between operational databases and analytics warehouses, and why analysts mostly query the latter.
    • Data models: Tables, keys, and the idea of fact and dimension tables (the star schema) that structure most analytics data.
    • ETL awareness: How raw data becomes clean, queryable tables, enough to know where numbers come from and where they can go wrong.

    Databases vs warehousesFact & dimension tablesStar schemaETL awarenessData lineage
12. 12

    Step 12AI-assisted analysis

    AI is changing analyst work, helping write queries, summarise findings, and explore data faster. Used well it’s a force multiplier; used carelessly it produces confident nonsense.

    • AI for querying & code: Using AI to draft SQL, write dataframe code, and explain unfamiliar data, with your review, since it can get joins and logic subtly wrong.
    • Summarising & exploring: Accelerating first-pass exploration and narrative drafts, while you verify every number against the source.
    • Judgement stays human: AI can compute, but framing the question, checking the data, and standing behind the conclusion remain yours.

    AI makes the mechanical parts faster, which raises the value of the judgement it can’t do: asking the right question, trusting the right data, and owning the recommendation.

    AI-drafted SQLSummarisationVerifying outputAnalytical judgement