Role roadmaps / Data Analyst Role flowchart
Turn messy tables into answers someone can act on. SQL, spreadsheets, a BI tool, and the nerve to say the number is wrong.
Also called: Business Analyst (data-heavy), Reporting Analyst
You ship: Trusted dashboards, ad-hoc SQL, and a written recommendation — not a model.
8–12 weeks to be useful; months to be trusted
Roadmap Coding Interview Question bank Switch roleData Analyst BI Analyst Data Engineer Analytics Engineer Data Scientist ML Engineer MLOps Engineer AI Engineer Data Architect Nearby roles:BI Analyst , Analytics Engineer , Data Scientist
Checked skills
0% · click a node for concepts, stacks, access, and what to produce · drag to pan · Ctrl-scroll to zoom
− Fit + Clear checks
Start · Data Analyst Foundations SQL you can defend Filters, joins, GROUP BY, and grain. If you cannot say what one row means, stop. Spreadsheets without shame Pivot, VLOOKUP/XLOOKUP, and why a CSV is not a warehouse. Stats that change a decision Distributions, sample vs population, and “is this lift real?” Daily craft Clean before you chart Nulls, duplicates, timezone, and currency. Document the grain. Exploratory analysis Cut the number by segment before you average it. Write the question first. One BI tool, deeply Looker, Tableau, or Power BI — pick the one your company pays for. How you ship Metric definitions Write the SQL behind the KPI. Version it. Argue the filter with finance. Write the so-what One chart, one sentence, one recommended action. Hide the 14 extras. SQL that does not melt the warehouse Pruning, partitions, and not SELECT * into a workbook. Career moves The analyst interview loop SQL pad, a take-home, and “tell me about a time the number was wrong.” Where this role goes Analytics engineer if you like dbt. Scientist if you like experiments. Manager if you like people. Keep practicing Study this node
SQL you can defend Filters, joins, GROUP BY, and grain. If you cannot say what one row means, stop.
Analyst screens open with SQL. Windows come next, not first.
Write SQL you can defend: one row has a meaning, the join does not fan out, and the filter matches the question.
Time to learn 2–3 weeks of daily SQL before you lean on windows.
Prerequisites SELECT / FROM / WHERE without copying a tutorial A fact table whose grain you can say in one sentence What you need This site’s SQL sandbox or a warehouse login (read is enough) A orders/customers-style schema (star, not a dump) A notebook or comment habit: grain above every saved query Someone who will tell you if revenue looks wrong Concepts to study Grain: what one row represents after the FROM/JOIN INNER vs LEFT JOIN and when a NULL is a finding GROUP BY vs window functions — collapse vs keep rows WHERE vs HAVING; filter before vs after aggregation NULL three-valued logic (unknown is not false) Anti-joins (NOT EXISTS / LEFT JOIN … IS NULL) for “without” Tech stacks PostgreSQL / warehouse SQL — The dialect of the pad and of productionSQLite (this site) — Practice joins and windows in the browserdbt / a SQL IDE — Version the query once it is a metricHow to study Memorize grain first: customers, orders, order_items — say it out loud before SELECT. Drill filters, GROUP BY, and INNER/LEFT until they are muscle memory. Add windows only after you can explain PARTITION BY vs GROUP BY. Read one window-functions guide; then redo the same KPI as a window and as an aggregate. What to produce Completed revenue by country you can defend out loud A query that would double revenue if you joined the wrong grain — and the fix A two-minute answer: “what does one row mean?” Practice on this site SQL Foundations track, then Warehouse SQL Core Open the sandbox and reconstruct “completed revenue by country” from memory Write the grain in a comment above every query you keep Pitfalls Joining a fact to a dimension at the wrong grain (fan-out that inflates revenue) Filtering on a LEFT JOIN column in WHERE, which quietly turns it into INNER SELECT * into a spreadsheet and calling it analysis Interview prompts “What does one row in this result mean?” Write revenue for completed orders by country — then name the join that would double it. Links
Mark as done All Articles Stacks Practice Learn Pages
No results. Try analyst, numpy, snowflake, or roadmap.
↑↓ navigate↵ openesc closeAliases: sf, numpy, analyst, mlops
