Be sure of the number before you say it.
Guides: think with data, check the answer
Short, practical guides on asking precise data questions, checking numbers from dashboards and AI assistants, and learning SQL and data analysis from zero. Worked examples use a fictional practice marketplace or clearly labelled sample data.
Start hereAtterna SQLWrite it, read it, check it: SQL for the age of AI.6 questionsFind your levelWhere to start: SQL from zero, the case rounds or the full simulator.Free toolAI Answer CheckerPaste an AI answer: see every number it leaves undefined and the question to send back.Free · in your browserData analyst CV checkSee which skills your CV shows, with the lines behind each verdict and free practice for your gaps.Free toolSRM checkerIs your A/B test split broken? A chi-square check in your browser.Task · 30 secondsWeekly puzzlePick an answer, read the explanation and try the free case.Game · 3 minutesWrong NumberOne database, three colleagues and three right answers. No SQL.Interview · 2 minutesInterview questionA junior analyst interview question on real-looking data, with the reasoning behind it.
Check an AI’s answer
Numbers and SQL from ChatGPT, Claude or a dashboard: what to ask before you forward them.
- How to check numbers an AI gives you: a 7-question checklistChatGPT, Claude or a dashboard gave you a number. Before you forward it, ask seven questions: the definition, the period, the filters, duplicated rows, the denominator, the average and a known total. Worked examples on a fictional marketplace dataset.6 min read · 1 task inside
- How accurate is ChatGPT at data analysis? A small test on 12 SQL questionsWe asked an OpenAI model 12 everyday analyst questions about a practice marketplace, ran every SQL query it wrote and checked each number against our own. The SQL itself was mostly fine. The misses were about meaning — and they are exactly what to check.7 min read · 2 tasks inside
- AI-generated SQL — 6 mistakes that run without an error (and how to catch them)ChatGPT, Claude and Copilot write SQL that runs and looks right, yet double-counts after a join, drops rows in a LEFT JOIN or counts orders as customers. Six real mistakes to try on this page, and the questions that catch them.4 min read · 6 tasks inside
- How to ask an AI a question about your data (so the answer is usable)A vague question can produce a precise answer to the wrong problem. A five-part template — decision, population, measure, period, exclusions — with before-and-after prompts and what to ask after the answer.4 min read · 1 task inside
- Data literacy for AI: the skills to check an answer before you actLearn the data skills you need to use AI well: define metrics, check denominators, spot misleading comparisons and verify SQL. A worked example and a free practice path.8 min read · 1 task inside
Learn SQL
From the first table to the traps interviewers ask about — with tasks on the page.
- Learn SQL from zero on a realistic marketplace dataset: a 10-day planA day-by-day plan for learning SQL from scratch on business data — from reading a table to JOIN traps and window functions — free in your browser: days 1–2 with no account, all ten days with a free account.4 min read · 1 task inside
- Why your SQL JOIN returns duplicate rows (and doubles your total)Two tables, ten rows each: how many after a JOIN? Practise this SQL interview question, watch the rows multiply, then fix a sales total without losing real orders.6 min read · 2 tasks inside
- LEFT JOIN with a WHERE clause: why rows disappear (and how to fix it)You wrote LEFT JOIN to keep every buyer, added WHERE o.status = 'paid', and almost half of them vanished. Why a WHERE on the right table turns a LEFT JOIN into an inner join, how ON fixes it, and how to spot it by counting rows.7 min read · 2 tasks inside
- SQL interview questions for data analysts — 8 practice problems with answersEight common SQL interview questions for data analysts — COUNT vs COUNT DISTINCT, NULL, WHERE vs HAVING, grouping, JOIN totals, LEFT JOIN filters, keys and one analyst trap — each one you can answer on this page, with the reason behind every option.3 min read · 8 tasks inside
- Average vs median: which one to report, and why the average misleads with moneyThe average order on one marketplace is €60.92, the median €43.96, and the top 10% of orders bring almost a third of the money. When to use the mean, when the median, and how to report both without misleading anyone.4 min read · 1 task inside
Metrics and decisions
Define what you count, read an experiment, and say what a number cannot tell yet.
- One question, three right answers: define the metric before you countOne database, three colleagues, three different customer definitions. Learn which questions to ask before counting.3 min read · 1 task inside
- Metric definition template (with example): agree what to count before writing SQLYou send the number. “That’s not what I meant.” A practical metric definition template, clarifying questions and a worked conversion example to use with a colleague or an AI assistant.6 min read · 1 task inside
- How long should you run an A/B test? Sample size, duration and the checks before you read itThe duration of an A/B test comes from the smallest effect you care about, your baseline rate and your traffic — not from when the result looks significant. A worked example with real numbers, and the sample ratio check to run before any result.5 min read · 1 task inside
- Before you forward that number: a data check for managers, marketers and foundersYou don’t need SQL to catch most wrong numbers. Five habits for product managers, marketers and founders who get figures from dashboards, analysts or AI assistants — and a one-minute check before a number goes into a deck.3 min read · 1 task inside
Becoming an analyst
Where to learn, how interviews test you and what to put in a portfolio when AI writes the SQL.
- Where to learn data analysis in 2026: choose by the skill you needCompare SQL tutorials, broad courses, interview practice and Atterna’s data reasoning exercises. Choose by what you need to practise: writing queries, defining metrics or checking AI-generated analysis.5 min read · 1 task inside
- How to learn data analysis in 2026, when AI writes the SQLAI can write SQL, so what should a beginner learn? A practical path from zero: reading tables, SQL as a language of checking, metric definitions, and judgement — with free resources and honest time estimates.5 min read · 1 task inside
- Data analyst case interview: 8 practice questions and how to answer themThe SQL round tests your queries; the case round tests whether you can take a vague business question to a checked answer. Eight typical case questions — revenue dropped, is the test significant, define active users — with a structure for answering each.5 min read · 1 task inside
- Product analyst case study interview: 5 practice cases with how to answerRetention dropped, conversion fell, a launch “worked”, an A/B test “won”, revenue is down 15%. Five product analyst case study questions with what to clarify, the metric tree, the checks, a strong answer and the weak one interviewers hear most.8 min read · 2 tasks inside
- A data analyst portfolio project in the AI era: show how you checked, not just what you builtAn AI can generate a polished dashboard in an evening, so a pretty project proves little. What makes a portfolio project convincing in 2026 — a real question, a written definition, visible checks, a decision — and a one-page template.4 min read · 1 task inside