AI skills test for data analysts, on real data. A good analyst with AI lets it write the queries and keeps the judgment: what the number means, whether it is right, and what the people asking need to hear.
Try it free on 3 candidates.
Who the data analyst AI skills test is for Data analysts Analytics engineers Product data scientists People who answer business questions with data. They pin down what a number means, run the queries, and explain the answer to someone who is not technical.
Level Junior to staff
Hiring ML engineers, data engineers or research scientists? Bring your own task
An example AI skills task for data analysts One of our data tasks. Your dashboard has the rest, and you can bring your own.
LemonsignalYour task Recording DeadlineFri 24 Oct The brief Explain a revenue drop The dashboard shows revenue down 12% in March. Explain why to the CFO in five lines.
Then do Q2 and Q3.
Files Time Work in Start on my computer Start in the browser
Where AI goes wrong Every task has a few spots where AI typically gets it wrong. That’s how you see who checks the work. The part that repeats After March, they explain Q2 and Q3. You see whether they reuse what they built the first time, or start again by hand. Time You set it: a deadline, a time limit, or both. We recommend a deadline and no time limit, so they work the way they would in the job. Where they work Claude Code, Codex, Claude Cowork or ChatGPT, on their own computer or in the browser. All tasks The AI skills we look for in data analysts Understand Checks how the dashboard defines revenue before asking for an answer. Brief Gives the AI the tables and the definition, and tells it to use only these numbers. Check Checks the total against a second source, and a second month to rule out a one-off. Correct Says what is wrong with a query, and adds a check so it cannot happen again. Sign off Five lines, as asked, with the assumptions stated and what is still unverified. Reuse Does Q2 and Q3 the same way, and hands in the query or prompt they built. Plus the skills for the job Read from the same task.
Defining the metric Rigour in the analysis Explaining it to people who aren’t technical What you get for each data candidate LemonsignalAnton Lindqvist Result His work
44% Below the bar
Took the AI’s first query without checking it. Sent the CFO a number he had not checked. Built a script that made Q2 and Q3 fast and the same. Understand Asked what “revenue” means here, but only after the first answer. Turn 9 Brief Gave it the tables, but not the definition. Turn 1 Check Took the first query as it was, and never matched the total to a second source. Turn 4 Correct When a query failed, said which join was wrong. Turn 6 Sign off Five lines, as asked, built on a number he had not checked. Reuse Saved the query as a script and ran Q2 and Q3 with it. Skills for the job Defining the metric Below the bar Rigour in the analysis Below the bar Explaining it to people who aren’t technical At the bar The main gap Checking the AI’s number against a second source before sending it.
Ask him in the next interview: “Your note says revenue fell 12%. How would you check that number before the CFO reads it?”
See his work The five lines for the CFO and his script’s numbers for Q1 to Q3 What he handed in, as he handed it in. Each mark is a skill from his result.
Weak Partial To the CFO March revenue fell 12% Because of the price change on 1 March.
What happened: revenue fell from €1.24m in February to €1.09m in March.The evidence: orders held up (+2%), so each order is worth less.Where: annual plans fell most.Next: Q2 and Q3, run the same way. Show all Strong revenue_by_month.py, run for Q1 to Q3
Show all 9 rows Questions about testing data analysts’ AI skills Do analysts need SQL for this task? It helps, but they can do it in any of the four tools. Claude Code and Codex can run the queries; Claude Cowork and ChatGPT can read the files.
Why does a strong Reuse not make this a good match? Each role has must-haves. For an analyst, Check is one: a fast, repeatable answer that is wrong is worse than a slow one. You can change the must-haves before you send the link.
Can we use our own data? Yes. Bring your own task. Confidential company files stay with your hiring team, the candidate and Lemonsignal.
Claude Code
Codex
Claude Cowork
ChatGPT