Independently researched from published sources. Last researched: April 2026. Results vary: this article teaches AI skills, not employment outcomes. See Terms and Privacy.
Most people type 'write me a SQL query' into ChatGPT and get back something that technically runs but misses half the edge cases. The fix isn't a better model. It's a better prompt. One that tells the model your database engine, your schema, and exactly what shape the output should take.
The six prompts below come from independently researched phrasing built for the work Data Analysts actually do every week: writing queries, investigating anomalies, summarizing findings for people who won't read past the first paragraph. Each prompt has placeholders in brackets. Replace those with your real details, paste the whole thing in, and you'll get a usable first draft instead of a generic one.
Pick the prompt that matches whatever is on your plate right now. You can have a solid first draft of a SQL query or an executive summary in the next ten minutes.
Translating a stakeholder's plain-English question into clean SQL is probably the most repeated task in data analysis. This prompt forces the model to use CTEs for readability, handle NULLs, optimize for your specific database engine, and output business-friendly column names. That's the difference between a query you need to rewrite and one you can actually run.
How to use it: Replace the database engine, paste your table schema (column names, types, join relationships), and drop in the stakeholder's question verbatim. Review the query's NULL handling and stated assumptions before running it against production.
Stakeholders rarely ask the precise analytical question they need answered. 'Why did revenue drop?' could mean five different analyses depending on context. This prompt makes you structure your thinking before touching SQL: restate the question precisely, list hypotheses, identify the right segmentation, plan the deliverable format. It catches scope creep before you've spent three hours on the wrong cut of data.
How to use it: Paste the stakeholder's exact words, their role, and any context about why they're asking. Use the output as your analysis plan before writing a single query.
You finished the analysis. Now you need to communicate it to someone who will spend 90 seconds reading. This prompt structures raw findings into a bottom-line-first format with specific numbers, recommended actions, and risks. It forces the 'so what' to the top, which is exactly where most analyst write-ups fall apart.
How to use it: Paste your raw analysis bullets, numbers, and observations. Specify your audience (C-suite, VP, department head). Edit the output for accuracy, since the model may reframe your numbers in ways that shift meaning.
Finding these useful? The full guide has 17 of them, plus tool reviews and a 30-day plan. Get it for $29.
A metric just spiked or dropped and your Slack is blowing up. This prompt generates a structured investigation checklist, diagnostic SQL queries for your specific schema, and a stakeholder communication draft under 100 words. It starts with data issues before business causes, which is the right order: anomalies often turn out to be pipeline problems rather than market shifts.
How to use it: Fill in the metric name, its normal range, the anomalous value, when it started, and your database schema. Run the diagnostic queries it generates in order, crossing off causes as you go.
You built a chart. Someone in a meeting has 30 seconds to absorb it. This prompt generates a title that communicates the insight, not just what the axes show, plus a plain-English description for non-technical audiences and a prebuilt counterargument for the skeptic in the room. It also suggests the next analysis, which saves you the follow-up question.
How to use it: Describe your chart type, what it visualizes, and paste the underlying data or key data points. Use the generated title and description directly in your slide deck or report.
Most A/B test summaries either over-claim significance or bury an inconclusive result in optimistic language. This prompt asks for sample sizes, duration, and metric values, then checks statistical significance, practical effect size, and common problems like sample ratio mismatch. It ends with a plain-English summary you can send to a product manager who doesn't speak statistics.
How to use it: Fill in the hypothesis, control and variant descriptions, sample sizes, test duration, and metric values for each group. Verify the statistical calculations independently before making the final call.
The full Data Analyst guide goes much further: 17 copy-paste prompts, honest reviews of 13 tools with current prices, a dos and don'ts chapter, and a 30-day plan to put it all into practice.
Get the full guide. $29One-time purchase. Instant download. Or read more about what's inside.