ChatGPT for Data Analysts: Prompts That Speed Up Every Stage of Analysis
I work with data every day. SQL queries, Python scripts, stakeholder decks, ad-hoc requests from every direction. The part that used to grind me down wasn't the analysis itself — it was the surrounding work. Writing query boilerplate, translating findings into plain English, turning a table of numbers into a narrative someone in marketing actually cares about.
ChatGPT changed that. Not the core analysis — I still do that — but everything wrapped around it.
Here are the prompts I use regularly.
SQL Query Generation
When I need a query structure I haven't written in a while, I draft it with ChatGPT and then audit it myself. I'm not trusting it blind — I'm using it to skip the blank-page problem.
Prompt:
"Write a SQL query that identifies customers who made at least 3 purchases in the last 90 days but haven't purchased in the last 30 days. Use a 'transactions' table with columns: customer_id, purchase_date, amount. Return customer_id and their last purchase date, sorted by last purchase date ascending."
I get a solid working draft, check the logic, adjust for my schema. Saves 10-15 minutes on standard queries.
Debugging and Explaining Queries
When a query returns unexpected results or a colleague sends me something that doesn't make sense, I paste it in and ask ChatGPT to explain it line by line.
Prompt:
"Explain what this SQL query does, step by step, in plain English. Then identify any potential issues — like performance problems, incorrect joins, or edge cases where the output might be wrong. [paste query here]"
Great for onboarding, great for debugging, great for explaining what inherited code is actually doing.
Data Interpretation Summaries
The hardest part of analysis isn't running the numbers — it's writing the summary that makes the numbers mean something to someone who didn't build the model. ChatGPT is surprisingly good at this if you give it the right framing.
Prompt:
"I ran an analysis comparing customer retention rates across 4 cohorts from Q1 through Q4. Here are the results: [paste table]. Write a 3-paragraph summary for a marketing director. Start with the most important finding, explain what's driving it, and suggest one concrete action they should take."
The output is always a draft — I fact-check every claim — but the structure and tone come out right.
Stakeholder Presentation Outlines
When I need to present analysis to leadership, I use ChatGPT to structure the deck before I open PowerPoint.
Prompt:
"I need to present the results of a pricing analysis to a VP and CFO. The key finding is that a 10% price increase on our mid-tier product would generate an estimated $2.1M in additional annual revenue with minimal churn risk based on our elasticity model. Create a 6-slide deck outline. Include slide titles, the main point of each slide, and what chart or data visualization would support it."
Six slides with clear story arc, in 2 minutes. Then I fill it in.
Hypothesis Generation
When I'm starting a new analysis, I use ChatGPT to pressure-test my hypotheses before I go looking for data to confirm them.
Prompt:
"I'm analyzing why our mobile conversion rate dropped 18% over the past 6 weeks. Here's what I know: the drop started after a product release in week 2, it's worse on Android than iOS, and cart abandonment went up while traffic stayed flat. What are 5 plausible hypotheses I should investigate? Rank them by likelihood and explain your reasoning."
It surfaces angles I'd overlook when I'm already anchored to one explanation.
Writing Python Analysis Boilerplate
For repetitive setup — reading CSVs, cleaning data, building basic visualizations — I draft the boilerplate fast.
Prompt:
"Write a Python script using pandas that reads a CSV file, removes duplicate rows, fills missing values in numeric columns with the column median, and generates a correlation matrix heatmap using seaborn. Include comments. Assume the file path is passed as a command-line argument."
I review every line, but I'm not writing it from scratch at 5pm on a Friday.
Plain-English Metric Definitions
When someone asks "what does this metric actually mean?" and I need to answer in a Slack message without writing a wall of text, I draft it here first.
Prompt:
"Write a plain-English definition of 'customer lifetime value' for a non-technical marketing manager. Explain what it measures, why it matters, and what a company should do when it goes down. Keep it under 150 words and avoid jargon."
Crisp, accurate, usable.
Ad-Hoc Request Triage
When I get slammed with ad-hoc requests and need to push back with something more than "I'm busy," I use ChatGPT to help me frame a scoping response.
Prompt:
"I received a request from a product manager asking for 'a full analysis of how users engage with Feature X.' I need to scope this into 3 concrete questions I can actually answer, estimate the data requirements and time for each, and set expectations about what I can deliver in 3 days vs. 2 weeks. Help me draft that response."
It helps me think through prioritization out loud and write a professional response instead of a frustrated one.
These prompts don't do the analysis for me. They handle the surrounding work — setup, translation, communication — so I can spend more time on the parts that actually require judgment.
If you want a full library of prompts organized by role and task — 100+ prompts covering analysis, communication, SQL, Python, and more — I put together The ChatGPT Prompt Playbook for $27. One-time purchase, immediate download.
Spend your time on analysis. Let ChatGPT handle the rest.
Top comments (0)