Module 13 · Lesson 03

Data Analysis with Claude

Reading time: 15 minutes Track: Claude Fluency for Teams · Knowledge worker path

What Claude is good at in data analysis

Claude's data capabilities are best described as: excellent at reasoning about data, limited at computing with data.

What this means in practice:

Claude is excellent at:

• Writing analysis code (Python, SQL, R) you then run
• Interpreting results and explaining what they mean
• Designing an analysis approach for a question
• Spotting methodological issues
• Translating findings into plain language

Claude is limited at:

• Performing arithmetic on large datasets directly (it will make errors)
• Running code on your actual data (unless given a code execution tool)
• Accessing your databases or BI tools directly (unless connected via MCP)

Use Claude as your analysis strategist and interpreter; use your actual tools for computation.

The analysis workflow

Step 1: Design before computing

I want to understand whether our customer churn is seasonal.
I have a dataset with columns: customer_id, subscription_start, subscription_end (null if active), plan_type.
What analysis would answer this question? What are the key metrics to compute and what would I look for?

Claude will describe the approach. You confirm it makes sense before writing any code.

Step 2: Write the code

Write Python code using pandas to:
[description of analysis from Step 1]
The dataframe is called df with these columns: [columns]
Include comments explaining each step. Return results as a summary dictionary.

Run this code on your actual data.

Step 3: Interpret results

Here are the results of the churn seasonality analysis:
[paste results]

Interpret these results. What do they tell us about seasonal patterns?
What's the strength of evidence? What caveats should I note?
What follow-up analysis would strengthen or challenge this conclusion?

SQL analysis patterns

For SQL analysis questions:

Here is my database schema: [paste schema]
Question: [your business question]
Write a SQL query to answer this question. Use [PostgreSQL/MySQL/BigQuery] syntax.
Include a brief explanation of what the query does.

Then run the query, paste results back, and ask Claude to interpret them.

Data storytelling

One of Claude's clearest wins: turning results into communication.

Here are the key metrics from our Q2 analysis:
[paste results]

I need to present this to our non-technical leadership team in 5 minutes.
Write a 3-sentence verbal summary of the most important finding and what it means for the business.

The chart shows [describe chart and data].
Write a 2-sentence caption for this chart for an executive audience.
Highlight the key insight, not just what the chart shows.

A note on numerical outputs

Never use Claude's direct numerical outputs in a deliverable. If Claude says "the churn rate is 8.3%," verify that against your actual data. Claude computes arithmetic through its language model, not a calculator, and makes errors — especially on multi-step calculations. Use Claude for code and interpretation; use your actual data tools for numbers.