Using Claude for Data Analysis: From Raw Numbers to Clear Insights

Data analysis used to require either a statistics background or the budget to hire someone who had one. Claude doesn't replace a data scientist for complex analytical work, but it dramatically expands what a non-specialist can do with the data they already have.

Here's how to get genuine analytical value from Claude, even if you don't know your p-values from your quartiles.

The Foundation: How to Share Data with Claude

Pasting CSV Data

Copy your data from Excel, Google Sheets, or any CSV source and paste it directly:

Here is sales data from Q1. Please analyze it:

Date,Product,Region,Units,Revenue
2026-01-03,Widget A,North,42,4200
2026-01-03,Widget B,South,17,2550
[continue rows...]

I want to understand: which product is performing best, 
which region is strongest, and any patterns in the data.

For large datasets, paste a representative sample:

This is a sample of 50 rows from a 10,000-row dataset.
The full dataset has the same columns and spans January 
through April 2026. Please analyze this sample and tell 
me what analyses would be most useful to run on the full dataset.

Describing Data Structure

If your data is too large to paste, describe its structure:

I have a spreadsheet with 8,000 rows and these columns:
- Customer ID (text)
- Purchase date (date)
- Product category (text: Electronics/Clothing/Home)
- Purchase amount (number)
- Customer acquisition channel (text: Organic/Paid/Referral)
- Customer age bracket (text: 18-24/25-34/35-49/50+)

I want to understand: which acquisition channels bring 
the highest-value customers, and whether age bracket 
correlates with product category preference.

What analysis should I run? Give me the exact formulas 
for Excel/Google Sheets.

Core Analysis Tasks

Descriptive Summary

The starting point for any dataset:

Analyze this data and give me:
1. A plain-English summary of what the data shows (3-4 sentences)
2. The most important numbers (top/bottom, highest/lowest, averages)
3. The most surprising or counterintuitive finding
4. What's missing or what questions this data can't answer

[paste data]

Trend Identification

Here is monthly revenue data for 18 months: [paste data]

Identify:
1. The overall trend (growing/declining/flat)
2. Any seasonal patterns
3. Any months that broke the pattern and why that might be
4. The growth rate month-over-month and year-over-year
5. If this trend continues, what would Q4 look like?

Segmentation Analysis

I have customer data segmented by [dimension]. 

[paste data]

Compare the segments on:
- Revenue contribution (absolute and percentage)
- Average transaction value
- Transaction frequency
- Growth trajectory

Which segment deserves more investment? Which is underperforming 
relative to its size?

Finding Anomalies

Review this data and flag anything that looks unusual, 
inconsistent, or potentially incorrect:
- Values that seem too high or too low
- Patterns that break suddenly
- Missing values that should be present
- Internal inconsistencies (e.g., refunds larger than purchases)

[paste data]

This is particularly useful for catching data quality issues before they become decisions based on bad data.

Formula Generation

One of Claude's most practical data uses: generating formulas you'd otherwise spend 30 minutes Googling.

Excel and Google Sheets Formulas

Write an Excel formula that:
- Looks up a customer name in column A
- Finds their most recent purchase date in column C
- Returns the purchase amount from column D for that date

My data starts in row 2. Columns: A=Customer, B=Product, 
C=Date, D=Amount

I need a Google Sheets formula that:
- Counts how many times each product appears in column B
- Only counts rows where column F says "Completed"
- Returns the count next to each product name in a summary table

My Excel sheet has dates in column A. Write a formula 
that groups them by week number and sums the values in 
column C for each week.

Complex Conditional Logic

Write an Excel IF formula that:
- Returns "High" if value in column D is above 1000
- Returns "Medium" if between 500-1000
- Returns "Low" if below 500
- Returns "Error" if the cell is blank or negative

Data Cleaning Formulas

I have a column of phone numbers in inconsistent formats:
(555) 123-4567
555-123-4567
5551234567
+1 555 123 4567

Write an Excel formula that standardizes all of these 
to the format: 555-123-4567

Interpreting Statistical Output

If you run analysis in Excel, Python, or any tool and get output you don't fully understand:

I ran a correlation analysis and got these results:
[paste correlation matrix or statistical output]

Explain:
1. What the numbers mean in plain English
2. Which relationships are strong vs. weak
3. Which correlations, if any, should influence business decisions
4. Any caveats or misinterpretations I should avoid

My survey has 450 responses. I compared two groups 
(users who churned vs. retained) on satisfaction scores.
Here are the results: [paste]

Is this difference statistically meaningful? 
What should I conclude and what can't I conclude from this?

Building Dashboards and Reports

Structuring a Business Review

I'm presenting monthly business results to my management team.
Here is the raw data: [paste]

Help me structure a 5-minute data review that:
1. Leads with the most important number
2. Shows trend vs. last month and last year
3. Identifies one win and one concern
4. Ends with one concrete recommendation

Format it as talking points, not a script.

Writing Data Narratives

Data presented as numbers without narrative often doesn't drive decisions. Claude is good at turning numbers into stories:

Here are our Q1 results: [paste summary data]

Write a 3-paragraph executive summary that:
- Opens with the key headline (what does this quarter mean?)
- Explains what drove performance (not just what happened)
- Addresses what we should do differently in Q2

Tone: Direct and honest. Don't bury bad news.

What Claude Can't Do With Data

Run actual code: Claude can write Python, R, or SQL for you, but it can't execute it. Paste the results back for interpretation.

Create charts: Claude can't generate visual charts. Use Excel, Google Sheets, or Tableau for visuals, then Claude for interpreting them.

Guarantee calculation accuracy: Always verify important calculations independently. For financial reporting or decisions with significant consequences, have a human check the math.

Access your live data: Claude works with data you paste into the conversation. It can't connect to your database, Google Sheets, or analytics platform directly (without API integrations).

The Workflow That Works

The most effective pattern for data work with Claude:

1. Paste data, ask for initial summary → Understand what you have
2. Ask specific questions → Dig into what matters
3. Get formulas for calculations → Run them in your actual tool
4. Paste results back → Get interpretation
5. Ask for the narrative → Turn numbers into decisions

This workflow works whether you're a marketing analyst looking at campaign performance, an operations manager reviewing efficiency metrics, or a small business owner trying to understand your own numbers.

The goal isn't to make Claude your data scientist — it's to eliminate the "I have this data but I don't know what to do with it" problem that leaves valuable information unused.