Data Analysis

Last verified: 13 February 2026 | Applies to: All plans (basic in Chat; full power with Data plugin in Cowork)

In 30 seconds

Claude can analyse data from CSVs, spreadsheets, databases, and connected tools — producing summaries, visualisations, and interactive dashboards without requiring you to write SQL or code. In Chat, upload a file and ask questions. In Cowork with the Data plugin, Claude builds full dashboards and runs statistical analysis.

Prerequisites

• For basic analysis (Chat): Enable “Code execution and file creation” in Settings → Capabilities. Upload your data file.
• For advanced analysis (Cowork): Install the Data plugin from claude.com/plugins or the Cowork interface.
• For connected data: Set up connectors for your data sources.

Two approaches

Chat: Quick analysis

Upload a CSV, Excel file, or paste data directly. Then ask questions in natural language.

Explore your data:

Here's our Q1 sales data [upload file]. Give me a summary: total revenue, average order value, top 5 products by revenue, and any trends you notice.

Ask specific questions:

Which region had the highest growth rate quarter over quarter? Which had the steepest decline?

Create visualisations:

Create a bar chart showing monthly revenue by region. Include a trend line.

Claude generates charts as artifacts you can view inline and download.

Compare datasets:

Here are two files — last year's Q1 and this year's Q1 [upload both]. Compare them: what improved, what declined, and what stayed flat?

Cowork + Data plugin: Full power

The Data plugin adds SQL support, interactive dashboards, and statistical analysis.

Query a database:

Connect to our PostgreSQL database and show me the top 10 customers by lifetime value who haven't placed an order in the last 90 days.

Claude writes and executes the SQL, then presents results in a readable format.

Build a dashboard:

Create an interactive dashboard from the data in sales-q1.csv. Include: revenue over time (line chart), revenue by region (bar chart), product breakdown (pie chart), and a summary metrics row at the top.

Claude generates an HTML dashboard you can open in your browser.

Statistical analysis:

Run a correlation analysis between marketing spend and revenue across all regions. Is the relationship statistically significant?

Common analysis patterns

The morning briefing

Read the latest data from [connected tool/uploaded file]. What are the three most important things I should know today?

The anomaly finder

Analyse this data for anything unusual. Flag any metrics that deviate more than 15% from the 30-day average.

The comparison report

Compare this month's performance to the same month last year. Highlight improvements, declines, and anything that changed by more than 10%.

The forecast

Based on the last 12 months of data, project the next 3 months. Show the projected values with confidence ranges. Note any assumptions.

Caveat: Claude’s projections are statistical extrapolations, not business forecasts. Use them as starting points, not decisions.

How operators actually use it

What operators say

“I upload our weekly sales CSV every Monday and ask Claude for the top 3 insights. It usually catches things I wouldn’t have noticed — like a product category that’s growing 40% but only accounts for 2% of revenue.”

“The Data plugin’s dashboard feature is genuinely useful. I created an interactive sales dashboard that my team now uses as our primary view. Took about 20 minutes vs. the two weeks it would have taken to build in a BI tool.”

What to watch out for

• Verify calculations. Claude can make arithmetic errors, especially with percentages, weighted averages, and complex formulas. Spot-check any numbers you’ll share with stakeholders.
• Large datasets have limits. Claude’s context window limits how much data it can process at once. For very large datasets (100,000+ rows), filter or sample before uploading.
• Statistical claims need scrutiny. When Claude says something is “statistically significant,” verify the methodology. Claude can overstate confidence in patterns.
• Dashboards are static exports. The HTML dashboards Claude creates don’t update automatically. They’re snapshots you need to regenerate when data changes.
• The Data plugin handles SQL across dialects. PostgreSQL, MySQL, SQLite, BigQuery — specify which dialect if Claude gets the syntax wrong.

Related

• Document Creation — turn analysis into reports and presentations
• Data Plugin — full plugin reference
• Connecting Your Tools — connect data sources directly
• Prompting for Operators — better prompts produce better analysis

---

Something wrong or outdated? Let us know →

Get weekly workflows — subscribe to the newsletter.