> ## Documentation Index
> Fetch the complete documentation index at: https://sourcetable.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Exploratory data analysis

> Automatically profile datasets, analyze distributions, and discover patterns in your data.

Exploratory data analysis (EDA) is usually the first step in any data workflow. Sourcetable automates the tedious parts — profiling, distribution analysis, and correlation — so you can jump straight to insights.

## Automated data profiling

Ask the AI to profile your dataset and it generates a comprehensive summary:

```
"Profile this dataset and give me a summary of all columns"
```

The AI examines every column and reports:

* **Data types** — numeric, categorical, datetime, text, boolean
* **Missing values** — count and percentage per column
* **Unique values** — cardinality for each column
* **Basic statistics** — mean, median, mode, std dev, min, max, quartiles
* **Distribution shape** — skewness and kurtosis for numeric columns

## Distribution analysis

```
"Show the distribution of revenue across all customers"
```

The AI generates histograms, box plots, or violin plots depending on your data. It identifies:

* Normal vs. skewed distributions
* Outliers beyond 1.5x IQR
* Bimodal or multimodal patterns
* Log-normal distributions common in financial data

## Correlation analysis

```
"Run a correlation analysis on all numeric columns and show a heatmap"
```

Sourcetable calculates Pearson, Spearman, or Kendall correlations and renders an interactive heatmap. It highlights:

* Strong positive correlations (> 0.7)
* Strong negative correlations (\< -0.7)
* Multicollinearity between features
* Unexpected relationships

## Automated insights

```
"What are the most interesting patterns in this data?"
```

The AI scans your dataset and surfaces:

* Columns with high missing value rates
* Highly correlated feature pairs
* Categorical columns with imbalanced classes
* Temporal trends and seasonality
* Potential data quality issues (duplicates, inconsistent formats)

## Example prompts

| Goal               | Prompt                                                                                    |
| ------------------ | ----------------------------------------------------------------------------------------- |
| Full profile       | "Profile this dataset — show data types, missing values, and statistics for every column" |
| Compare groups     | "Compare the distribution of salary between departments"                                  |
| Find outliers      | "Identify outliers in the revenue column using IQR and Z-score methods"                   |
| Time patterns      | "Show how monthly sales have trended over the past 2 years"                               |
| Category breakdown | "Break down customer count by region and show percentages"                                |