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

# Create a Comparison

> Step-by-step guide to configuring and running a comparison between two datasets.

Navigate to **Comparisons** in the left sidebar and click **New Comparison** to open the creation wizard.

## Step 1 — Select comparison type

| Type        | When to use                                                                                   |
| ----------- | --------------------------------------------------------------------------------------------- |
| **Deep**    | When you need a complete diff — all differences identified and reported.                      |
| **Shallow** | When you only need to know if differences exist, not where. Stops at the first mismatch.      |
| **Schema**  | When you only need to compare structure (columns, types, constraints) without examining data. |

## Step 2 — Name the comparison

Enter a descriptive name to identify this comparison in the list.

## Step 3 — Choose data sources

Select the **source** and **target** data sources from your connected integrations.

## Step 4 — Select comparison method

| Method    | Description                                                                                                                          |
| --------- | ------------------------------------------------------------------------------------------------------------------------------------ |
| **Table** | Compare data directly from a database table. Select a dataset and optionally apply a SQL filter clause (e.g. `product.price > 100`). |
| **Query** | Compare datasets using custom SQL queries. Useful when the comparison scope doesn't map to a single table.                           |

## Step 5 — Define primary keys

Select the columns that uniquely identify each row in the source and target. Primary keys are used to align records accurately across both datasets.

## Step 6 — Select columns

Choose which columns to include in the comparison. Columns exclusive to one dataset cannot be selected. Enable **Case Sensitive** to treat uppercase and lowercase values as distinct.

## Step 7 — Map columns (optional)

If source and target columns have different names, define the mapping using **Add Column Mapping Pair**. Select the source column and its corresponding target column for each pair.

## Step 8 — Attach rules (optional)

Rules control how specific differences are handled — for example, ignoring whitespace differences or rounding numeric values before comparison. Click **+ Add Rule** to apply a pre-configured rule.

## Step 9 — Configure semantic similarity (Shallow only)

For Shallow comparisons, you can enable semantic similarity to detect near-matches in text columns:

| Setting                 | Description                                                                  |
| ----------------------- | ---------------------------------------------------------------------------- |
| **Model**               | The text analysis model to use.                                              |
| **Pre-processing**      | Functions applied before comparison — lowercasing, punctuation removal, etc. |
| **Similarity function** | The algorithm used to score similarity (e.g. Levenshtein Distance).          |
| **Match threshold**     | A score between 0 and 1. Values above this are treated as matches.           |

## Step 10 — Advanced configuration (optional)

| Parameter                 | Description                                                                                 |
| ------------------------- | ------------------------------------------------------------------------------------------- |
| **Bisection Threshold**   | Minimum record count above which the dataset is split into smaller segments for processing. |
| **Bisection Factor**      | The division factor used when breaking down large datasets.                                 |
| **Max Threadpool Size**   | Maximum number of parallel threads used during comparison.                                  |
| **Egress Limit**          | Maximum number of differing rows before the comparison automatically stops.                 |
| **Per Column Diff Limit** | Maximum number of differences detected per column.                                          |
| **Timeout Limit**         | Maximum allowed duration in minutes for the comparison job.                                 |

## Step 11 — Submit

Click **Submit** to create and start the comparison. You can monitor progress in [Jobs](/reference/jobs).