Creating and Understanding Custom Targets

Custom Targets allow you to define the exact values, taxonomies, or patterns you want your data to map to, giving you complete control over your classification structure. Once created, custom targets can be added to any data pipeline.

Written By Andrew Sheridan

Last updated 3 months ago

1. Accessing the Custom Targets Menu

Action: Navigate to the Custom Targets creation screen by visiting http://app.mappingclarity.com/custom-targets.

2. Choosing a Target Type

You must select one of three target types, based on the nature of the data you are mapping against.

A. List – Definition (Recommended for Accuracy)

This is the most powerful option as it provides the AI with maximum context.

  • Use Case: Ideal for defined taxonomies, benchmarking categories, or internal account codes where specific, written rules or explanations exist for each value.

  • AI Behavior: By including definitions, you give the AI greater context, enabling it to make smarter and more confident mapping allocations than with a simple list.

  • File Structure for Upload:

    • The file must have headings on Row 1.

    • If you have multiple classification levels (e.g., Level 1, Level 2), repeat the value in Column A for Level 1, and place the specific item in Column B (Level 2).

B. List Only

This type is used when you have a defined, static set of values, but no formal definitions.

  • Use Case: Best for lists where definitions are unnecessary, such as a master list of vendor names, product IDs, or internal codes.

  • AI Behavior: The algorithm will strictly map uploaded data to the values provided in this list.

  • File Structure for Upload:

    • The structure is the same as the List – Definition type (headings on Row 1, multi-level support).

    • Crucially: Definition columns are omitted. Only the list columns (Level 1, Level 2, etc.) are included.

C. Pattern

This type uses unstructured data to help the AI learn how to extract or normalize specific information.

  • Use Case: For situations where you need to extract specific text or normalize data based on contextual patterns, such as extracting a hashtag value (e.g., extracting "something" from a text containing #something.

  • AI Behavior: The AI attempts to understand the general pattern and apply that logic to new, incoming rows.

  • Recommendation: It is highly recommended to upload a sample of previously mapped data (i.e., known inputs and correct outputs) alongside the Pattern type to establish a contextual reference for the AI and significantly improve result accuracy.

3. Creating the Custom Target

Action: After deciding on the type, provide a name for the custom target, select the corresponding type (List–Definition, List Only, or Pattern), and upload your structured file (if applicable).

Once created, your new Custom Target will be available for selection when you create or edit a Data Pipeline.