Merchant-Level Categorisation

How It Works

• Extraction: Merchant information and transaction details are extracted from the receipt or invoice.
• Analysis: The extracted data is analysed using Taggun Machine Learning algorithms.
• Categorisation: An expense category is assigned to the transaction (from a finite list of categories).
• Confidence Scoring: A confidence level is provided for the assigned category to indicate how close it is to that category.

Set-Up Process

1. Contact Taggun

Reach out to [email protected] to enable this feature for your account.

2. Define your categories

Define your categories in a CSV file with the template provided.

Downloadable CSV Template:

categories-example.csv

CSV Template Structure:

• categoryName: The category identifier returned in API responses
• description: (optional) Clear natural language explanation of what belongs in this category.
  • This helps with specification and handling ambiguity.
  • Only use descriptionwhen necessary as it may increase response time.

3. Upload Your Categories

• Upload your categories using the API Endpoint: Upload Categories
• Review your current category mappings via the API Endpoint: Export Categories

4. Make an API call

Make an API call to one of Taggun's Verbose Data Extraction API Endpoints

• Upload File (verbose)
• Encoded File (verbose)
• URL (verbose)

Access the assigned category in the response in: entities.category.data

No other changes are required. This feature automatically returns data once it's enabled by Taggun.

Example

Image

Response

🔐

Access Path

Access the category via the path entities.category.data

A snippet from the JSON Payload:

{
  "entities": {
    "category": {
      "data": "food & drink",
      "confidenceLevel": 0.81
    }
  }
}

Use Cases

• Corporate expense management
• Personal finance tracking
• Market analysis and consumer behaviour insights

Best Practices

• Use the confidence level to determine when manual review might be necessary.
• Combine with Product-Level Categorisation for more detailed insights.

Customisation

The transaction categorisation feature is powerful, but it may need fine-tuning to be truly effective for your unique setup. If you are getting unreliable results, edit the description for the individual categories, or reach out to us at [email protected] for further support.