classify

business analytics

slug: step-classify

Purpose

Assigns categorical labels to dataset rows based on user-defined conditional rules. Useful for segmenting customers, scoring leads, or creating flags based on business logic.

When to Use

• Segment customers based on RFM (Recency, Frequency, Monetary) or other scoring systems
• Flag or categorize data based on complex business rules
• Simplify downstream analysis by creating a categorical field from numeric or textual data

How It Works

1. Receives a dataset (data) and applies a series of conditional rules to each row.

2. Each rule contains a when condition and a then value:

   • The first rule whose condition evaluates to true is applied.
   • If no rule matches and an else clause is defined, that value is used.
   • If no match and no else, the output value is set to null.
3. The evaluated value is assigned to a new or existing column (outputColumn).

Important: Conditions are evaluated sequentially, and the first match wins.

Parameters

Required

• outputColumn (string) – Name of the column to store the classification result.

• rules (array) – Ordered list of rules with the following structure:

  - when: "condition_expression"
  then: "label_value"
  - else: "default_label"

  • when expressions are evaluated per row.
  • else is optional and provides a default classification.

Input Requirements

• Any dataset containing the fields referenced in the when conditions
• Conditions can use logical operators (AND, OR, ==, >=, etc.)

Output

• data: Original dataset with an additional column (outputColumn) containing the classification
• pdv: Updated PDV metadata (adds outputColumn if new)
• extras: Passed through unchanged
• outputType: 'work'

Example Usage

steps:
  - classify:
      source: rfm_scores
      outputColumn: segment
      rules:
        - when: "rScore == 5 AND fScore == 5 AND mScore == 5"
          then: "Champions"
        - when: "rScore >= 4 AND fScore >= 4 AND mScore >= 4"
          then: "Loyal Customers"
        - when: "rScore <= 2 AND fScore <= 2"
          then: "At Risk"
        - else: "Other"

Explanation:

• Rows with perfect RFM scores are labeled as “Champions.”
• Rows with generally high scores are labeled “Loyal Customers.”
• Low-scoring rows are labeled “At Risk.”
• All other rows default to “Other.”

Notes & Best Practices

• Ensure all fields referenced in when conditions exist in the dataset.
• Logical operators should match the DSL’s evaluation syntax.
• Order matters: the first matching rule is applied; place more specific rules first.
• Combine with prior calculation or transformation steps to generate derived fields for classification.
• Consider providing an else clause to avoid null classifications.