←back to features

Data Extraction

Turn unstructured conversations into structured data and metrics using a JSON schema, ready for analytics and other systems.

ChatBotKit's Extract integration bridges the gap between unstructured conversation data and structured databases. It activates when a conversation becomes idle, uses a JSON schema you provide to pull specific information from the dialogue, and stores the result in the conversation metadata, accessible through the ChatBotKit API or dashboard.

This turns everyday customer interactions into structured data you can use to populate databases, feed other systems, and build analytics around what your users actually need.

What You Can Do

  • Define a schema: Specify the fields to capture with a JSON schema.
  • Extract automatically: When a conversation goes idle, the integration extracts the defined data.
  • Store and access it: Results are written to conversation metadata and available via the API or dashboard.
  • Collect numeric metrics: Mark numeric fields with collect: true to track them as dedicated metrics for reporting.
  • Feed other systems: Use extracted data to populate databases and downstream tools.

How It Works

Each Extract integration watches conversations and runs when they become idle. It applies your JSON schema to the conversation, extracts matching values, and stores them in the conversation's metadata. When a field is flagged for metric collection, its numeric value is captured as a scoped metric - order amounts, quantities, ratings, and similar values become organized data points. Metric collection, metadata updates, and webhook notifications run in parallel, so extraction keeps up even during high conversation volume.

Setup

Create an Extract integration in ChatBotKit, define your extraction schema, and add collect: true to any numeric fields you want tracked as metrics. The integration then runs automatically as conversations go idle.

Practical Uses

Data extraction is valuable for analytics and operations: e-commerce teams track order values and quantities, service teams monitor satisfaction ratings, and subscription businesses analyze usage - all captured automatically from natural conversations. The result is business intelligence drawn directly from customer interactions.

data extractionextractstructured dataanalyticsmetricsintegrationsconversations