The classification object is a powerful, flexible structure that allows you to add layers of categorical and contextual understanding to any semantic event. Think of it as adding smart tags to your data. An event isn't just a single occurrence; it can represent a customer's intent, fall into a specific category, or be assigned a priority. A single event can have multiple classifications, allowing for a rich, multi-faceted description.
This classification can be assigned when the event is created or, more powerfully, it can be added automatically by downstream systems like the Context Suite. Using Natural Language Processing (NLP) and Large Language Models (LLMs), the system can analyze the content of an event and enrich it with classifications, turning raw data into structured, actionable intelligence without any manual effort.
Properties
The following properties make up a single classification object. An event can contain a list of these objects.
| Property | Description | Provided By |
|---|---|---|
type | The type of classification being applied (e.g., 'Intent', 'Tag', 'Priority'). | User-Provided |
value | The specific value for the classification type (e.g., 'Purchase', 'Customer Support', 'High'). | User-Provided |
reasoning | (Advanced) An explanation for why this classification was assigned, often provided by an AI model. | User-Provided (Optional) |
score | (Advanced) A numerical score associated with the classification. | User-Provided (Optional) |
confidence | (Advanced) The confidence level (typically 0.0 to 1.0) from a model that assigned the classification. | User-Provided (Optional) |
weight | (Advanced) The relevance or importance of this specific classification relative to others on the event. | User-Provided (Optional) |
Common Classification Types
While you can use any custom classification type that makes sense for your business, the Context Suite has a set of standardized types that enable powerful, pre-built analytics.
For Business Analysts: These types help in understanding customer behavior and business funnels.
Intent: Describes the underlying goal of the user (e.g., 'Make Purchase', 'Request Refund', 'Get Information').Intent Category: A broader grouping for intents (e.g., 'Sales', 'Support').Priority: Assigns an importance level to the event (e.g., 'High', 'Medium', 'Low').Segmentation: Places the event or user into a specific business segment (e.g., 'High-Value Customers', 'Churn Risk').Category/Subcategory: Standard business-defined categories for products, services, or issues.
For Data Engineers & Developers: These types are often useful for technical routing, filtering, and process automation.
Tag: A general-purpose tag, similar to labels in other systems.Keyword: Extracts specific keywords from event content.Inbox: Used to route an event to a specific team or processing queue (e.g., 'SupportInbox', 'SalesLeadInbox').
Example: Automatic Classification
Imagine a customer submits a generic feedback form on your website. You send a track event with their raw message in the content property.
1. Initial Event Sent by User
A simple jitsu.track call captures the customer's feedback.
1jitsu.track('Feedback Submitted', {
2 content: {
3 body: "Hi, my order just arrived but the screen is cracked. I'm really unhappy about this and need to get a replacement sent out immediately. My order number is #A-123."
4 },
5 properties: {
6 source: 'Website Feedback Form'
7 }
8});2. The Enriched Event (After Automatic Classification)
The Context Suite receives this event. Its AI engine analyzes the content.body and automatically adds a classification array to the event, transforming it into a structured, intelligent record.
1{
2 "event": "Feedback Submitted",
3 "type": "track",
4 "content": {
5 "body": "Hi, my order just arrived but the screen is cracked. I'm really unhappy about this and need to get a replacement sent out immediately. My order number is #A-123."
6 },
7 "properties": {
8 "source": "Website Feedback Form"
9 },
10 "classification": [
11 {
12 "type": "Intent",
13 "value": "Request Replacement",
14 "confidence": 0.95
15 },
16 {
17 "type": "Intent Category",
18 "value": "Support",
19 "confidence": 0.98
20 },
21 {
22 "type": "Category",
23 "value": "Damaged Goods",
24 "confidence": 0.99
25 },
26 {
27 "type": "Priority",
28 "value": "High",
29 "reasoning": "The user expressed unhappiness and requested an immediate replacement.",
30 "confidence": 0.89
31 },
32 {
33 "type": "Inbox",
34 "value": "Returns"
35 }
36 ]
37 // ...other event properties
38}As you can see, the single, unstructured feedback message has been automatically categorized, prioritized, and routed, making it immediately actionable for your support and logistics teams.