Segment data with custom properties

Use Custom Properties to segment your data and derive meaningful insights. This feature can help you understand the costs and behavior of different user groups, and gain other insights to help inform strategic decisions and optimizations.

Here are some methods that we recommend for data segmentation:

​

Use Case 1: Tracking Environments

Organizations use Custom Properties to track different environments (i.e. development, staging, and production). To distinguish between these environments, you can create a Helicone-Property-Environment property.

​

Quick Start

1

Add the 'Environment' property and assign a value.

openai.Completion.create(
    # ...
    headers={
        "Helicone-Property-Environment": "development",
    }
)

2

Send a request.

You will then see the Environment property appear in the Requests page.

In this example, we now have a property called 'environment'.

You can choose to hide the custom property by deselecting it under Columns.

You can modify other properties in a similar manner.

3

Filter by custom properties

Filter all requests labelled as 'development'.

4

View metrics associated with a custom property.

Go to the Properties page, and select Environment. You will see metrics associated with this custom property.

View total costs, requests and average latency for `Environment'.

​

Use Case 2: User Segmentation

A common method of data segmentation is by user type. For example, you might want to distinguish between paying and free users to understand their behaviors and costs.

​

Quick Start

To do this, create a user_type custom property and assign either “paid” or “free”.

openai.Completion.create(
    # ...
    headers={
        "Helicone-Property-User-Type": "free",
    }
)

Then, you can filter by paid/free users, or view associated metrics in the same way.

​

Further Segmentation

Suppose you want to understand the behavior of paying users when they use a specific feature (i.e. spellcheck). You can add a Feature custom property.

openai.Completion.create(
    # ...
    headers={
        "Helicone-Property-User-Type": "paid",
        "Helicone-Property-Feature": "spellcheck",
    }
)

You can create highly detailed segment by adding even more custom properties. For example, you may segment users further by plan and Job ID. There are no limits on the number of custom properties you can add.

openai.Completion.create(
    # ...
    headers={
        "Helicone-Property-User-Type": "paid",
        "Helicone-Property-Feature": "spellcheck",
        "Helicone-Property-Plan": "enterprise",
        "Helicone-Property-Job-UUID": "1234-5678-9012-3456",
    }
)

​

Analyzing Segmented Data

Segmented data can provide you with invaluable insights. For example, you might discover that your free users are using the spellcheck feature more than your paid users. This could signal an opportunity to market this feature more aggressively within your premium plans.

​

Use Case 3: Advanced Segmentation

You can refine your segments further by incorporating other properties. The more detailed your segments, the more accurate insights you can derive. Here are some examples:

• Location: Helicone-Property-Location
• Device type: Helicone-Property-Device-Type
• Use activity level: Helicone-Property-Activity-Level

Remember, the key is to select properties that best align with your objectives and that will yield valuable insights upon analysis.

​

Questions?