Product Category Field
Overview
This document provides information about the custom Zendesk field “Product Category”, and how to populate it. This is an expected part of the Closing phase of ticket handling.
What is the Product Category field?
The Product Category field is a multi-select field in Zendesk that helps us identify which GitLab product areas are at the heart of each support ticket. This field is not yet required to be populated before setting any ticket to Solved status, however it is a highly beneficial field for providing customer insights to our product and development teams.
The data collected through this field enables:
- Analytics on which GitLab features generate the most support volume
- Trend identification for product team feedback
- Improved knowledge management and documentation prioritization
The options in the field have been populated from the categories.yaml file.
Using the Product Category field
When to populate the field
The Product Category field needs to be populated as part of the ticket closure process, though it can be updated at any time prior to the ticket closing. Setting the ticket to “Solved” will often be the most appropriate time to set this, as we tend to gain information during the course of solving a ticket that helps us better understand what was the primary cause of the problem the customer experienced.
How to populate the field
You can populate the Product Category field in two ways:
-
Manual selection: If you clearly know which category applies to the ticket, select the appropriate options directly from the multi-select field. More than one category can apply to a single ticket.
-
ZenDuo assistance: Use the “Product Category” ZenDuo prompt to help identify the most relevant categories based on the ticket content.
Using the ZenDuo “Product Category” prompt
What is the ZenDuo Product Category prompt?
The ZenDuo “Product Category” prompt is a specialized prompt designed to analyze ticket content and suggest appropriate Product Category selections. This prompt was developed through extensive iteration to provide consistent and helpful categorization suggestions. As an LLM tool, it won’t always get it right, but generally it gets us close enough to be helpful.
How to use the ZenDuo prompt
NOTE: The prompt takes some time to process - it will send at least 4 chunks of data to Duo Chat, and depending on the ticket complexity, it could be significantly more. It is recommended to set it running while you work on other tickets in different tabs.
To use the ZenDuo Product Category prompt:
- Navigate to the ticket you want to categorize
- Open ZenDuo App
- Select the “Product Category” prompt and wait for the suggested categories.
- Use the suggestions to populate the Product Category field. (This is a manual step for now.)
Expected behavior and limitations
- The prompt analyzes the entire ticket conversation to identify relevant product areas
- Results are generally consistent, but may occasionally provide unhelpful suggestions
- Processing time varies based on ticket length and complexity
- Manual review of suggestions is always recommended before field completion
When the prompt doesn’t work well
If you find the ZenDuo prompt consistently providing unhelpful results for certain types of tickets, please provide feedback through the feedback issue. This helps improve the prompt’s effectiveness over time.
Troubleshooting
Common issues
- ZenDuo prompt takes too long: This is expected behavior for complex tickets. Consider starting the prompt early in your ticket work or while handling other tasks.
- Prompt suggests irrelevant categories: Manual review and selection is always recommended. Use your judgment to select the most appropriate categories. Provide feedback in the feedback issue as the prompt may need tweaking. You can also use the Chat feature of ZenDuo and add a follow up prompt of “those categories are not in the filtered-categories.yml file”
Getting help
If you encounter issues with the Product Category field or ZenDuo prompt:
- Provide feedback (feedback issue) on prompt effectiveness to help improve future iterations
References
- Roadmap issue where this was introduced
971545fe)
