Scorecard Fundamentals
Scorecards are the foundation for evaluation in Voxjar. They define the rating criteria used for both human and AI evaluations. Scorecards allow you to create custom questions and scoring systems tailored to your specific quality assurance needs.
Creating Scorecards
There are three ways to create a scorecard in Voxjar:
Starting from scratch
Using templates
Customizing existing scorecards
To access the scorecard creation options:
Navigate to the Scorecards section in the main menu.
Look for options to create a new scorecard or use a template.
Templates are always available and can be a good starting point, especially if you're new to creating scorecards.
Scorecard Builder Guide
The Scorecard Builder is where you'll create and edit your scorecards. Here's an overview of its components:
Scorecard Header:
Name: Give your scorecard a descriptive title.
Description: Explain the purpose of the scorecard.
AI Model: Choose which AI model to use (e.g., GPT-4, Llama, Voxjar Instruct model, etc).
Scorecard Prompt: This acts as a system prompt for the AI. It's recommended to keep this light and focus on writing good questions.
Passing Score: Enable or disable a passing score and set the threshold.
Sections:
You can create sections to organize your questions.
Sections are primarily for visual organization and don't affect the AI evaluation.
Questions: There are three types of questions:
Single Selection: Choose one answer from multiple options.
Sliders: Set a value along a continuous range.
Multiple Choice: Select one or more answers from multiple options.
For each question:
Write your question clearly.
Set up answer options.
Assign point values to each answer.
Optionally add explanations or examples to guide the AI.
You can add as many questions as needed. Each question can have its own point value, allowing you to weigh different aspects of the evaluation as desired.
Testing Scorecards with AI
You can test both individual questions and entire scorecards with AI:
Look for a "Test with AI" option in the Scorecard Builder.
You'll need to provide a sample call. You can either:
Upload an audio file, which will be automatically transcribed.
Choose from your existing calls (limited to the last 7 days).
Run the test. This typically takes 30 seconds to 4 minutes, depending on the scorecard size.
Review the AI's responses. For each question, you'll see the AI's answer and explanation.
This testing feature allows you to verify that your questions are clear and that the AI understands and can accurately evaluate based on your criteria.
Learn more about testing your scorecards with AI here
Managing Scorecards
Best practices for managing scorecards:
Use the duplicate function to create new versions of scorecards.
Avoid editing scorecards that are actively being used for evaluations.
If you need to make changes, create a new version and use that going forward.
You can delete scorecards, but be aware that this will delete all scores associated with that scorecard.
Scorecard Reporting
To view scorecard results:
Navigate to the Scorecards section.
Select the scorecard you want to review.
Look for a "View Responses" or similar option.
In the results view:
You can filter by date range, AI or human evaluator, and other criteria.
Each question is displayed with a distribution of responses.
You can drill down into individual questions to see all responses.
For each response, you can see the evaluator, score, and any notes provided.
This reporting feature allows you to quickly identify trends, problem areas, and high-performing aspects of your customer interactions.
Scorecards are a powerful tool in Voxjar, allowing you to customize your quality assurance process to your specific needs. By carefully crafting your scorecards and regularly reviewing the results, you can gain deep insights into your customer interactions and drive continuous improvement in your team's performance.
A more detailed guide can be found here