Understanding the Dashboard
When you log into Voxjar, the main dashboard provides an overview of your quality assurance data. Here are the key components:
Overview: At the top of the dashboard, you'll see a trend line showing how many evaluations have happened on a daily basis. This chart also displays the average score and pass rate over time. You can turn these metrics on or off using the toggles provided.
Scorecard Summaries: Below the overview, you'll find summaries of the scorecards you've created. Each scorecard shows how many evaluations have been done, the average score, and the pass rate. The colors are indicative of performance: green for scores above 75%, yellow for scores between 50-75%, and red for scores below 50%. These thresholds are arbitrary and may be subject to change.
Recent Evaluations: For each scorecard, you can see the most recent five evaluations, color-coded to indicate if they passed or failed.
Customizable Charts: The dashboard includes two main charts - Score Distribution and Evaluator Distribution.
Interpreting Data Visualizations
Score Distribution Chart:
This chart shows the distribution of scores across all evaluations.
It includes several key indicators:
A dotted line on the left represents the bottom quartile (25th percentile)
A blue line in the middle shows the median score
A dashed line on the right indicates the top quartile (75th percentile)
Below the chart, you'll see the average score, the median score, and the scores for the bottom 25% and top 25% of evaluations.
The chart also shows the change in these values compared to the previous period.
Evaluator Distribution Chart:
This chart shows how evaluations are distributed among different evaluators.
It displays the number of evaluations, average score, and pass rate for each evaluator.
This can help you compare AI performance against human evaluators.
Note: The charts may not align perfectly with each other. This is a known issue that hasn't been fully ironed out yet.
Filtering Dashboard Data
You can refine the data shown on the dashboard using various filters:
AI Scores: Filter to show only AI-generated scores.
Language Models: If you're using multiple AI models, you can filter by specific models.
Score Units: Choose between points, percentage, or pass/fail.
Agents: Filter data for specific agents.
Scorecard Titles: Focus on particular scorecards.
Date Range: Adjust the time period for the displayed data.
To apply filters:
Look for the filter options near the top of the dashboard.
Select your desired filters.
The dashboard will automatically update to reflect your selections.
Note: Filters are resource-specific. This means that the filter options available on the dashboard may differ from those in other parts of Voxjar, such as the Interactions page.
Exporting Scorecard Results
To export your data:
Look for an export option, typically represented by a download or export icon.
Clicking this will generate a CSV export.
Important notes about exports:
Exports are grouped by scorecard. You'll receive multiple exports, one for each scorecard involved in your selected data.
Each export will include a reference to the interaction and a link to view it in Voxjar.
You can customize the columns included in your export:
Look for options to choose columns in the export interface.
You can select from standard metadata as well as custom metadata fields.
Your column preferences can be saved for future exports.
Remember, the availability of custom metadata depends on your specific integrations. Some integrations offer an extensive amount of metadata, while others provide just a handful of fields.
The dashboard and reporting features in Voxjar are designed to give you a clear overview of your QA process. By understanding these tools, you can quickly identify trends, spot areas for improvement, and make data-driven decisions to enhance your team's performance.