Ask or search…
⌃K
Links

Getting Started With Dashboards

A Quickstart Guide To Start Using Dashboards

When To Use Dashboards

Dashboard are a powerful tool for teams to easily view, share, and compare key model health metrics. Explore the dashboard widget and templates pages to dive into dashboard use cases for your ad hoc analysis.
​
✅
Check Key Metics Daily
Suggested templates: Top Features, Feature Slice Performance, Feature Drift
​
🔍
Investigate Root Causes
Suggested templates: Regression, Classification, Ranking Performance, Pre-Production Performance
​
📈
Identify Trends
Suggested templates: Compare A/B, Production vs. Training, Drift Monitors

Sharing Dashboards With Your Team

Share dashboards across teams to easily communicate ML model analytics. Simply copy the dashboard URL to share with your team.

How To Create A Dashboard

Dashboards are composed of widgets that you can assemble from scratch or create via pre-existing templates.
Learn how to create a new dashboard below.

Create A Dashboard via UI

Step 1: Pick A Template

Use a template to easily create a new dashboard view. Learn more about templates here.

Step 2: Specify Model Parameters

Fill in your dashboard name, model, and other required fields.

Step 3: Edit & Add Widgets

Edit and add new widgets to customize your dashboard. Learn more about how to use widgets here.

Create A Dashboard via GraphQL

Programmatically create any dashboard using CreateDashboardFromTemplateMutationInput in the GraphQL explorer.
mutation {
createDashboardFromTemplate(input: {
name: "example monitor dashboard",
modelId: "example1id", #add cooresponding model id
modelVersionIds: [],
modelBId: "",
modelBVersionIds: [],
template: driftMonitor, #select template
positiveClass: null,
featureValues: [],
numSlices: 0,
monitorIds: [],
​
}) {
dashboard {
id
monitorLineChartWidgets{
edges{
node{
monitor{
...on WidgetMonitor{
id
name
}
}
}
}
}
}
}
}

Export Data From A Dashboard Widget

Export dashboard data to Phoenix or a notebook for detailed analysis
Questions? Email us at [email protected] or Slack us in the #arize-support channel