AI agent skill · by @serac-labs
Performance Analytics
Build ServiceNow Performance Analytics — pa_indicators (count/sum/avg/percentage), pa_breakdowns, pa_thresholds with severity colors, pa_widgets, and pa_dashboards for KPI tracking.
Archived 2026-10-23 · Source updated 2026-09-10 · License: Apache-2.0. Check the original repository for the latest version.
SKILL.md instructions
# Performance Analytics for ServiceNow
Performance Analytics (PA) provides advanced reporting and KPI tracking capabilities for measuring and improving business processes.
## PA Architecture
### Component Hierarchy
```
PA Dashboard
├── Widgets
│ ├── Scorecard Widget
│ ├── Time Series Chart
│ ├── Breakdown Pie Chart
│ └── Single Score
├── Indicators
│ ├── Number of Open Incidents
│ ├── Average Resolution Time
│ └── SLA Compliance Rate
├── Breakdowns
│ ├── By Priority
│ ├── By Assignment Group
│ └── By Category
└── Thresholds
├── Critical: > 100
└── Warning: > 50
```
### Key Tables
| Table | Purpose |
| ------------------------- | ------------------------- |
| `pa_indicators` | KPI definitions |
| `pa_indicator_breakdowns` | Indicator-breakdown links |
| `pa_breakdowns` | Breakdown definitions |
| `pa_thresholds` | Threshold rules |
| `pa_widgets` | Dashboard widgets |
| `pa_dashboards` | Dashboard containers |
## Indicators
### Indicator Types
| Type | Aggregation | Example |
| -------------- | ---------------- | ------------------- |
| **Count** | COUNT(\*) | Number of incidents |
| **Sum** | SUM(field) | Total cost |
| **Average** | AVG(field) | Avg resolution time |
| **Percentage** | (A/B)\*100 | SLA compliance % |
| **Duration** | Time calculation | Avg time to resolve |
### Creating Count Indicator (ES5)
```javascript
// Create "Open Incidents" indicator
var indicator = new GlideRecord("pa_indicators")
indicator.initialize()
indicator.setValue("name", "Open Incidents")
indicator.setValue("description", "Number of currently open incidents")
// Source configuration
indicator.setValue("cube", "pa_cubes_incident") // or table-based
indicator.setValue("facts_table", "incident")
indicator.setValue("conditions", "active=true")
// Aggregation
indicator.setValue("aggregate", "COUNT") // COUNT, SUM, AVG
// Display settings
indicator.setValue("direction", 2) // 1=Up is good, 2=Down is good
indicator.setValue("unit", "Incidents")
indicator.setValue("precision", 0) // Decimal places
// Scoring
indicator.setValue("frequency", "daily")
indicator.insert()
```
### Creating Average Indicator (ES5)
```javascript
// Create "Average Resolution Time" indicator
var indicator = new GlideRecord("pa_indicators")
indicator.initialize()
indicator.setValue("name", "Avg Resolution Time")
indicator.setValue("description", "Average time to resolve incidents")
indicator.setValue("facts_table", "incident")
indicator.setValue("conditions", "state=6") // Resolved
// Aggregation on duration
indicator.setValue("aggregate", "AVG")
indicator.setValue("field", "calendar_duration") // Duration field
// Time unit
indicator.setValue("unit", "Hours")
indicator.setValue("unit_conversion", 3600) // Seconds to hours
indicator.setValue("direction", 2) // Lower is better
indicator.setValue("frequency", "daily")
indicator.insert()
```
### Creating Percentage Indicator (ES5)
```javascript
// Create "SLA Compliance" percentage indicator
var indicator = new GlideRecord("pa_indicators")
indicator.initialize()
indicator.setValue("name", "SLA Compliance Rate")
indicator.setValue("description", "Percentage of incidents meeting SLA")
indicator.setValue("facts_table", "task_sla")
indicator.setValue("conditions", "task.sys_class_name=incident")
// Formula-based percentage
indicator.setValue("aggregate", "FORMULA")
indicator.setValue("formula", "(COUNT(has_breached=false) / COUNT(*)) * 100")
indicator.setValue("unit", "%")
indicator.setValue("direction", 1) // Higher is better
indicator.setValue("precision", 1)
indicator.insert()
```
## Breakdowns
### Common Breakdowns
| Breakdown | Field | Use Case |
| --------- | ---------------- | --------------------- |
| Priority | priority | Incidents by P1/P2/P3 |
| Category | category | Incidents by type |
| Group | assignment_group | Team performance |
| Location | location | Geographic analysis |
| Time | opened_at | Trend analysis |
### Creating Breakdown (ES5)
```javascript
// Create "Priority" breakdown
var breakdown = new GlideRecord("pa_breakdowns")
breakdown.initialize()
breakdown.setValue("name", "Priority")
breakdown.setValue("description", "Breakdown by incident priority")
breakdown.setValue("facts_table", "incident")
breakdown.setValue("dimension_field", "priority")
// Sorting
breakdown.setValue("sort_field", "priority")
breakdown.setValue("sort_order", "ASC")
// Element mapping (optional)
breakdown.setValue("use_element_mapping", true)
breakdown.insert()
```
### Linking Breakdown to Indicator (ES5)
```javascript
// Link breakdown to indicator
var link = new GlideRecord("pa_indicator_breakdowns")
link.initialize()
link.setValue("indicator", indicatorSysId)
link.setValue("breakdown", breakdownSysId)
link.setValue("active", true)
link.insert()
```
## Thresholds
### Creating Thresholds (ES5)
```javascript
// Create threshold for Open Incidents
var threshold = new GlideRecord("pa_thresholds")
threshold.initialize()
threshold.setValue("indicator", indicatorSysId)
threshold.setValue("name", "Critical Level")
// Threshold conditions
threshold.setValue("operator", ">=") // >=, <=, =, >, <
threshold.setValue("value", 100)
// Visual styling
threshold.setValue("color", "red")
threshold.setValue("icon", "exclamation-circle")
// Notification
threshold.setValue("notification_user", adminSysId)
threshold.setValue("notification_script", thresholdScript)
threshold.insert()
// Add warning threshold
var warning = new GlideRecord("pa_thresholds")
warning.initialize()
warning.setValue("indicator", indicatorSysId)
warning.setValue("name", "Warning Level")
warning.setValue("operator", ">=")
warning.setValue("value", 50)
warning.setValue("color", "orange")
warning.insert()
```
## Widgets
### Widget Types
| Type | Use Case | Shows |
| ---------------- | ------------------- | -------------------- |
| **Single Score** | Current value | "127 Open Incidents" |
| **Scorecard** | Value + trend | Current + sparkline |
| **Time Series** | Trend over time | Line/bar chart |
| **Breakdown** | By dimension | Pie/bar chart |
| **Comparison** | Multiple indicators | Side-by-side |
### Creating Widget (ES5)
```javascript
// Create scorecard widget
var widget = new GlideRecord("pa_widgets")
widget.initialize()
widget.setValue("name", "Open Incidents Scorecard")
widget.setValue("type", "scorecard")
// Indicator
widget.setValue("indicator", indicatorSysId)
// Time range
widget.setValue("time_range", "last_30_days")
widget.setValue("show_trend", true)
widget.setValue("compare_to", "previous_period")
// Display
widget.setValue("show_breakdown", true)
widget.setValue("breakdown", priorityBreakdownSysId)
widget.setValue("chart_type", "bar")
widget.insert()
```
## Dashboards
### Creating Dashboard (ES5)
```javascript
// Create PA Dashboard
var dashboard = new GlideRecord("pa_dashboards")
dashboard.initialize()
dashboard.setValue("name", "Incident Management Dashboard")
dashboard.setValue("description", "Key metrics for incident management")
// Layout
dashboard.setValue("layout", "2-column")
// Access control
dashboard.setValue("public", true)
dashboard.setValue("owner", gs.getUserID())
var dashboardSysId = dashboard.insert()
// Add widgets to dashboard
function addWidgetToDashboard(dashboardId, widgetId, row, column) {
var placement = new GlideRecord("pa_dashboard_widgets")
placement.initialize()
placement.setValue("dashboard", dashboardId)
placement.setValue("widget", widgetId)
placement.setValue("row", row)
placement.setValue("column", column)
placement.insert()
}
addWidgetToDashboard(dashboardSysId, openIncWidget, 0, 0)
addWidgetToDashboard(dashboardSysId, avgTimeWidget, 0, 1)
addWidgetToDashboard(dashboardSysId, slaWidget, 1, 0)
```
## Data Collection
### Manual Score Collection (ES5)
```javascript
// Collect scores for an indicator
var job = new PAScoreCollector()
job.collectIndicatorScores(indicatorSysId)
```
### Scheduled Collection
```javascript
// PA uses scheduled jobs for data collection
// Default: Daily at midnight
// Configure via: Performance Analytics > Data Collection > Jobs
```
## MCP Tool Integration
### Available PA Tools
| Tool | Purpose |
| ------------------------------------------ | ------------------ |
| `snow_pa_create` (action='indicator') | Create indicator |
| `snow_pa_create` (action='breakdown') | Create breakdown |
| `snow_pa_create` (action='threshold') | Create threshold |
| `snow_pa_create` (action='widget') | Create widget |
| `snow_pa_operate` (action='get_scores') | Retrieve scores |
| `snow_pa_operate` (action='collect_data') | Trigger collection |
| `snow_pa_discover` (action='indicators') | Find indicators |
### Example Workflow
```javascript
// 1. Create indicator
var indicatorResult = await snow_pa_create({
action: "indicator",
name: "Open P1 Incidents",
table: "incident",
conditions: "active=true^priority=1",
aggregate: "COUNT",
})
var indicatorId = indicatorResult.sys_id
// 2. Create breakdown
var breakdownResult = await snow_pa_create({
action: "breakdown",
name: "By Assignment Group",
table: "incident",
field: "assignment_group",
})
var breakdownId = breakdownResult.sys_id
// 3. Attach the breakdown to the indicator
// (add_breakdown creates the breakdown + the indicator link in one step)
await snow_pa_indicator_manage({
action: "add_breakdown",
sys_id: indicatorId,
breakdown_name: "By Assignment Group",
breakdown_table: "incident",
breakdown_field: "assignment_group",
})
// 4. Create threshold
await snow_pa_create({
action: "threshold",
indicator: indicatorId,
type: "critical",
operator: ">=",
value: 10,
})
// 5. Create widget
await snow_pa_create({
action: "widget",
name: "P1 Incidents Scorecard",
type: "scorecard",
indicator: indicatorId,
breakdown: breakdownId,
})
// 6. Get current scores
var scores = await snow_pa_operate({
action: "get_scores",
indicator_sys_id: indicatorId,
time_range: "last_30_days",
})
```
## Best Practices
1. **Direction Matters** - Set correctly (up/down is good)
2. **Meaningful Thresholds** - Based on business requirements
3. **Consistent Frequency** - Match data volatility
4. **Use Breakdowns** - Enable drill-down analysis
5. **Dashboard Purpose** - One focus per dashboard
6. **Trend Analysis** - Always show comparison
7. **Performance** - Limit active indicators
8. **Documentation** - Clear indicator descriptions