The PA program shipped 47 indicators, the dashboard takes 90 seconds to render, and the executive who sponsored it has not opened it in three months. Performance Analytics is genuinely powerful and routinely overbuilt. The indicators that drive decisions tend to be a small subset; the dashboards that get opened weekly are usually the ones that started narrow and earned their breakdowns by demonstrated use.
PA vs Live Queries
Standard queries are point-in-time. Performance Analytics is time-series — the same metric measured daily, weekly, or monthly, with targets, breakdowns, and drill-through to underlying records. If the question is “what is the trend,” use PA. If the question is “what is the current state right now,” use a live query. Mixing them on a single dashboard is fine; using one when the other is right for the job is the most common modeling mistake.
Indicators Are Your Building Block
An indicator is a metric definition: the source query, the aggregation type, the collection frequency, and any explicit filters. Start small — 10 core indicators that map to the operations leadership’s weekly conversation (incidents created, MTTR, backlog age, SLA breach rate, change success rate). Add dimensions over time as questions emerge from the dashboard, not speculatively. Indicators are easy to add and painful to remove once dashboards depend on them.
Starter indicator set:
incidents_opened (daily)
incidents_resolved (daily)
incident_backlog_age (daily)
mttr_p2 (daily)
sla_attainment_rate (daily)
change_success_rate (weekly)
change_volume (weekly)
csat (weekly)
problem_closed_with_change (monthly)
knowledge_article_helpfulness (weekly)Breakdowns Unlock Value
An indicator without breakdowns is one number. Add breakdowns (priority, assignment group, category) and the same indicator answers ten questions via drill-through. Do not over-break — keep breakdowns to dimensions that drive action. A 200-value breakdown produces a chart no human reads. Use element grouping to bucket long-tail values into “top 10 + rest” for routine views; analysts can drill into the long tail when needed.
Targets and Thresholds
A dashboard without targets is trivia. Set baseline targets (last year’s average or current operational floor) and aspirational targets (the goal). Thresholds convert numbers into traffic lights — green within target, yellow within warning band, red beyond. Executives want the traffic light, not the number. The threshold definition is the conversation with leadership about what “good” means; do not skip it.
// PA threshold check helper
function evaluateThreshold(score, indicator) {
if (indicator.direction === 'up') {
return score >= indicator.target_value ? 'green' :
score >= indicator.warning_value ? 'yellow' : 'red';
} else {
return score <= indicator.target_value ? 'green' :
score <= indicator.warning_value ? 'yellow' : 'red';
}
}Data Collector Load
PA runs scheduled data collectors against the source tables. On large instances, uncoordinated schedules can load the database during the same windows as user queries. Stagger collectors across off-peak hours, monitor scheduled job queue depth, and budget the collection cost per indicator. A PA dashboard that tanks instance performance during business hours is not worth having; tune the schedule before the user complaints.
Common Failure Modes
Indicators built for a stakeholder who left the organization — quarterly review of indicator usage and ownership; orphaned indicators with no recent dashboard views are retirement candidates. Breakdowns with cardinality that exceeds the user’s screen — convert to element grouping or split into separate indicators. Collection schedules that run hourly when daily would suffice — the cost compounds quickly with breakdowns; default to daily and justify higher frequencies.
What Changed in 2026
Workspace dashboards in Next Experience lazy-load tiles below the fold, which mitigates the cost of mixing PA tiles with heavier live-query tiles. The AI Control Tower and the platform’s anomaly detection (where licensed) can flag scoring anomalies for investigation rather than waiting for the human to notice. Older Service Portal dashboards render everything eagerly and need stricter tile budgets.
Cost Considerations
Each indicator consumes scheduled job time and storage in pa_scores. A breakdown with 100 dimension values multiplies score storage by 100. The cost is not just dollars — it is database time during collection windows. Monitor pa_scores row count per indicator and retire indicators whose marginal value does not justify the storage and collection cost.
Implementation Sequence
Build the executive dashboard first with five indicators, three breakdowns each, and explicit targets. Validate against three weeks of leadership conversations — does the dashboard support the questions actually being asked? Add additional dashboards (operations team, individual support groups) only after the executive surface is stable. Indicators that nobody opens for 30 days enter a deprecation queue; cleanup is part of PA hygiene, not a separate project.
What to do this week: open your top three PA dashboards and identify which indicators have been viewed in the last 30 days; the ones that have not are your retirement candidates.
