KPI Control Tower
The Control Tower turns scattered product and operational signals into an ordered decision queue. It keeps the evidence, the verdict, and the response together: see what moved, decide what matters, then start the workflow that should change it.
The operating loop
Map evidence
Choose a numeric integration metric or create a manual KPI.
Define success
Set the target direction, value, evidence window, scope, and priority.
Triage the queue
Evidence problems and off-track outcomes rise above healthy KPIs.
Start an intervention
Run the bound workflow or create one with the KPI evidence attached.
What appears in the queue
Fleet separates an unhealthy outcome from unhealthy evidence. A broken data source never masquerades as a healthy KPI.
| Evidence problem | Fleet cannot assign a trustworthy verdict because evidence is missing, stale, incomplete, invalid, unreadable, or overdue for evaluation. |
| Needs attention | The KPI is breached or at risk. These outcomes are ranked ahead of healthy KPIs so the next decision is visible. |
| Healthy | Current evidence satisfies the condition. Healthy KPIs are collapsed by default to keep the queue focused. |
KPI definitions
Evidence
A numeric metric published by a connected integration, or a manually recorded observation.
Condition
At least or at most a target value, with an aggregation and evidence window for integration metrics.
Scope
Company, product, or repository ownership, plus priority.
Response
An optional saved workflow to use when the KPI is breached or at risk.
Preview before activation
Preview runs the real evaluator against current evidence. Activation is blocked when evidence is insufficient, stale, or changed after the preview. When you edit an active KPI, the active definition keeps running until the new draft passes preview and is activated.
Evidence sources
Integration evidence
Once a connected integration publishes a numeric metric descriptor, it is available in the mapping flow. Select required dimensions, aggregation, and the evidence window Fleet should evaluate.
Manual evidence
Start with a current value, unit, target, and update cadence. Record new observations in Fleet. When the metric becomes available, open the manual definition and choose Map integration evidence to keep the KPI identity while replacing its source.
Interventions
A breached or at-risk KPI can start its default workflow directly from the evidence rail. If no workflow is bound, Fleet opens the workflow builder with the KPI context attached. An active intervention is shown in place so the same response is not launched twice. Evidence problems must be repaired before a response can start because Fleet will not automate from an untrusted verdict.
Management and access
Daily triage lives in the Control Tower; setup and lifecycle work lives in Manage KPIs. Operators with integration-mapping permission can create, edit, preview, activate, pause, resume, archive, delete, record manual observations, and start interventions. Other signed-in users can inspect the queue, evidence, and definitions without changing them.