KPIs & OKRs
Define outcomes, connect measurable work, collect evidence and maintain the system.
Start with a task
| I need to | Open |
|---|---|
| Follow collection from setup to result | Collect results |
| Define or propose an objective | Manage objectives |
| Connect a KPI to executable work | Manage tasks |
| Define a measurement and target | Define metrics · Set targets |
| Request recurring updates | Schedule collection |
| Respond with evidence | Submit results · Manual results |
| Check collection quality | Review results |
| Understand fields and dependencies | Entities · Architecture |
| Maintain or recover the automation | Automations · Troubleshooting |
Recorded and achieved
Metric defines the measurement; Metric Target defines the expected value for a subject and period. Recurrence Configuration requests a response; Observation stores it. Milestone holds an objective connected to targets through Task. Recorded confirms collection, not achievement. Evaluation is the separate interpretation of results under a named policy; no workflow in the current collection system fills it.
What runs today
The base holds the KPIs migrated on 2026-10-21: 159 Metrics, 193 Metric Targets and 193 Recurrence Configurations, all Active when migrated and with the same counts on 2026-10-28. Daily Digest asks the accountable people in Slack every day and has sent real messages since 2026-10-26. Answers come back through Reading Reply as Observations. People work the base in Data views and grid intakes; Airtable Interfaces and forms are not an operating path. No Airtable automation runs in this base.
The legacy Operations base is a different, older base. Its active KPI rows were copied here, but it is not part of this system. See Architecture.
Coverage and boundaries
These pages cover 26 current tables and 539 fields, operational fields, local mirror usage, all 67 Data views, three active n8n workflows, dependencies and recovery. Shared definitions and source fields live in Core. English is the source; PT-BR is equivalent and retains English product names. Screenshots use existing fictitious records and include genuine empty queues. The known gaps separate configuration, observed execution and missing implementation.