.svg){kind=logo}
Trace → Assign → Enforce → Prove is Anavsan’s closed accountability loop. Trace finds recurring, fixable waste — not a spend leaderboard. Assign names a party from evidence, or says needs owner instead of guessing. Enforce drives the exact change through your review process until it is applied, deferred with a reason, or dismissed. Prove closes the finding when the metric returns to a normal band. Together they are how workload governance is executed.
Key takeaways
The four stages
Find the waste, not the spend
Detect a recurring, fixable inefficiency and quantify the excess: the gap between what you paid and what an efficient configuration would have cost. Recurring remote spillage, idle compute, and refresh storms are mechanisms. A one-off spike is not Trace. A busy, efficient warehouse never earns a row. Mapping every credit on a spend leaderboard is not the whole story.
Name a party, never the wrong one
Resolve the accountable person or team from real ownership evidence — tags, owning role, organisation map. When ownership traces to a generic admin role, mark it needs owner instead of guessing. Confidence gates the channel, never the ranking. Inferred usage is a suggestion, never an auto-assignment or a personal accusation. The Private Knowledge Graph is what makes this specific.
Drive it to a decision
Give the owner the exact change, the risk, and the savings, then track it through your own review and pull-request process until it is applied, deferred with a reason, or dismissed. Archive-with-reason stays distinct from “the waste stopped.” The Enforcement Desk is the queue. Anavsan does not silently alter production warehouses.
Close it from the data
Watch the underlying metric return to a normal band and resolve the finding automatically. Nobody self-reports a win. Simulation can sit beside the loop as change-safety before a change lands; it is not how closure is proved, and it is not a banked-dollar receipt against the estimate.
Incomplete loops
| What teams have | Missing stage | What happens |
|---|---|---|
| Dashboards and alerts | Assign, Enforce, Prove | Alert fatigue |
| Tickets without a named owner | Assign | Nobody decides; the finding ages out |
| Advice in Slack, no PR | Enforce | No applied, deferred, or dismissed record |
| A rewrite or a simulation with no follow-up | Prove | The estimate is treated as the win; the pattern returns |
| Caps and resource monitors | Trace + Assign | Surprise suspends, not cheaper queries |
Where it runs
On Anavsan, the engine is APEX. The Enforcement Desk is the ranked worklist. Ask APEX is grounded chat over the same findings. The Native App stops at visibility — useful when security will not approve anything else, incomplete when alerts already exist and stall. Choose from the constraint in front of you, not from a feature matrix.
Related terms
APEX
The engine that runs the loop.
Read →Workload Governance
The discipline the loop operationalizes.
Read →Enforcement Desk
The ranked worklist where Enforce and Prove are visible as status.
Read →Cost Accountability
The outcome when all four stages complete.
Read →