Stress-test an uptime commitment before the SLA goes live and proves the credit math is soft, the dependency map is incomplete, or the status page was theater.
Uptime commitments fail when credits are vague, when third-party outages still count against you with no carve-out, when maintenance windows are informal, and when sales already promised numbers engineering will not sign. A calendar invite labeled SLA is not evidence that the company can hold the line without chaos.
What to put on the table
One sentence for why the commitment exists, the measurement window you will honor, who owns credits and incident language, and the rollback trigger if the commitment cannot meet the reliability goal. Attach the dependency map, the status-page rules, the open risk list, and the customer language you would use. If product, eng, and support disagree on what "up" means, stop and reconcile first.
Name the decision you will make if the stress test finds nothing new, and the delay criteria if credit math or dependency owners are missing.
Attack surfaces
- Credit theater: refunds that exist on paper and fail under volume.
- Dependency silence: vendors and regions that can take you down without a named carve-out.
- Status-page gaps: incidents that never appear or appear too late to matter.
- Promise collisions: sales, partners, or launches already booked above the real ceiling.
- Domain overlap: commitments that share fate with a migration or vendor exit already in flight.
Optional customer seat if enterprise contracts sit inside the new number. Optional finance seat if credits or revenue recognition depend on uptime you just promised.
How to run it
Feed Pingpong the SLA draft, dependency map, and open risk list. Early passes steelman the plan. Later passes attack from product, eng, sales, and support seats. End with a pass that turns surviving objections into a clearer credit ladder, a named owner, or a hold. Delete invented "just this once" carve-outs and dual-counted spare capacity.
When the commitment coincides with a migration or status-page rewrite, force eng and support seats to map every path that still assumes quiet uptime. Maintenance, partial degradations, and partner enablement count. An uptime commitment that looks clean in a deck while sales still demos unfinished regions will fail on the first escalated deal.
Force a day-after narrative: what happens if a Sev-1 needs a credit, if a vendor outage slips into the window, or if two teams disagree on whether a degradation is an outage. If those stories are stronger than your mitigation plan, fix the package before you lock numbers. Related: stress-test an SLA change, stress-test a status page update, stress-test a backup plan, stress-test a security incident response, and the war-game decisions hub. Process: how to run a Pingpong.