Separate dependency failure from the customer-facing objective
The team wants to exclude payment-provider outages from checkout reliability.
- Focused work estimate
- 3h + prerequisites
- Priority in the scenario
- Medium
- Engineering practice
- Reliability analysis
Estimated field mix
- Site reliability80%
- System design20%
Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence.
Review it, then add it to your workspace.
The board opens an editable draft; nothing is saved until you confirm it. Sign-in and workspace permissions apply, and Demo boards remain ephemeral.
Project context
A fictional checkout service reports process uptime while customers experience failed orders and slow confirmations.
Setup prerequisites
- Create synthetic request and order-event traces plus a local query or metrics fixture; no production telemetry required.
Preceding work
Complete these dependencies, or supply their agreed outputs before taking this ticket.
- BSLO-101 · Define eligible checkout attempts and successful completion
- BSLO-102 · Deduplicate retried checkout attempts in indicator calculations
- BSLO-103 · Compute completion latency from consistent event pairs
- BSLO-104 · Keep missing telemetry distinct from successful service
- BSLO-105 · Implement multi-window error-budget burn calculations
Acceptance criteria
- Keep customer outcome accounting intact.
- Add dependency attribution as a separate diagnostic dimension.
- Expose uncertain attribution without changing the denominator.
Implementation constraints
- Do not transfer responsibility by redefining failed outcomes away.
Verification to include
- Count a dependency-caused customer failure.
- Show attribution missing while retaining the failure.
Deliverables
- Attribution dashboard contract.
Rollout and recovery
Add diagnostics alongside the objective; preserve historical customer outcomes.
Value of the work
For the engineer: Practice SLI semantics, alert evaluation, and reliability tradeoffs.
For the team: Align reliability discussions with customer outcomes and explicit measurement limits.
Evidence boundaries
Outcome Evidence: Tests, patches, and runbooks are requested deliverables. They become Outcome Evidence only through a qualified Mission and immutable Evidence IDs.
Ownership Evidence: Independent adaptation must be observed under a declared verification policy and cite immutable Evidence IDs. Completing a planning ticket establishes no Ownership Evidence.