{{project}} / review owner {{business}} / approver {{contact}} / target {{deadline}}
Purpose: make a bounded release decision from reproducible evidence. A checked box is not a test result.
Candidate version/build [ ]; staging address [ ]; production address [ ]; scope freeze date [ ].
Primary visitor task [ ]; completion evidence [ ]; consequence if broken [ ].
Pages, forms, payments, downloads, and external dependencies actually in scope [ ]. Mark absent features not applicable with a reason; do not imply a checkout was tested on an enquiry-only site.
Test environment matrix
ID | browser/version | operating system/version | viewport/zoom | real device or simulation | input/assistive tool | rationale
[Environment | exact versions | system | size | method | keyboard/screen reader where used | audience evidence]
State what is not tested and who accepts or investigates that gap. Emulation is not evidence of behavior on a real device.
Test record
ID | candidate/environment | starting state and fixture | action | expected observable result | actual result | pass/fail/blocked/not applicable | evidence | tester/time
[Case | version | precondition | reproducible steps | completion/failure state | observation | status | sanitized record | name/date]
Keep expected behavior separate from what happened. A blocked test stays blocked; a production test does not inherit a staging pass.
Release gates agreed before testing
G1 Primary task completes and the confirmation means what the copy promises.
G2 Required-field, failed-request, unknown-outcome, retry, and fallback states are usable for the supported route.
G3 Offer/contact information and applicable delivery behavior match the reviewed version.
G4 Task-blocking barriers are resolved; accepted minor issues have an explicit owner and date.
G5 The previous version and a tested recovery method are available to the release owner.
Adapt these gates and define any additional requirements. Record unmet gates rather than lowering them after a failure.
Fixtures must be non-sensitive. Use an isolated test environment for injected failures; do not disrupt visitors or probe unrelated systems. Identify who can authorize production actions and any payment rehearsal before it occurs. Do not store credentials, payment details, full session links, or real enquiry text in shared evidence.
Decision: proceed to production verification / hold / revise. Reason and evidence references [ ]; approver/date [ ]; untested scope and next owner [ ].
Your decision
Which broken journey would prevent a visitor completing their task?
My answer: [ ]
Source and date: [ ]
Still to check: [ ]
Next action, owner, and due date: [ ]
Before you start
- Preview site: [Source or notes]
- Critical visitor tasks: [Source or notes]
- Rollback access: [Source or notes]
What you will make
A release decision with recorded tests and a named person responsible for rollback.
View plain worksheet text
{{project}} / review owner {{business}} / approver {{contact}} / target {{deadline}}
Purpose: make a bounded release decision from reproducible evidence. A checked box is not a test result.
Candidate version/build [ ]; staging address [ ]; production address [ ]; scope freeze date [ ].
Primary visitor task [ ]; completion evidence [ ]; consequence if broken [ ].
Pages, forms, payments, downloads, and external dependencies actually in scope [ ]. Mark absent features not applicable with a reason; do not imply a checkout was tested on an enquiry-only site.
Test environment matrix
ID | browser/version | operating system/version | viewport/zoom | real device or simulation | input/assistive tool | rationale
[Environment | exact versions | system | size | method | keyboard/screen reader where used | audience evidence]
State what is not tested and who accepts or investigates that gap. Emulation is not evidence of behavior on a real device.
Test record
ID | candidate/environment | starting state and fixture | action | expected observable result | actual result | pass/fail/blocked/not applicable | evidence | tester/time
[Case | version | precondition | reproducible steps | completion/failure state | observation | status | sanitized record | name/date]
Keep expected behavior separate from what happened. A blocked test stays blocked; a production test does not inherit a staging pass.
Release gates agreed before testing
G1 Primary task completes and the confirmation means what the copy promises.
G2 Required-field, failed-request, unknown-outcome, retry, and fallback states are usable for the supported route.
G3 Offer/contact information and applicable delivery behavior match the reviewed version.
G4 Task-blocking barriers are resolved; accepted minor issues have an explicit owner and date.
G5 The previous version and a tested recovery method are available to the release owner.
Adapt these gates and define any additional requirements. Record unmet gates rather than lowering them after a failure.
Fixtures must be non-sensitive. Use an isolated test environment for injected failures; do not disrupt visitors or probe unrelated systems. Identify who can authorize production actions and any payment rehearsal before it occurs. Do not store credentials, payment details, full session links, or real enquiry text in shared evidence.
Decision: proceed to production verification / hold / revise. Reason and evidence references [ ]; approver/date [ ]; untested scope and next owner [ ].
Your decision
Which broken journey would prevent a visitor completing their task?
My answer: [ ]
Source and date: [ ]
Still to check: [ ]
Next action, owner, and due date: [ ]
Before you start
- Preview site: [Source or notes]
- Critical visitor tasks: [Source or notes]
- Rollback access: [Source or notes]
What you will make
A release decision with recorded tests and a named person responsible for rollback.