This guide is for product owners, engineering leads and buyers of outsourced development who need a release decision that does not rest on confidence alone. Our software quality governance guide explains who owns quality and how the gate fits a vendor relationship; this page gives the working checklists, item by item.
In this guide
- Two checklists, two moments
- Checklist 1: is the story ready to build?
- Criteria patterns by story type
- Checklist 2: is the release ready to go live?
- Size the gate to the risk
- No-go triggers
- Run the go or no-go decision
- AI in acceptance and release work
- Alternatives: how to gate a release
- How Netbase runs acceptance and release readiness
- What delivery record exists, and what does not
- Limits of this guide
- Frequently asked questions
- Next step
Two checklists, two moments
Acceptance criteria are agreed before a story is built. Release readiness is judged after the build, on a candidate that could ship. Mixing them is how teams discover in the final week that nobody wrote down what "done" meant. Keep the first in the backlog and the second in the release record, and make the first feed the second: every acceptance criterion that passed becomes a line of release evidence.
Checklist 1: is the story ready to build?
The Scrum Guide says work that does not meet the Definition of Done cannot be part of an Increment, and that refinement breaks items into smaller, more precise pieces. A story is ready to build when:
- The outcome is observable. A tester who missed the planning meeting can decide pass or fail from what the screen, the data or the log shows.
- Limits are stated. File sizes, volumes, response times you care about, supported browsers and devices, and the roles allowed to do it.
- Failure cases are written. What happens on invalid input, a timeout, a double click, an expired session and a refused payment.
- Test data is named. The roles, records and edge cases needed to test it exist in a test environment.
- Non-functional needs are attached. Accessibility, performance and security needs sit in the criteria or in the shared definition of done.
- Dependencies are known. Other teams, systems or decisions the story waits for are listed, with an owner.
- The size fits one iteration. If the criteria need a page of text, split the story.
- The product owner has approved the wording. Approval happens before the story enters a sprint, not during testing.
Criteria patterns by story type
| Story type | Criteria to write | Failure cases to include |
|---|---|---|
| Sign-in and permissions | Which role sees and does what; what a signed-out user sees | Wrong role, expired session, a link opened by someone else |
| Checkout or payment | Totals, taxes and statuses at each step; what the customer receives | Declined payment, double submit, abandoned and resumed order |
| Data import | Accepted formats and limits; what a finished import reports | Bad rows, duplicates, a file that is too large, a partial failure |
| Search and lists | Sort order, filters, paging and empty results | Very long lists, special characters, no match |
| Email and notifications | Trigger, recipient, content and timing | Bounce, unsubscribed recipient, repeated trigger |
| Reports and exports | Fields, filters, totals and who may export | Empty period, large export, data from another account |
Write each criterion as a given, when, then sentence, and keep the table above as a prompt, not as a template to fill blindly. The point is that a failing case is described before the code exists.
Checklist 2: is the release ready to go live?
Google's Launch Coordination Checklist, written for its own services, covers architecture, capacity, failure modes, monitoring, security and the rollout plan. The same areas apply to most products, and each needs evidence you can open:
- Scope. Release notes list the stories included and the known issues accepted, each with an owner.
- Functional quality. Acceptance tests and the regression suite pass on the release candidate, verified by someone other than the author.
- Security. Scans ran, findings were reviewed, and every open finding has a named owner who accepted it.
- Performance. Key journeys were measured against the limits agreed in the criteria.
- Data and migrations. Each migration was rehearsed on a copy of real-shaped data, and a backup was taken before the change.
- Rollback. A rollback was tested, or a feature flag can switch the change off without a deploy.
- Operations. Monitoring and alerts cover the new behaviour, the runbook is updated, and support knows what changed.
- Dependencies and configuration. Environment settings, third-party keys and scheduled jobs match between test and production.
- Communication. Customer notices, help pages and internal announcements are ready where users will notice a change.
- Decision record. Who decided, when, on which evidence, and which risks were accepted.
Size the gate to the risk
Not every release needs every line. Three tiers keep the checklist alive instead of ignored:
- Routine change. A small fix passes on automated evidence: tests green, review done, rollback by redeploy.
- Standard release. The full list above, with the evidence attached to the release record.
- High-risk release. Data migration, payments, permissions or a change many customers will see. Add a rehearsal in a production-like environment, a staged rollout and a named person on watch after launch.
No-go triggers
Agree these before the meeting so nobody negotiates them under deadline pressure:
- An acceptance criterion fails or has no test.
- A critical or high security finding is open without a named owner's acceptance.
- No tested rollback exists for a change that cannot be switched off.
- A migration was not rehearsed.
- Monitoring cannot show whether the new behaviour works.
- The person who must sign is not available or was not shown the evidence.
Run the go or no-go decision
-
Freeze the candidate
Name the exact build and the stories in it; changes after the freeze restart the checklist for the affected items.
-
Collect the evidence
Owners attach test results, scan reports, migration rehearsal notes and the rollback record to the release record.
-
Walk the list
In a short meeting or an asynchronous review, each owner states pass, fail or accepted risk.
-
Decide and record
The release owner decides go or no-go, and the record names accepted risks and who accepted them.
-
Release in stages where the risk calls for it
A flag, a canary group or an off-peak window limits what a miss can cost.
-
Watch and close
Someone watches errors, key journeys and support requests for an agreed period, then closes the record or triggers the rollback.
AI in acceptance and release work
AI shortens the drafting work in both checklists and adds a new kind of acceptance test. Drafting test cases and edge cases from approved criteria, generating synthetic test data and grouping duplicate defects are useful, but each output is a draft that a named person accepts, and a go decision is never automated. Features that use AI need acceptance as a measured result on an agreed evaluation set, re-run when the model or prompt changes. Netbase works with the major commercial and open-source AI models, chosen per project. Each item below states how mature it is at Netbase.
-
In delivered commerce work
Recommendation engine
Built for 4over4's online printing store; a web store feature, not a release-testing record.
-
Available capability
AI-assisted test generation and AI feature evaluation
Machine learning, NLP and generative AI capabilities; not yet tied to a published quality engineering case.
Alternatives: how to gate a release
| Model | Strength | Weakness | Choose it when |
|---|---|---|---|
| Manual checklist and meeting | Cheap, flexible, forces a conversation | Depends on discipline; slow if every release needs a meeting | Few releases, mixed evidence, or a buyer who signs off |
| Automated pipeline gates | Fast and consistent; blocks a release on failing checks | Only checks what someone automated | Frequent releases with a solid regression suite |
| Progressive delivery with flags or staged rollout | Limits the damage of a miss and shortens rollback | Needs flag hygiene and good monitoring | Many users, high change rate, or risky features |
| Buyer-side acceptance testing | The person who pays confirms the outcome | Late feedback if criteria were vague | Outsourced delivery, contractual milestones |
Choose by four criteria: how often you release, how costly a failure is to customers, how much of the evidence is already automated, and who is contractually accountable for sign-off. Most teams combine the first two or three rows.
How Netbase runs acceptance and release readiness
Netbase runs quality assurance inside delivery: acceptance criteria written with the product owner, automated regression, performance budgets and a release readiness gate, as described on the quality engineering and testing service. Delivery is governed by weekly reviews, KPI dashboards and a dedicated account and project manager, in teams of 3 to 30 that combine analysts, developers, QA and designers. Security practice includes secure code review, vulnerability scanning and penetration testing; Netbase's own credentials cover how Netbase works, not a client's product. Communication is in English, and support runs Monday to Saturday, Sunday off. Most Netbase projects are delivered on fixed-price contracts agreed after discovery, which makes written acceptance criteria the basis of scope; see dedicated team vs fixed price. For a vendor's security answers beside the release evidence, see security questions for a software development partner; the SaaS MVP roadmap shows a launch readiness list for a first release.
What delivery record exists, and what does not
- What exists. The USticker and PrintLeo records describe commerce work with performance and stability improvements, and Netbase's published governance model includes weekly reviews and client dashboards.
- What does not. No Netbase record publishes defect rates, release frequency, rollback counts or the outcome of a release gate, and neither case is offered as proof that a given checklist prevents failures. The checklists here are general practice, not results.
Limits of this guide
- The lists are a starting point; regulated products add their own approvals, such as change records or audit trails, so take specialist advice.
- Metrics such as DORA's change fail rate, the share of deployments that need immediate intervention, show whether the gate works over time; they are outcomes, not items to tick.
- A checklist records evidence; it cannot make a weak test suite strong.
Plan the next step with a Netbase consultant
Frequently asked questions
Each item has its own owner, such as the QA lead for test evidence, a security owner for findings and the product owner for scope; one release owner makes the final decision and records it.
Detailed enough that a tester who never attended planning can decide pass or fail, and no longer than one iteration's worth of work. If the criteria run to a page, split the story.
Yes, when each one is known, rated and accepted by a named owner in the release record. An unknown or unowned defect is a no-go.
Use the routine tier for small changes and keep the no-go triggers for every release; skipping evidence on high-risk changes is how failures reach customers.
Next step
Share a recent release, its acceptance criteria and who signed it off, and we will book a solution review to adapt both checklists to your team. You can also see quality engineering and testing or more Netbase insights.
Related services and solutions
AI-assisted QA and software testing that run inside delivery
Netbase provides software testing and QA services for product and commerce teams to release often without breaking what customers use. QA runs inside delivery, not after it: acceptance criteria, automated regression with AI-generated test drafts, performance budgets and release readiness checks. Page speed shows the result: USticker's page load fell 21%, and PrintLeo's improved 35%.
Learn More
Web-to-print platform with AI design assistance, from online design to a print-ready order
A web-to-print platform is an online ordering, design and prepress workflow that helps print businesses sell custom products: customers configure, design and approve their order online, AI can suggest layouts and catch artwork problems, and production receives a print-ready file. Netbase has delivered 50+ custom web-to-print platforms across apparel, packaging, signage, promotional merchandise and corporate B2B portals.
Learn More
Discuss a project
Netbase JSC helps organizations design, build, modernize, and operate digital products and AI-enabled business systems.+84 937 869 689
91 Nguyen Chi Thanh, Dong Da, Hanoi, Vietnam
Get in touch
Tell us what you want to build, modernize, or operate.