Set go-live criteria before software can go live: measurable pass conditions, named owners, evidence, and clear reasons to stop. Your approval should reflect the risk customers will face, including whether essential services can recover if the release fails.
A completed build doesn’t prove operational readiness. You need evidence of business readiness, supported by a decision process that holds up under deadline pressure.
Start by separating mandatory launch gates from improvements that can safely follow.
Key Takeaways
- Treat customer harm, material data discrepancies, and untested recovery as launch barriers.
- Require signed readiness evidence before committing, then hold a final go/no-go meeting on launch day.
- Name one approving executive and publish everyone’s responsibilities.
- Keep support and monitoring active through the first billing, close, and reporting cycle.
Turn Go-Live Criteria Into a Decision Template
Go-live means making software available for real production use. An implementation checklist tracks the work; go-live criteria define the conditions for approving customer exposure.
Each gate needs decision criteria: a business consequence, pass condition, supporting evidence, owner, and deadline for approving customer exposure in the production instance. Use “yes,” “no,” or “not applicable with rationale,” rather than an unexplained green status.
Use this go-live checklist as the front page of your readiness record.
| Launch gate | Pass condition | Accountable owner |
|---|---|---|
| Customer workflows | All launch-critical acceptance tests pass; customer sign-off is recorded. | Business sponsor |
| Integrations | Production dependencies exchange and reconcile transactions correctly. | Technology lead |
| Performance | Peak-volume tests meet approved customer response-time requirements. | Engineering lead |
| Data | Migration reconciles with source records; no material discrepancies remain. | Data owner |
| Security | Required access and privacy controls work; no launch-blocking findings remain. | Security lead |
| Recovery and support | Recovery rehearsal meets business recovery objectives; support coverage is staffed. | Operations lead |
Attach dated test results and approvals behind the table. For exceptions, record the exposure, temporary control, accountable executive, and expiry date.
Avoid replacing failed gates with average scores. Five passing checks don’t cancel one defect that exposes another customer’s data.

Publish Decision Rights and Launch-Day Responsibilities
Separate delivery ownership from approval authority
Name a business sponsor who owns the customer outcome and a technology lead who owns the delivery recommendation. Assign one executive final go/no-go authority.
Publish a decision-rights map that names project stakeholders and defines their roles and responsibilities. Cover who recommends, tests, approves, funds, executes, and escalates. Include the project manager, system and database administrators, developers, team leads, and user representatives. Give each person a timeline and backup.
If a technology leadership gap leaves vendors effectively approving their own work, fractional CTO and interim CTO services can add independent executive judgment. Your business still retains approval authority.
Give communications a dedicated owner
Assign one person whose first priority is inbound launch-day communication. Don’t make the engineer diagnosing a failure also manage every stakeholder message.
Agree on communication channels and update frequencies for executives, employees, customers, and technology partners. Name who approves customer notices and emergency spending.
Also prepare customer-facing staff to explain disruption and workarounds. That preparation helps protect customer trust during upgrades.
Require Testing That Proves Customer Readiness
Complete acceptance, integration, and regression testing
Require end-to-end user acceptance testing with customer sign-off. The record should show that acceptance issues are resolved and connections between vendor and customer instances work.
Use integration testing across the complete chain of connected systems, including ERP, payments, identity, messaging, and fulfillment where applicable. A successful API response doesn’t prove that the correct transaction reached Finance.
Automate regression testing where practical. For a warehouse management system, cover receiving, putaway, picking, inventory management, and dispatch. For a customer portal, cover login, permissions, purchases, cancellations, and refunds.
Test peak demand and failure behavior
Performance testing and volume testing should use expected peak production volumes and realistic concurrent activity. Google’s Spanner launch checklist emphasizes traffic spikes and realistic concurrency.
Require the customer or accountable business owner to sign off that measured response times meet approved requirements.
Test failures too: an unavailable payment provider, an expired session, a duplicate submission, or a delayed downstream system. Evidence should show whether customers receive a useful response and whether staff can resolve the exception without losing transactions.
Sign Off Data, Hardware, and the Cutover Sequence
Your pre go-live checklist needs sign-off before you commit to launch. Then control changes so the signed evidence still matches the release you deploy.
For data conversion, rehearse the load and reconcile master data and legitimate transactional data against source records. Remove test transactions and mock data from production. Never treat cleanup as permission to delete real customer activity.
For a WMS cutover that requires a clean close, complete open orders and receipts in the current system. Document approved transfer exceptions and reconcile them explicitly.
Sequence migrations, access changes, rollouts, and hotfixes in a timed cutover plan. Google’s Compute Engine launch checklist calls for testing migration procedures alongside functional and load tests.
Where physical operations depend on the release, test configured workstations, report and label printers, and RF devices or scanners against production connections. Make several test prints and verify barcode reading works.
For web software, validate the production instance’s configuration, access, and customer-facing settings, including DNS, valid TLS certificates, HTTPS redirects, and analytics. Remove unintended staging links and test accounts. For public websites, check indexing settings, sitemaps, and error pages.

Prove Recovery Before Approving Customer Exposure
Rehearse restoration and operational fallback
Backup plans should cover the entire launch window: approval, deployment, partial failure, extended cutover, containment, and rollback. Give every high-risk dependency an owner and stopping point.
Test backup plans through restoration, emergency access, vendor escalation, and essential manual procedures against business recovery objectives. This tests disaster recovery through real restoration and fallback, not documentation alone. For inventory-dependent operations, include a controlled master inventory reload option.
Google Cloud’s recovery-testing guidance calls for safeguards such as monitoring, backups, and manual rollback procedures.
A tabletop discussion tests understanding. An executed restore tests whether recovery works. Retain timestamps, system outputs, decisions, rollback results, and unresolved findings.
Protect transactions created after cutover
Restoring the old system can discard legitimate orders, payments, or account changes entered after launch. Your rollback plan must explain how you capture and reconcile that activity.
Where reversal isn’t feasible, document containment and business continuity arrangements. Untested recovery and unresolved material discrepancies remain decision barriers.
Define extensions in advance. State which milestone permits more time, who approves it, and when the team must stop. A deadline extension needs evidence that progress remains safe.
Run the Final Go/No-Go Meeting on Launch Day
Hold the final launch-day no/go meeting with every required decision owner present, including relevant customer and technology partner representatives. Review the agreed criteria against current evidence.
The approving executive chooses whether to proceed, narrow exposure, delay, or stop. Record the decision, remaining risks, authorized exceptions, and next checkpoint.
A failed mandatory gate means no-go. An improvement can follow launch only when the accountable executive accepts its consequences and assigns a deadline.
A working website with failing payment reconciliation is a no-go. A cosmetic defect can proceed only if it doesn’t impair accessibility, customer completion, or another agreed launch requirement.
After deployment, run production smoke tests before expanding access. Microsoft’s cloud-native deployment guidance includes smoke testing to confirm that services and core functions work in production.
If those tests fail, invoke the agreed containment or rollback decision.
Keep Hypercare Open Through the First Business Cycle
Agree on an intensive post go-live support period with all technology partners before launch. Publish coverage, escalation contacts, and response commitments. Prepare key users with practical training and documentation so they can support users and explain available workarounds.
Track success metrics and KPIs throughout the go-live period. Include customer completion rates, failed transactions, reconciliation differences, support demand, and revenue-impacting incidents. Compare results with approved launch expectations.
Continue post go-live support through the first billing run, financial close, and reporting cycle. Some defects emerge only when downstream processes execute.
Require project-team sign-off on a post go-live checklist confirming secure, dependable operation. Exit hypercare only when incidents fall within agreed limits, owners accept remaining work, and routine support can take over.
For directors, keep board-ready technology risk reporting focused on customer exposure, unresolved material risks, and decisions requiring approval. Move accepted technical debt into the technology roadmap with funding and ownership.
Frequently Asked Questions
What belongs in a go/no-go checklist?
Include customer acceptance, integrations, peak-volume performance, security, data reconciliation, production configuration, recovery, and support. Each item needs a pass condition, evidence, owner, and consequence of failure. Add system-specific gates rather than copying irrelevant checks.
Who should approve a software go-live?
One named executive should approve the business exposure. Your business sponsor and technology lead provide recommendations. Security, Finance, Legal, Operations, and customer representatives contribute evidence where their responsibilities apply. Vendors shouldn’t approve their own readiness unchallenged.
Can you launch with unresolved defects?
Yes, if they don’t violate mandatory launch gates. Record the customer impact, workaround, owner, resolution deadline, and executive acceptance. Defects involving material data loss, unauthorized access, or failed essential transactions should block launch.
Make Launch Approval an Evidence-Based Business Decision
Your go-live criteria should make the decision clearer before deadline pressure arrives. Require named ownership and tested evidence, then keep accountability active after customers begin using the software.
Start your next readiness conversation with the failed gates, recovery proof, and decisions requiring executive approval.
If those answers remain scattered, Get an Executive Technology Clarity Check to establish sharper priorities, clearer ownership, and a practical next step.