Agent.MD: Deliver Phase
Phase Overview
Phase: Deliver (Phase 5 of 5)
Purpose: Launch the solution, monitor performance, gather feedback, and iterate based on real-world usage.
Duration: Ongoing (continuous deployment and iteration)
Mindset: "Ship, learn, improve" - Release to users and iterate based on data
HCDAgile Deliver Principles
In the Deliver phase, we prioritize:
- User Value First: Launch features that provide real value
- Measured Rollout: Deliver gradually to manage risk
- Continuous Monitoring: Watch metrics and user feedback closely
- Rapid Response: Fix issues and iterate quickly
- Learning Culture: Treat deployment as learning opportunity
Primary Goals
- [ ] Deploy solution to production
- [ ] Monitor system performance and stability
- [ ] Track success metrics and KPIs
- [ ] Gather user feedback and usage data
- [ ] Identify and fix issues quickly
- [ ] Plan and implement improvements
- [ ] Maintain and support the solution
Key Activities
Pre-Deployment
- Final testing in staging environment
- Security and performance review
- Deployment plan creation
- Rollback plan preparation
- Communication plan to users
- Training materials preparation
Deployment
- Deploy to production environment
- Execute deployment plan
- Monitor deployment process
- Validate deployment success
- Enable monitoring and alerts
- Communicate to stakeholders
Post-Deployment
- Monitor system health
- Track success metrics
- Gather user feedback
- Respond to incidents
- Analyze usage patterns
- Plan iterations
Continuous Improvement
- Review metrics regularly
- Prioritize improvements
- Implement enhancements
- A/B test variations
- Optimize performance
- Address technical debt
Deliverables
- Deployment Plan: Step-by-step deployment process
- Rollback Plan: How to revert if issues arise
- Monitoring Dashboard: Real-time system health view
- User Communication: Release notes and announcements
- Support Documentation: User guides and troubleshooting
- Incident Response Plan: How to handle issues
- Performance Reports: Regular metrics analysis
- Iteration Roadmap: Plan for continuous improvement
AI Agent Instructions
Your Role in Deliver
As an AI agent in the Deliver phase, your role is to:
- Assist with deployment planning
- Monitor system performance
- Analyze usage data and metrics
- Identify patterns in user feedback
- Suggest improvements based on data
- Draft release notes and documentation
- Help troubleshoot issues
- Support incident response
- Track and report on KPIs
Deployment Best Practices
Before Deployment:
- Review all test results
- Verify staging environment works
- Prepare rollback plan
- Schedule deployment window
- Communicate to stakeholders
- Backup data
During Deployment:
- Follow deployment checklist
- Monitor for errors
- Verify each step succeeds
- Test critical paths
- Watch server metrics
- Be ready to rollback
After Deployment:
- Validate deployment success
- Monitor error rates
- Check user feedback
- Review performance metrics
- Communicate success
- Document lessons learned
Monitoring Focus Areas
Technical Metrics:
- Server response times
- Error rates
- API performance
- Database queries
- Resource utilization
- Uptime/availability
User Metrics:
- User adoption rate
- Feature usage
- Task completion rates
- User satisfaction scores
- Support ticket volume
- User feedback sentiment
Business Metrics:
- Conversion rates
- Engagement metrics
- Retention rates
- Revenue impact (if applicable)
- ROI calculations
- Goal achievement
Working Approach
- Data-Driven: Base decisions on metrics and feedback
- User-Focused: Prioritize user experience over technical perfection
- Proactive: Monitor and address issues before they escalate
- Iterative: Continuously improve based on learnings
- Collaborative: Work with all stakeholders
Prohibited Actions
- ❌ Do not deploy without adequate testing
- ❌ Do not deploy without rollback plan
- ❌ Do not ignore monitoring alerts
- ❌ Do not make changes without tracking
- ❌ Do not deploy during high-traffic periods (without planning)
- ❌ Do not skip post-deployment validation
- ❌ Do not ignore user feedback
Success Criteria
The Deliver phase is successful when:
- ✅ Deployment completed without critical issues
- ✅ Success metrics meet or exceed targets
- ✅ User feedback is positive
- ✅ System performance is stable
- ✅ No critical bugs in production
- ✅ Support team is prepared
- ✅ Continuous improvement cycle is established
- ✅ Stakeholders are satisfied
Common Pitfalls to Avoid
- Big Bang Release: Deploying everything at once without gradual rollout
- Insufficient Monitoring: Not tracking the right metrics
- Ignoring Early Signals: Missing warning signs of issues
- Poor Communication: Not informing users/stakeholders
- No Rollback Plan: Unable to revert if problems arise
- Deploy and Forget: Not continuing to monitor and iterate
- Feature Dumping: Deploying features users don't need
Deployment Strategies
Blue-Green Deployment
- Maintain two identical environments
- Deploy to inactive environment
- Switch traffic when validated
- Quick rollback if needed
Canary Deployment
- Deliver to small subset of users first
- Monitor for issues
- Gradually increase user percentage
- Full rollout when validated
Feature Flags
- Deploy code with features disabled
- Enable features gradually
- Test in production safely
- Quick rollback by disabling flag
Rolling Deployment
- Update servers one at a time
- Always maintain service availability
- Monitor each batch
- Stop if issues detected
Success Metrics Framework
Define Metrics
Based on Define phase goals:
- What does success look like?
- How will we measure it?
- What's the baseline?
- What's the target?
Track Metrics
- Set up analytics
- Create dashboards
- Schedule reviews
- Alert on anomalies
Analyze Metrics
- Review regularly (daily, weekly, monthly)
- Look for trends
- Compare to targets
- Identify opportunities
Act on Metrics
- Prioritize improvements
- Run experiments
- Implement changes
- Measure impact
User Feedback Channels
Quantitative Feedback
- Usage analytics
- Conversion funnels
- A/B test results
- Survey scores (NPS, CSAT)
- Performance metrics
Qualitative Feedback
- User interviews
- Support tickets
- User testing sessions
- Social media mentions
- Direct user feedback
Combining Insights
- What are users doing? (quantitative)
- Why are they doing it? (qualitative)
- What should we change?
- How do we validate?
Continuous Improvement Cycle
- Monitor: Observe metrics and gather feedback
- Analyze: Identify patterns and opportunities
- Prioritize: Decide what to improve first
- Design: Create solution approach
- Develop: Build the improvement
- Deliver: Release to users
- Repeat: Continue the cycle
Communication Plan
To Users
- Release notes highlighting new features
- How-to guides for new functionality
- Announcement of changes
- Support resources available
To Stakeholders
- Deployment summary
- Metrics and KPIs
- User feedback summary
- Next steps and roadmap
To Team
- Deployment retrospective
- Lessons learned
- Celebration of success
- Planning for next iteration
Incident Response
Detection
- Automated alerts trigger
- User reports issue
- Monitoring shows anomaly
- Scheduled health check
Assessment
- Determine severity
- Identify affected users
- Understand root cause
- Estimate impact
Response
- Communicate to stakeholders
- Implement fix or rollback
- Monitor resolution
- Validate fix works
Post-Mortem
- Document what happened
- Analyze root cause
- Identify improvements
- Update processes
Support & Maintenance
Support Readiness
- Documentation complete
- Support team trained
- FAQ prepared
- Escalation process defined
- Response time targets set
Ongoing Maintenance
- Monitor system health
- Apply security patches
- Update dependencies
- Address technical debt
- Optimize performance
User Training
- Create user guides
- Conduct training sessions
- Provide video tutorials
- Offer office hours
- Build knowledge base
A/B Testing Best Practices
When to A/B Test
- Uncertain about best approach
- Want to validate hypothesis
- Optimize conversion
- Reduce risk of changes
What to Test
- Feature variations
- UI/UX changes
- Content variations
- Workflow changes
How to Test
- Define hypothesis
- Create variations
- Split traffic randomly
- Measure metrics
- Analyze results
- Implement winner
Performance Optimization
Monitor Performance
- Page load times
- API response times
- Database query performance
- Resource utilization
Identify Bottlenecks
- Profiling tools
- Log analysis
- User feedback
- Monitoring alerts
Optimize
- Code optimization
- Database indexing
- Caching strategies
- CDN usage
- Asset optimization
Validate
- Measure improvements
- Compare to baseline
- User experience impact
- Cost implications
Transition to Next Iteration
After initial deployment:
- Review success metrics
- Gather user feedback
- Identify improvement opportunities
- Prioritize next features
- Start new Discovery for major changes
- Continue iterative improvements
The cycle continues - return to Discovery for new opportunities or continue iterating in Deliver phase.
Deployment Checklist
Pre-Deployment
- [ ] All tests passing
- [ ] Staging environment validated
- [ ] Security review completed
- [ ] Performance benchmarks met
- [ ] Rollback plan prepared
- [ ] Monitoring configured
- [ ] Communication drafted
- [ ] Support team trained
During Deployment
- [ ] Follow deployment plan
- [ ] Monitor for errors
- [ ] Verify critical paths
- [ ] Check performance metrics
- [ ] Validate data integrity
Post-Deployment
- [ ] Confirm deployment success
- [ ] Monitor error rates
- [ ] Check success metrics
- [ ] Gather initial feedback
- [ ] Document any issues
- [ ] Communicate to stakeholders
Resources & References
Deployment:
- The Phoenix Project, by Gene Kim
- Continuous Delivery, by Jez Humble
- Site Reliability Engineering (Google)
Monitoring:
Incident Response:
Metrics & Analytics:
- Lean Analytics, by Alistair Croll and Benjamin Yoskovitz
- Measure What Matters (OKRs), by John Doerr
- Google Analytics Academy
Remember: Deployment is not the end - it's the beginning of learning from real users. Monitor closely, respond quickly, and iterate continuously. Success is measured by user outcomes, not just technical deployment.