Stage Frameworks

Explicitly banned (money pits at this stage):

• EKS ($73/mo for nothing), NAT Gateway ($32/mo for nothing)
• Multi-AZ RDS, ElastiCache, AWS Config
• Custom CloudWatch metrics ($0.30/metric/month adds up)
• Service mesh, multi-region, dedicated CI/CD pipeline

You don’t need this yet: observability beyond basic alarms, load testing, DR plan, staging environment.

Seed (2-5 people, <1K users, $500K-$2M raised)

Principle: Every dollar on infra is a dollar not spent on product. Prove PMF first.

Key constraints:

• On-demand pricing only — don’t commit to provisioned anything
• One pipeline, one environment (prod). Staging is optional luxury.
• Single AWS account is still fine (use tags)
• No Savings Plans — you don’t know your baseline yet
• Cost target: $100-500/month, credits should cover 12-18 months

Now add (but keep simple):

• SQS for async work (decouple heavy processing from API)
• EventBridge for internal event routing
• IaC (CDK or Terraform) — one stack, not microstack per service
• CloudWatch alarms on: 5xx rate, latency p99, DynamoDB throttles

Still avoid: EKS, Redshift, Step Functions Standard ($0.025/1K transitions), multiple AWS accounts, Security Hub.

Trigger to graduate: >1K active users, raising Series A, team >5, first incident that costs you users.

Series A (5-15 engineers, 1K-100K users, $5M-$20M raised)

Principle: Harden what works. Add reliability and observability — don’t re-architect.

Key changes:

• Multi-AZ for all production workloads
• Separate AWS accounts (prod vs non-prod)
• No console changes to production — everything in IaC
• On-call rotation (minimum 4 people for sustainability)
• Weekly 5-minute cost review
• Cost target: $1K-10K/month

Now appropriate:

• Savings Plans after 3+ months stable data (1-year Compute SP, No Upfront, cover 50-70% baseline)
• ECS with EC2 for steady-state workloads where Fargate cost exceeds EC2 + SP
• RDS Proxy for Lambda → RDS connection pooling
• WAF if handling sensitive data or seeing abuse
• SOC2 prep (GuardDuty + Security Hub + Config)
• Blue/green or canary deploys

Trigger to graduate: 100K+ users, >15 engineers, need dedicated platform team, multi-region requirements.

Series B+ (15+ engineers, 100K+ users, dedicated platform team)

At this stage, standard AWS best practices apply. Startup-specific filtering dissolves.

What persists from startup thinking:

• Speed still matters — platform team enables, doesn’t gate
• Track cost per customer (investors care about unit economics)
• Still prefer managed services — your moat is product, not your K8s cluster
• Credits may still apply through Series B — check expiration dates

Stage Transition Checklist

Pre-Revenue → Seed

• Not done: Real users (not just friends testing)
• Not done: Hit a real limitation of pre-revenue architecture (not just “feels hacky”)
• Not done: Credits/funding for 12+ months of infra

Seed → Series A

• Not done: PMF evidence (retention, revenue, growth rate)
• Not done: At least one user-impacting incident
• Not done: Can afford $1K+/month in infra without stress
• Not done: Team growing and needs shared standards

Series A → Series B+

• Not done: Need dedicated platform/SRE team (not just one person part-time)
• Not done: Multi-region is a real requirement, not nice-to-have
• Not done: Compliance demands formal controls