AWS cost checklist: model the drivers that actually move the bill

If your AWS bill feels “mysterious”, it usually means you are missing a driver. Most costs scale with a small set of units: hours, requests, GB transferred, and GB-month stored. This checklist helps you capture the drivers that actually move spend.

1) Define scope (avoid mixing worlds)

• Accounts and regions included
• Environments (prod vs staging vs dev)
• Time window and seasonality (baseline vs busy month)

2) Identify the top 5 line items first

• Use billing/Cost Explorer to find the top services by spend.
• For each service, write down the primary driver unit (hours, requests, GB, GB-month).
• Do not optimize everything at once; fix the top drivers first.

3) Data transfer: the most common hidden bill

• Internet egress (bytes leaving AWS)
• Cross-AZ traffic (often accidental)
• Cross-region traffic and replication
• NAT gateways and load balancers can change where transfer shows up

4) Logs and metrics: ingestion + retention + scans

• Log ingestion (GB/day) and retention (GB-month)
• Query scan costs (Insights/Athena/SIEM)
• Custom metric cardinality and dashboard polling

5) Request-based services: count requests before pricing

• CDN requests and bandwidth (origin vs edge)
• API requests and response size (requests + transfer)
• Database requests (read/write units, query scans)

If you only know RPS, convert it to monthly requests and then model cost drivers.

6) Storage: GB-month + requests + snapshots

• Object storage: GB-month and request volume
• Block storage: volume GB-month, provisioned IOPS/throughput, snapshots
• Replication: storage copies and transfer

7) Reliability incidents are also cost incidents

• Retry storms multiply requests, logs, and transfer.
• Incidents increase query scans (people search logs repeatedly).
• Capacity spikes (autoscaling) increase hours and cross-AZ patterns.

8) Validation checklist (so the model is trusted)

• Validate each primary driver from measured telemetry (logs/metrics/billing) for a representative week.
• Use two scenarios: baseline and busy month.
• Confirm unit consistency (GB vs GiB, requests units, per-hour vs per-second).
• After changes, compare against the next week of real billing and adjust assumptions.

9) Make the model operational

• Write down assumptions and owners (who updates request volume, who updates storage growth).
• Link each driver to the place you measure it (metrics, logs, billing report).
• Update monthly: keep a small set of calculators and checklists close to the team.

Sources

• AWS pricing overview: aws.amazon.com/pricing

Related guides

Related calculators

FAQ