The Well-Architected Framework gets described as best practices, which makes it sound optional. It is closer to a list of questions AWS has watched people answer badly, written down. There are six pillars — sustainability joined the original five at the end of 2021, which is why half the articles online still say five.
Operational excellence
Can you deploy without ceremony, and do you find out about problems before your customers tell you? This pillar is about infrastructure as code, small frequent changes, runbooks that exist in writing, and post-incident reviews that produce a change rather than a feeling.
Security
Identity as the perimeter, least privilege, encryption in transit and at rest, and an audit trail you cannot quietly delete. Most of what a review flags here is unremarkable: root account without MFA, access keys older than the laptop they sit on, a security group open to the world on port 22 because someone needed to debug something in March.
Reliability
What happens when a component fails, and have you tested that answer? Multi-AZ deployment, health checks, automated recovery, and backups that someone has actually restored from. An untested backup is a hypothesis.
Performance efficiency
Are you using the right resource for the job, and do you revisit that as AWS releases new instance families and managed services? This pillar ages faster than the others — an architecture that was efficient in 2021 may now be a Graviton migration and a managed service away from being much better.
Cost optimisation
Are you paying for what you use, and does anyone own the number? Right-sizing, commitment coverage, storage lifecycle, and shutting down what nobody needs. The framework treats cost as an architectural property rather than a finance problem, which is the correct way round.
Sustainability
Minimising the energy and hardware footprint of what you run. In practice it overlaps heavily with cost — fewer idle instances, right-sized storage, efficient instance families, and workloads scheduled to run when they need to rather than continuously.
What a review actually involves
The AWS Well-Architected Tool walks through the questions per workload and produces a list of high and medium risk items. Run with a partner, it is typically a working session with your engineers rather than an audit — the value is in the discussion where somebody says "we have never tested that failover" out loud.
The output is a prioritised list of risks, an owner against each, and a plan. Not a certificate. Anyone selling you a Well-Architected badge has misunderstood the exercise.