The formula is useful, but the list of things that keep billing is what saves people from a disappointing first month. Disks, NAT gateways and idle public IPs often explain why a 64% compute saving turns into a much smaller drop on the invoice.

The RDS detail is worth repeating. A stopped instance starting itself again after seven days catches teams who assumed stopped meant stopped.

Your order of work matches what we do on client DevOps engagements: remove the resources paying for nothing first, then the disks of VMs that are not coming back, and only then agree a schedule with the people who use the environment. The first two need no meeting. The third needs a named owner and a way to wake an environment outside the window, otherwise someone disables the schedule after the first late release.

Do teams usually set the off hours window per environment or per project in Control Cloud?