K8s integration sounds useful, there are also many startups in this area, and it's trivial to implement, however please consider that just a workload is requesting 2 CPUs, does not mean you can divide the underlying EC2's price per vCPU or Memory. There is significant overhead running K8s, log collectors, agents, wasted space, and spot instance pricing etc. please be more insightful for that, not just divide by request/limits. Combining these with overall AWS costs would make more sense. However as I said, we tag by service or cost centers, so I would want to associate my Pods and its SQS queue cost together.

Chargebacks in AWS is very complex, i.e in cost explorer net amortized cost vs unblended cost is hard to understand for an average user. It'd be nice if you can have some clarifications on it, what does even amortized mean etc.

We use Cost Explorer API to identify anomalies. Even basic statistics works sometimes. But there is also new native feature that is good enough for us, so we removed that. But it was a very sensitive topic, i.e our S3 usage is steadily increasing %.x per month. Trend & seasonality analysis is a must.

Sorry for my quick and dirty English. This is a hot topic for me and I get excited.