One S3 configuration mistake cost us $8K in a single month.

The strange part?

We didn't suddenly start storing 10x more data.

The problem was how we were storing it.

A Rails SaaS application was generating large numbers of files, logs, exports, and temporary objects.

Storage kept growing.

But the real cost came from objects that were no longer needed but were never cleaned up.

We reviewed the S3 usage and found:
* Temporary files staying for months
* Old application exports never expiring
* Logs stored in expensive storage classes
* No lifecycle policies for different object types

So we changed the storage strategy:
* Added S3 lifecycle rules
* Moved older data to cheaper storage
* Automatically deleted temporary objects
* Added monitoring for unusual storage growth

The next month, the S3 bill dropped significantly.

The lesson?
Cloud cost problems don't always come from using too much infrastructure.

Sometimes they come from never defining when infrastructure should stop existing.

For S3, storage architecture is also cost architecture.