Splitting cost of Elastic Pool per client

We have a situation where we are hosting a database per client in an Elastic Pool. We are trying to determine how to fairly charge the customers based on how much they use the databases.

Whatever the metric it needs to be something we can pull per database per billing period.

• What would be some appropriate metrics to consider using for each client database?
  • DTU, vCore utilization over the month per database?
  • Avg CPU utilization per month?

Any suggestions, ideas would be greatly appreciated. I hope this makes some sense. Please feel free to ask clarifying questions and I'll provide as much feedback as I can.

I'm agree with your idea to amortize costs according to different purchase models.

1. For DTU-based purchasing model:
   Average DTU utilization

2. For vCore-based purchasing model:
   Average vCore utilization