And thanks @Bryan Ross, you're entirely right - chargeback is best, but can be hard to calculate sometimes across all platform services, so at a previous place we used a percentage of cloud spend to fund the platform team.
It always goes back to the metrics, and being able to translate metrics to money. Cost of delay (time to market) is a good one, but even that implies the product organisation can provide that information, and I very strongly doubt they can do so reliably, and I've unfortunately never worked in an organisation where they could.
I've never really found a way to directly translate platform or DX metrics into equivalent money amounts in a way that's acceptable to finance AND that's convincing to myself, it's always an argument based on very flawed and shaky assumptions. That's the nature of the field and I doubt it's possible to do rigorously, but would be great to have a standard approach to these things.
I usually fall back to things like cost of downtime which I think most organisations have a good handle on, and also rely on estimates of engineer-days saved (by the people who would have been doing the work) when improving efficiency to highlight cost avoidance. Given for the latter everyone is often just guessing more or less, it becomes an interesting exercise in story-telling 🙂