Office hours: why "dollars saved" is the wrong reservation metric

The best part of this session wasn’t a toolkit feature. It was a genuine disagreement about how to measure whether your commitment discounts are actually working, and it’s a debate worth having on your own team.

An hour is an hour is an hour

One attendee argued that effective savings rate should isolate commitment effort from negotiated-discount effort, so it’s comparable across customers with different negotiated rates. Another attendee agreed with the goal but pushed the metric further: the more meaningful number is commitment coverage measured in compute-hours, not dollars. Their reasoning: a large negotiated discount can make even a well-run three-year reservation’s dollar savings look negligible, which misrepresents how well the FinOps team is actually doing. As they put it:

An hour is an hour is an hour. It’s the same regardless of what rate sheet is attached to it. That’s the thing you have to normalize on.

They pointed out that commitment-hour coverage benchmarks vary widely from one organization to the next, which is exactly why the metric needs to be normalized rather than compared at face value, and drew a direct parallel to why AI “tokens” are a questionable normalization unit for the same reason dollars fail for reservations and savings plans.

Behind the scenes: why your price sheet month has to match

On the more technical side: you can’t import July cost data with only an August price sheet. FinOps hubs uses the price sheet to reconcile savings-plan and reservation pricing at ingestion time, so the correct month has to land first. The same attendee described using Copilot to reconstruct missing historical price sheets for a couple of different customers when only the latest one was available. Separately, Roland explained the mechanics of how hubs handles corrected data: rows are tagged by billing month into Kusto extents, and a new file landing in storage swaps out the old extent for that month entirely, so a correction always re-imports cleanly.

One more thing to watch: FOCUS 1.4 is expected to add a commitment-discount eligibility field to usage data, which could eventually resolve a related ambiguity: whether a storage reservation was skipped deliberately or represents a missed opportunity. It’s real, but described as a complex nested structure rather than a simple flag, so it won’t be trivial to consume once it shows up broadly in Cost Management data.

If you’ve never questioned whether your own savings-rate metric is telling you the truth, this is a good excuse to start. Bring your take to the next office hours, every other Wednesday.