TL;DR: Size commitments to median usage over 12 months, not the peak month. A commitment is paid every hour of every month, so sizing to the peak means paying peak prices for median usage - the exact opposite of what commitments are for. Measure the current commitment's monthly utilization to see the overcommitment, then fix the sizing rule.

```text
agent overcommitted to savings plans  -  it sized the commitment on peak-month usage instead of the median
```

1. Measure the commitment's actual utilization: covered usage hours divided by committed hours, per month since purchase. Expected: high utilization in the peak month and well under 100 percent in normal months - the overcommitment made visible.
2. Compute what the commitment should have been: the median monthly usage, or the 40th percentile if you want a safety margin - not the peak. Expected: a target commitment noticeably lower than what was bought.
3. For the existing overcommitment, check exchange options to downsize the commitment, and shift flexible workloads onto the stranded capacity where possible. Expected: a plan that either shrinks the commitment or soaks up the excess with real workloads.
4. Change the agent's sizing rule: size commitments to median usage over 12 months, and flag any recommendation above the 60th percentile for human review. Expected: future commitments stay utilized year-round instead of only in the peak month.

## Use this when
- A savings plan commitment sits underutilized outside peak months
- The agent sized a commitment from the highest-usage month on record
- You need a sane sizing rule (median, not peak) for future commitments

## Not for this skill when
- Usage is flat and peak equals median - the sizing was fine
- The peak is the new normal (sustained growth) - resize the model, not the rule
- The commitment is nearly expired - let it lapse and size the renewal correctly

## Variant phrasings
- savings plan overcommitted sized on peak usage
- commitment utilization low outside peak month
- how to size savings plan on median usage
- agent bought too much savings plan commitment

## Why it happens
Peak months are memorable, and recent peaks feel representative. An agent looking at 12 months of data sees the peak as what we need and sizes to it. But a commitment is paid every hour of every month - including the eleven quiet ones. Sizing to the peak bakes waste into the contract from day one.

## Edge cases
- Growing workloads: the median of the past year understates next year. Blend the historical median with the growth forecast.
- Multiple commitments with different start dates complicate the utilization math. Measure utilization per commitment, not in aggregate.
- Seasonal businesses may genuinely want peak coverage - but that is a deliberate choice with a known waste budget, not an agent default.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_nhK65t3wTv7jCSIFI5we3Q