Reserver internal business case
Reserved Instance coverage: the numbers to send upstairs
The situation
- The gap between reviews. Coverage drifts in the weeks after a quarterly review closes. New instances launch, old reservations expire, and nothing catches the difference until the next cycle.
- The renewal that slips a few days. Even disciplined teams miss a renewal date by days here and there, and each slip bills the gap at on-demand rates until someone repurchases.
- The purchasing habit that outlived its logic. A buying pattern gets set early and never gets re-derived from what the fleet actually looks like now.
What it's costing
Zoom in on one instance: at AWS's published reserved-vs-on-demand spread for a larger database class, a single unreserved instance left running 14 days past its term bleeds about $188 on its own. The figure above is a blended average across every reservation a fleet this size carries, not one instance.
What we're asking for
Autopilot is a flat monthly line item, never a cut of what it saves. Here's what it costs against what it returns.
What it returns
One prevented missed renewal alone pays for about 6 months of Reserver.
How we know
Conservative and disclosed as estimates, not a quote: 65% of typical compute spend is reservable capacity, discounted at a blended 45%. That's below the higher published ceilings for any single term, since this blends across a whole fleet. Reserver shows the exact math, per reservation, once you connect an account.