Case study — lean team
Six instances. One missed renewal, one quiet leak — together, more than a year of the tool that stops both.
A real production account: a lean, single-account AWS footprint. We read its reservation ledger, usage history, and running fleet. Every number below is measured, and priced at the rates this account actually pays for the exact classes below, us-east-1, mid-2026 — anonymized at the customer's request. See the method notes at the bottom.
Finding 01 — the renewal nobody owned
They bought right. Nobody put the date on a calendar.
At product-market fit, this team reserved its db.r6g.large Multi-AZ pair for a year — a good, deliberate call. A year later the term quietly ended. Nobody had a process for renewal dates, so the pair kept running — just at on-demand rates instead. It took 23 days for a routine bill check to notice.
Eighty-some dollars won't break a startup. But nobody was watching, and there was no reason it would have stopped at 23 days — our flagship study's own worst case ran 309 days on one instance before anyone noticed. Reserver's expiry radar flags the renewal date before the term ends, free on Watch. Renewal autopilot rebuys the same day a term expires, on Autopilot. Either one closes this category to zero, permanently.
Finding 02 — the instances that were never in the plan
The database got reserved. The rest of the fleet didn't.
Three m6g.large app-tier instances and one cache.r6g.large Redis node have run at on-demand rates since the day they went live — 18 months and counting. Nobody decided against reserving them. Nobody decided anything; they just never came up at a review, because there wasn't one.
None of this required a mistake — just the absence of a process. Reserver's age rule ("running longer than 30 days, uncovered → reserve it") would have ticketed all four instances in their first month on the fleet. On a fleet this size, that rule is the entire difference between watching it happen and catching it before it starts.
The bill
$2,195 found on a fleet you could count on one hand.
That's already more than a year of Autopilot ($1,788) — on a fleet of six instances. Watch, the tier built for exactly this size of account, costs nothing to run. There's no minimum reservation count where finding this stops being worth it.
Method notes — for the skeptical reader (we hope that's you)
- Anonymized, not invented — this is a real single-account fleet; the account and any identifying details are withheld at the customer's request, but every instance class, count, and dollar figure is its own.
- Every dollar figure is priced at the account's own AWS On-Demand and 1-year Reserved Instance (No Upfront) rates, US East (N. Virginia), mid-2026, for the exact classes named: db.r6g.large (RDS MySQL, Multi-AZ), cache.r6g.large (ElastiCache Redis), m6g.large (EC2 Linux).
- On-demand-minus-RI premiums use the No Upfront payment option specifically, because it carries the smallest discount of the three RI payment options — that keeps every premium in this study conservative, not best-case.
- The 23-day renewal gap and the 18-month never-reserved window are measured from the account's own history.
- No right-sizing, Spot, or workload changes are assumed anywhere on this page. This is purchasing mechanics only, on the fleet exactly as described.
Stop overpaying for a steady fleet.
Reserver is in early access. Join the list and we'll onboard you as capacity opens up — first recommendation within minutes of connecting.