Case studies
What Reserver finds in real AWS accounts.
Every study here reads a real account from end to end. We read its reservation ledger, usage history, and fleet. We price the findings at the rates the account pays. The customers asked us to anonymize each study. The fleets range from a six-instance startup to an AWS Organizations footprint with dozens of accounts.
RDS · ElastiCache · EC2 · 40 accounts · AWS Organizations
Real account40 AWS accounts, one blind spot: $250k/yr found
A real AWS Organizations footprint across 40 member accounts. Every account looks fine alone, and the money hides in the gaps between them. We priced three problems at the accounts' own rates. They are an org-wide renewal lapse, coverage drift that nobody could see whole, and reservations stranded by a re-org.
RDS · ElastiCache · EC2 · 6 regions
Real accountThe $2.4M account that did everything right, and still leaked $60k a year
A multi-region SaaS platform with twelve years of reservation discipline. It bought 550 reservations since 2014 and has coverage most teams would envy. We read the whole ledger. The money leaked between manual reviews, and through a purchasing habit that nobody checked against the data.
RDS · ElastiCache · EC2 · 1 account
Real accountSix instances, one missed renewal: $2,195 found
A real, lean, single-account fleet, the size a founder or a small platform team runs. We priced it at the account's own AWS on-demand and Reserved Instance rates for the exact classes named. The customer asked us to anonymize it. A missed renewal and a never-reserved tier both have a price.
More studies will come as design-partner accounts come online. Do you want us to read your account the same way, and price every gap, lapse, and term? Join the waitlist and be next.
Across all three
Three fleet sizes, the same three failures.
Six instances, a multi-region SaaS platform, and 40 AWS accounts fail in the same three ways. Only the scale changes. Each row below quotes the published figures of each study.
Nobody owned the renewal date.
A Reserved Instance term ends on a fixed date, and AWS never renews it. Someone must track that date. If nobody does, the discount ends without warning.
- Startup23 days passed before anyone caught the lapse.
- GrowthA median gap of 8 days ran between an expiry and its repurchase, across 550 reservations.
- EnterpriseOne reservation term lapsed on all 40 accounts in the same week.
Coverage slid down without anyone noticing.
Coverage rarely drops all at once. It falls in small steps between reviews. Every step bills at the on-demand rate.
- StartupThe database held a reservation. The rest of the fleet did not, for a starting coverage of 33%.
- GrowthCoverage slides between manual reviews, one instance at a time. No single review catches the whole drift.
- EnterpriseOrg-wide coverage measured 71% at scan time. Each of the 40 accounts looked correct alone.
The fleet changed shape. The reservations stayed where they were bought.
An instance generation retires, or a re-org moves a workload to a different account. A reservation is tied to a class and a region. A zonal one, or one in an account with sharing off, is also tied to its account. None of these follow the fleet on their own.
- GrowthFour renewal rounds bought about 35 reservations on a class the team was about to retire.
- EnterpriseA re-org moved a workload to a different account. The 45 reservations bought against it stayed behind, in the account it left.
Which fleet size are you?
Find the plan that fits your fleet.
Every figure below comes from the same model as the pricing page and the savings calculator. Find the study closest to your fleet. The plan below it is the match.
Watch tells you before it lapses. You click buy. A missed renewal here runs about $92. See the Watch plan.
Autopilot checks coverage every day and flags every renewal before it lapses, one flat price. A missed renewal here runs about $951. See the Autopilot plan.
One coverage view across every account you connect. Flat software price, never a cut of savings. A missed renewal here runs about $6,904. See the Scale plan.
How we read an account
The same method, every time.
Every study above used the same four steps that Reserver runs on a connected account. No study used a special method.
- Connect a read-only IAM role. The account grants Describe access only. Reserver never asks for a key or a password. See the full security model.
- Scan the running fleet. RDS, ElastiCache, and EC2 instances, across every connected account and region.
- Scan the reservation ledger the account already owns. Every Reserved Instance, its class, term, and end date.
- Match the two and price the gap. At the account's own on-demand and Reserved Instance rates, never at list price. See how matching works.
We anonymized every study on this page at the customer's request. We withhold the name, the account ID, and any other identifying detail. The instance classes, counts, and dollar figures are the account's own. We never use list price and never blend rates across accounts.
Questions
Case studies, answered plainly.
Are these real AWS accounts?
Yes. Every study reads the fleet and reservation ledger of a real account. We modeled none of the figures and invented none.
How are the accounts anonymized?
At each customer's request, we withhold the name, the account ID, and any identifying detail. The instance classes, counts, and dollar figures stay real.
Whose AWS rates set the dollar figures?
Each study prices its findings at the on-demand and Reserved Instance rates of that account. Reserver never uses list price and never blends rates across accounts.
Do I need a large fleet before Reserver finds anything?
No. The startup study found $2,195 on six instances. The method finds gaps at any fleet size.
Can Reserver read our AWS account the same way?
Yes. Connect a read-only IAM role. Reserver then scans your fleet and reservation ledger the same way, priced at your account's own rates.
Stop overpaying for a steady fleet.
Reserver is in early access. Join the list, and we will onboard you when we have capacity. Your first recommendation lands within minutes of connecting.