All Case Studies

Case Study

A SaaS platform saved $590,909 on AWS in a year. 25% more than their own Savings Plans.

They were already good at this. They bought their own Savings Plans and Reserved Instances from internal analysis, and it was working. Handing commitment buying to Autopilot saved an extra $118,320 on the same workloads, with nothing migrated and no long-term contract.

SaaS Automotive

$590,909

saved

59.6%

average discount off on-demand

+25%

versus their own strategy

1 month

commitment

April 2025 to April 2026. Single customer account.

The Challenge

A SaaS platform serving the automotive industry was managing AWS rate optimisation by hand, buying Savings Plans and Reserved Instances from internal analysis.

It worked. They were saving real money and their team knew what they were doing. But commitment management across a large EC2 and RDS estate is a continuous job, and doing it well means constantly choosing between over-committing, which wastes money, and under-committing, which leaves it behind.

They wanted the savings without the long-term lock-in, and without an engineer spending part of every month on purchasing decisions.

The Approach

Autopilot took over commitment buying in April 2025.

Instead of broad Savings Plans, it uses Reserved Instances matched to the exact instance type, size and availability zone actually running, bought and sold on the AWS Reserved Instance Marketplace.

The term lengths are deliberately split. Across twelve months Autopilot made 93 purchases covering 1,046 instances. 82 of them ran one to twelve months, bought short for the part of the estate that moves. Eleven ran longer, from thirteen months to three years, for the stable floor that was never going anywhere: 346 r5.xlarge on a thirteen-month term, 40 r5.large and 9 r5.4xlarge on three years, 37 r5a.large on twenty-seven months. Those eleven hold 474 of the 1,046 instances and most of the savings still running today.

Glassity holds all of it. The customer's own exposure is one month, cancellable on 30 days notice, and they keep on-demand flexibility throughout. When a workload changes, the position gets sold back rather than stranded.

The Results

$590,909

saved over twelve months

59.6%

weighted average discount off on-demand pricing

$118,320

more than their own Savings Plans strategy would have delivered

Their existing approach was already achieving a 47.6% discount. Autopilot reached 59.6% on the same workloads, which is 12 percentage points higher and 25% more money.

Nothing was migrated, nothing was renegotiated, and their existing commitments were left in place.

Why It Worked

Savings Plans are broad by design. One commitment covers a wide family of instances, and the discount is set by the breadth of that coverage. That makes them simple to buy and imprecise by nature.

Reserved Instances can be far more precise, but only if someone keeps matching them to what is actually running. Done by hand that is a full-time job nobody has. Done continuously it was worth 12 percentage points.

Three things a Reserved Instance can do that a Savings Plan cannot:

  • • Match the exact instance type. Not a family, the specific type and size in the specific availability zone.
  • • Buy short terms on the Reserved Instance Marketplace. Commitments of a few months rather than a minimum of one year.
  • • Sell the position back. When a workload changes, the reservation is resold rather than left to expire unused.

That third one is why the customer's commitment is one month while the underlying reservations run for up to three years. Glassity carries the term and can exit it.

What the customer kept

  • ✓ On-demand flexibility. Instances can be resized, migrated or retired without stranding a commitment.
  • ✓ A one-month commitment. Glassity carries the multi-year risk. Cancel on 30 days notice.
  • ✓ Their engineering time. No analysis, no purchasing decisions, no monthly review.
  • ✓ Their existing Savings Plans. Nothing was cancelled.
  • ✓ A success fee only. 10% of savings verified in their own AWS bill.

Frequently asked questions

They can, but only with continuous management. Savings Plans apply broad coverage across instance families at a discount set by that breadth. Reserved Instances can be matched to a specific instance type, size, availability zone and term, capturing a deeper discount and allowing regional price differences to be used. In one twelve-month customer deployment, moving from manually purchased Savings Plans to precision-targeted Reserved Instances raised the discount from 47.6% to 59.6% off on-demand pricing.

Discounts depend on commitment type and how closely commitments match real usage. AWS advertises up to 72% with Savings Plans; most organisations achieve closer to 23% in practice. A SaaS platform using Glassity Autopilot achieved a 59.6% weighted average discount off on-demand pricing across EC2 and RDS over twelve months, saving $590,909.

Yes, through a third party that holds the commitment on your behalf. Glassity Autopilot purchases and rebalances Reserved Instances itself, so the customer commits for one month and can cancel on 30 days notice while still receiving discounted rates. The multi-year risk sits with Glassity rather than the customer.

Savings begin appearing in the AWS bill in the first billing cycle after commitments are purchased. In one customer deployment beginning April 2025, savings accumulated to $590,909 over the following twelve months at a 59.6% average discount.

A stranded commitment is a Reserved Instance or Savings Plan that no longer matches the workload it was bought for, usually after instances are resized, migrated or retired. The customer keeps paying for unused capacity with limited options to exit. Short commitment terms and continuous rebalancing avoid it by ensuring coverage expires or moves before the workload does.

See what's available on your own bill

Free savings assessment. Fifteen minutes, read-only access, no card and no call required.

Get my free savings assessment Or book a call with Ernesto