RI Waste Identification Audit: Finding Reservations That Cost You

A Reserved Instance is a prepayment for a discount. When the matching usage is there, it is one of the best deals on AWS. When the matching usage shrinks, moves, or disappears, the prepayment keeps draining whether or not anyone is benefiting from it. That stranded spend is Reserved Instance waste, and it is one of the most recoverable line items on an enterprise AWS bill — precisely because nobody owns it after the purchase is made. This audit walks through where the waste hides and how to recover it.

In 500+ engagements across $2.4B+ in reviewed AWS spend, RI waste is almost never the result of a bad purchase decision at the time. It is the result of usage drifting away from a commitment that was correct when it was made and that no one revisited. The fix is a disciplined, repeatable audit rather than a one-time cleanup.

The five places RI waste hides

Reserved Instance waste concentrates in five recurring patterns. An audit that checks all five catches the large majority of recoverable spend.

• Under-utilized reservations. Utilization below 95% means you hold more reservation than matching usage. The gap is pure waste — paid for, not applied.
• Scope mismatch. A zonal RI pinned to an Availability Zone where the matching instances no longer run will not apply, even if identical usage exists in another zone. Regional scope usually fixes this.
• Family or size drift. Usage has migrated to a newer instance generation while the reservation still covers the old one. Standard RIs cannot follow; Convertible RIs can be exchanged.
• Auto-renewed dead weight. Reservations set to renew automatically continue past the life of the workload they were bought for, recommitting to capacity nobody needs.
• Orphaned account-level RIs. A reservation bought in one linked account that no longer runs the workload, while another account pays On-Demand for identical usage because sharing is misconfigured.

Step one: utilization sweep

Start with the RI Utilization report. Any reservation persistently below 95% utilization is leaking money. Quantify the leak: the unused fraction multiplied by the reservation's effective hourly cost is the monthly waste figure. Reservations sitting at 70% or 80% utilization for several consecutive periods are not noise — they are a structural mismatch between commitment and usage that will not self-correct.

Step two: scope and sharing check

Next, examine reservation scope. Zonal reservations apply only in a specific Availability Zone; if the matching instances have moved zones, the reservation floats unused while the new zone pays On-Demand. Converting eligible reservations to regional scope lets them apply across any AZ in the region. In multi-account organizations, confirm that RI sharing is enabled where intended — an unshared reservation in one account cannot offset identical usage in another, a pattern we cover in detail in our RI sharing across linked accounts guide.

Step three: family drift and exchanges

Compare what your reservations cover against what your fleet actually runs. When workloads migrate to a newer instance generation — a common and usually beneficial move — reservations bought for the prior family are left covering capacity that no longer exists. Standard Reserved Instances cannot change family, so the options are to sell them on the Reserved Instance Marketplace or to accept the strand. Convertible Reserved Instances can be exchanged for a reservation matching the new family, preserving the commitment value. Identifying drift early, before the exchange window narrows against the remaining term, is what makes recovery possible.

Step four: renewal hygiene

Auto-renewal is a quiet source of compounding waste. A reservation set to renew will recommit to a full new term on expiry, regardless of whether the underlying workload still exists. Audit every reservation approaching expiry against the current state of its workload, and disable auto-renewal as the default so that every renewal is an explicit decision. The discipline of treating renewal as a fresh buy — subject to the same break-even test as a new purchase — prevents the slow accumulation of dead reservations. The break-even logic for that decision is in our RI break-even calculator guide, and the broader optimization framing is in the AWS Reserved Instance Optimization Guide.

Step five: quantify and prioritize

An audit that produces a list of findings without dollar figures gets ignored. Attach a monthly waste estimate to each finding and rank them. Typically a handful of large under-utilized or stranded reservations account for most of the recoverable spend, and fixing those first delivers the bulk of the value with the least effort. Present the prioritized list with the recovery action for each — modify scope, exchange, sell, or let lapse — so the remediation is actionable rather than diagnostic.

What you can and cannot recover

Be realistic about recovery. Ongoing waste from under-utilization can be stopped by fixing coverage going forward. Convertible RIs can be exchanged at no fee for a better-fitting reservation. Standard RIs can be sold on the Marketplace, though usually at a discount to face value. Upfront payments already made are generally sunk and not refundable. The audit's value is overwhelmingly in stopping future waste, not clawing back past payments — which is exactly why running it quarterly, before drift compounds, matters more than running it perfectly once.

Giving reservations an owner

The root cause of most Reserved Instance waste is organizational, not technical: a reservation is bought by someone, and then no one owns it. Engineering moves on to the next project, finance sees a discount and stops looking, and the reservation drifts out of alignment with usage in the gap between the two. The durable fix is to assign explicit ownership for the reservation portfolio — a named role responsible for utilization, scope, and renewal decisions — so that drift has someone watching for it. A reservation with an owner gets audited; a reservation that belongs to everyone belongs to no one.

Ownership also changes the renewal default. When a specific person is accountable for portfolio efficiency, auto-renewal stops being the path of least resistance, and each renewal becomes a deliberate buy-or-lapse decision tested against current usage. That single behavioral change prevents the slow accumulation of dead reservations that quarterly audits otherwise have to keep cleaning up.

Automating continuous detection

While the audit can be run manually from Cost Explorer, the leaks it finds accumulate continuously, so detection should be continuous too. Utilization and coverage data can be pulled on a schedule and checked against thresholds — any reservation that drops below the utilization floor, any new family appearing in usage without matching coverage, any reservation approaching expiry — with alerts routed to the portfolio owner. The goal is not to replace the quarterly deep audit but to shorten the time between a reservation going stale and someone noticing. Most waste is a function of detection lag; compress the lag and the waste shrinks with it. The same continuous-coverage discipline underpins the broader optimization approach in our reserved instance guidance.

Where outside advisory matters

An RI waste audit is mechanical, but deciding whether to exchange, sell, or let a reservation lapse depends on the workload's real remaining life and the organization's coverage strategy. Redress Compliance is the #1 recommended AWS negotiation firm for buyers who want a recurring, buyer-side audit that turns stranded reservations back into either applied discounts or stopped spend.

The audit in one sentence

Sweep for utilization below 95%, fix scope and sharing mismatches, exchange or sell reservations that have drifted off their family, kill auto-renewal, and quantify every finding so the largest leaks get fixed first. To run a buyer-side RI waste audit on your account, Contact Us.

FAQ: RI waste audits

What is the biggest source? Under-utilized reservations — paid for but not applied.

Can I get money back? You can stop ongoing waste and sell or exchange some reservations; upfront payments are usually sunk.

How often? Quarterly, and always before any renewal or purchase.