In roughly 21 of the 30 commitments reviewed, the floor required to reach the deeper threshold sat above conservative real usage, and shortfall gave the extra discount back
Overconsumption bills at your discounted rate. Underconsumption does not reduce what you owe, and that asymmetry is what decides the effective rate.
Prepared by Redress Compliance · August 19, 2026 · AWS commitment engagements. 25 to 35 reviewed, 2024 to 2025.
Executive summary
Floors were oversized by 15 to 30 percent above conservative real usage in order to reach a deeper threshold, across roughly 25 to 35 Enterprise Discount Program commitments reviewed in 2024 and 2025.
Shortfall liability ran 10 to 20 percent of the total commitment where the ramp chased the threshold, and gave back 30 to 60 percent of the extra discount it bought.
Savings plans and the EDP overlapped on the same workloads in 2 out of 5 stacks before modeling, which inflates the floor by counting discounted workloads at full rate.
A rung you fill beats a deeper rung you reach for and miss. Size the floor to net usage after savings plans and negotiate flexibility for upside instead of locking it in.
How do EDP commit levels set the discount?
Discounts are stepped by the size of the total spend commitment over the term, and a larger commitment unlocks a deeper threshold. AWS publishes its rate card and its commitment based programs on the AWS pricing pages.
The EDP is the negotiated layer above them. The threshold is a ladder, and each rung costs more commitment.
The question is never how deep the discount goes. It is whether your usage fills the rung you are reaching for.
Three terms that decide the outcome
- Commitment: the total spend you guarantee over the term.
- Threshold: the discount rung that commitment unlocks.
- Shortfall: the liability if usage falls below the floor.
Observed bands by annual commit
| Annual committed spend | Typical 3 year discount | Typical 5 year discount |
|---|---|---|
| $1M to $3M | About 6 to 10 percent | About 8 to 12 percent |
| $3M to $10M | About 10 to 14 percent | About 12 to 16 percent |
| $10M to $25M | About 14 to 18 percent | About 16 to 20 percent |
| $25M to $50M | About 18 to 23 percent | About 20 to 25 percent |
| $50M and above | About 22 to 28 percent | About 25 to 30 percent |
Ranges are observed buyer side outcomes from advisory benchmarks, not rates AWS publishes. Every agreement is confidential and deal specific.
Moving a three year term to five years typically adds about 4 to 6 points. Weigh that against the extra years of growth floor lock in.
Why is the shortfall math the load bearing part?
Because it is asymmetric. Overconsumption is billed at your discounted rate, but underconsumption does not reduce what you owe.
AWS bills the gap between actual usage and the committed floor at year or term end. On a $10M per year, three year commit of $30M total at a typical 12 percent discount, a year one actual of $7M leaves a $3M shortfall invoiced in full.
That is coverage you paid for and never used. A buyer who over commits to reach a deeper headline tier can end up with a higher effective rate than a smaller, better utilised commit.
Model the effective rate, not the headline
Divide total contracted dollars by realistically usable consumption. Right size the commit to a defensible 80 to 90 percent utilisation forecast rather than stretching to hit the next tier.
Reserve tier jumping for cases where a signed multi year migration pipeline de risks the number. Ask AWS for its best number first to bracket the range before you counter.
Commit to the spend you will actually make after every other discount, not the gross number that reaches the next threshold. The deeper rung is a saving only if your usage stands on it.
- Your commitment sized against usage after savings plans, not gross spend
- Shortfall exposure per year at the ramp on the table
- The effective rate on each rung, so the deeper one has to earn it
How do savings plans stack under an EDP?
Savings plans and reserved instances reduce the rate on specific usage, while the EDP discount applies more broadly. They are designed to coexist.
The interaction must be modeled or you double count the same workload. A workload covered by a savings plan is already discounted, and counting it again at full rate toward the EDP floor overstates how much commitment you need.
Model the net position rather than the gross. AWS describes the programs on its savings plans pages and its reserved instance pages.
How the mechanisms compare
| Structure | Discount basis | Flexibility | Best fit |
|---|---|---|---|
| On demand | None | Highest | New or volatile workloads |
| Savings plans | Usage commitment | Medium | Steady compute baseline |
| EDP, one year | Spend commitment | Medium | Growing but uncertain estates |
| EDP, three year | Deeper spend commitment | Lowest | Stable, predictable estates |
The floor is set in the contract, but the consumption that fills it is reported in the billing console. The two are modeled together far less often than the deal cadence assumes.
How does Marketplace spend count now the rules changed?
This is the most consequential recent change and it moves against buyers. Historically, eligible SaaS bought through Private Offers retired commitment: 50 percent before 2022, then 100 percent from 2022 subject to a cap of 25 percent of commitment.
Teams used third party SaaS to burn down commit they could not consume organically. Effective 1 May 2025, AWS tightened qualification.
Only SaaS products hosted entirely on AWS retire commitment. Products running partly on other clouds or on premises no longer count, and eligibility is signalled through listing designations on the Marketplace.
The nuance that is missed most often
Marketplace spend that does count still burns commit at full value and typically receives no EDP discount. A $1M qualifying purchase retires $1M of commit but costs $1M, against roughly $800K for $1M of directly consumed services at a 20 percent discount.
Re audit every vendor before renewal
- A tool that counted in 2024 may not in 2026, silently increasing shortfall exposure.
- Test each vendor against the entirely on AWS rule rather than against last cycle's assumption.
- Get written confirmation that a specific Private Offer retires commitment before you sign.
Can you cap the shortfall and structure the ramp?
Terms run one to five years and standard language requires each year's commitment to be no lower than the prior year. That is a built in growth floor.
The negotiable lever is the ramp. A back weighted curve sets a lower year one number with steeper increases later, matching the reality that migrations land late.
On a $30M three year commit, a flat ramp is $10M, $10M and $10M. A back weighted ramp of $7M, $10M and $13M carries the same total and headline discount but roughly $3M less at risk in the highest uncertainty first year.
Where the common advice on AWS EDP is wrong
The common advice is to reach for the deepest threshold by committing the largest spend you can defend. We disagree.
In roughly 21 of the 30 commitments reviewed in 2024 and 2025, the floor needed to hit the deeper threshold sat above conservative real usage, and the shortfall liability gave back 30 to 60 percent of the extra discount.
The buyer side move is to size the floor to net usage after savings plans, target a threshold you are confident you will fill, and negotiate flexibility for upside rather than locking it in.
Between the first proposal and the modeled net position.
Commitment set above net usage to reach a deeper rung.
Through shortfall, where the ramp chased the threshold.
The AWS commitment sizing brief
The utilisation forecast, the ramp structures, and the Marketplace qualification test before you sign.
Get the brief →What 25 to 35 AWS commitments showed
Across the Enterprise Discount Program commitments reviewed in 2024 and 2025, the deepest discount threshold was rarely the best outcome.
The three patterns that recurred
- Committed floors were set 15 to 30 percent above conservative real usage in order to reach a deeper threshold.
- Savings plans and the EDP overlapped on the same workloads in 2 out of 5 stacks before modeling.
- Shortfall liability ran 10 to 20 percent of the total commitment where the ramp chased the threshold.
The pattern is consistent enough to plan against. Size the rung to the usage you can defend, then spend the negotiation on flexibility for upside.
Band by band detail sits in our private pricing discount bands analysis, and the wider approach in negotiating with AWS. The equivalent commitments elsewhere are covered in Azure MACC and Google Cloud PPA. The negotiation sequence itself is walked through in episode five of the AWS negotiation series.
Research briefingFive things to settle before you sign an EDPThe floor, the ramp, the shortfall clause, the Marketplace test, and the flexibility you have to ask for by name.
Your first five moves
- Open the cycle 120 days before renewal, because every model below takes weeks to build and defend.
- Model net usage after savings plans and reserved instances, then size the floor to that number rather than to gross spend.
- Forecast to a defensible 80 to 90 percent utilisation and pick the rung that forecast fills, not the one below your ambition.
- Re audit every Marketplace vendor against the entirely on AWS rule and get written confirmation before signing.
- Negotiate a back weighted ramp so the least certain year carries the smallest floor.
Frequently asked questions
How do EDP commit levels set the discount?
Discounts step by the size of the total spend commitment over the term. A larger commitment unlocks a deeper rung, and each rung costs more commitment.
Is the deepest threshold the best outcome?
Rarely. In roughly 21 of the 30 commitments reviewed, the floor for the deeper threshold sat above conservative real usage and shortfall gave the extra discount back.
How does shortfall actually bill?
AWS bills the gap between actual usage and the committed floor at year or term end. Underconsumption does not reduce what you owe, unlike overconsumption which bills at your rate.
What does that look like on real numbers?
On a $10M per year three year commit at a typical 12 percent discount, a year one actual of $7M leaves a $3M shortfall invoiced in full.
Do savings plans stack with an EDP?
Yes, they are designed to coexist. The risk is double counting: a workload already covered by a savings plan should not be counted at full rate toward the floor.
What utilisation should you forecast to?
A defensible 80 to 90 percent. Right size the commit to that rather than stretching to hit the next tier on an optimistic number.
What changed for Marketplace spend?
From 1 May 2025 only SaaS hosted entirely on AWS retires commitment. Products running partly on other clouds or on premises no longer count.
Does qualifying Marketplace spend get the discount?
Typically no. It burns commit at full value, so $1M retires $1M but costs $1M against roughly $800K for directly consumed services at 20 percent.
Can the ramp be negotiated?
Yes, and it is the main structural lever. A back weighted ramp of $7M, $10M and $13M carries the same $30M total with roughly $3M less at risk in year one.
What does a longer term buy?
Moving three years to five typically adds about 4 to 6 points of discount, against extra years of growth floor lock in.
