The headline discount is not the effective rate, and a 15 percent headline routinely lands at 11
Every instrument in an AWS agreement is quoted as a percentage, and no two of them apply to the same base. The number that matters is the one after they interact.
Prepared by Redress Compliance · August 17, 2026 · AWS advisory. Redress Compliance advisory engagement file, 2024 to 2025.
Executive summary
A 15 percent headline routinely lands at an 11 percent effective rate. Because the headline applies to one base, Enterprise Support takes back up to 10 percent of spend, and the instruments do not overlap cleanly.
EDP discounts run 5 to 20 percent, stepped by commitment threshold. That is the whole published range. Anything quoted above it is either a different instrument or a different base.
Savings Plans and Reservations reach up to 72 percent, but only on covered compute. Which is why an estate average and a covered workload rate are entirely different numbers and should never be compared.
Marketplace burns down EDP at 100 percent credit. That is shortfall insurance rather than a saving, and treating it as a discount is the most common modelling error we see.
Five instruments, five different bases
Each is quoted as a percentage. None of them apply to the same thing, which is why they cannot simply be added.
| Instrument | Range | Applies to |
|---|---|---|
| EDP | 5 to 20 percent | Total commitment, stepped by threshold |
| PPA | 5 to 40 percent | Specific service lines only |
| Savings Plans and Reservations | Up to 72 percent | Covered compute only |
| Enterprise Support | Up to 10 percent | Taken back, off total spend |
| Marketplace | 100 percent credit | Burns down commitment, not price |
The fourth row is the one that turns 15 into 11, and it points the opposite way to the others. Four instruments reduce what you pay and one adds to it, and the one that adds is calculated on total spend rather than on a subset. Any model that sums the discounts without subtracting support produces a number the invoice will never match, which is how a well negotiated agreement still disappoints the people who approved it.
The headline applies to one base. The bill is the whole estate
An AWS agreement is not one discount. It is a stack of instruments, each quoted as a percentage, each applying to a different base, and one of them running in the opposite direction. EDP sits at 5 to 20 percent, stepped by commitment threshold. PPA reaches 5 to 40 percent but only on specific service lines. Savings Plans and Reservations reach up to 72 percent, but only on covered compute. Enterprise Support then takes up to 10 percent of spend back off the top.
Put those together and the arithmetic stops being additive. A 15 percent headline routinely lands at an 11 percent effective rate once the instrument that only touches part of the estate is applied to part of the estate, and the instrument that touches all of it is subtracted from all of it. Nothing has gone wrong in that negotiation. The number simply described a base narrower than the bill.
The comparison error this produces is worth naming, because it recurs. Savings Plans reaching 72 percent on covered compute is a genuine figure and it is not comparable to an EDP percentage, because one is a rate on a workload and the other is a rate on a commitment. Buyers who benchmark an estate average against somebody else's covered compute rate conclude they are being badly treated when they are not, and buyers who quote their own covered rate as an estate figure conclude the opposite. Both are comparing instruments that do not measure the same thing.
Marketplace deserves its own caution. It burns down EDP commitment at 100 percent credit, which is valuable, and it is not a discount. It converts spend you were making anyway into progress against a commitment, which protects you from shortfall. Treating that as a saving double counts it, because the money left the business either way. Model the effective rate across the whole estate, keep each instrument attached to its own base, subtract support rather than forgetting it, and treat marketplace as insurance. The commitment mechanics sit in the EDP pillar, the aggregate discount in the PPA pillar, and the library in the AWS practice.
- Your quote benchmarked against 500,000+ real closed deals, adjusted for size, region, and industry
- Commitment sized against measured usage rather than a forecast built to justify a rate
- Every risky clause flagged with the exact quote, the page, and the replacement language
The AWS EDP negotiation brief
The commitment sizing model, the tier thresholds, the marketplace mechanics, and the buyer side moves across the full AWS agreement.
Get the brief →Modelling the effective rate
- Model the effective rate across the whole estate, not the headline, since that is the only figure the invoice will reflect.
- Keep each instrument attached to its own base, because EDP, PPA, and Savings Plans measure different things and cannot be summed.
- Subtract Enterprise Support explicitly, at up to 10 percent of spend, which is what turns a 15 percent headline into 11.
- Never compare a covered compute rate to an estate average, in either direction, since both errors lead to the wrong conclusion about the deal.
- Treat marketplace as shortfall insurance, not a saving, because it burns commitment at 100 percent credit on money that left the business anyway.
- Hold EDP expectations inside 5 to 20 percent, stepped by commitment threshold, and read anything quoted above that as a different instrument.
What the instruments actually reach
From the Redress Compliance advisory engagement file, 2024 to 2025:
What a negotiated headline becomes once each instrument is applied to its own base and Enterprise Support is subtracted from the whole.
Savings Plans and Reservations, a genuine figure that is not comparable to an EDP percentage because it is a rate on a workload rather than on a commitment.
EDP runs 5 to 20 percent stepped by commitment threshold, and PPA reaches 5 to 40 percent on specific service lines. Enterprise Support takes up to 10 percent of spend back off the top.
Marketplace burns down EDP commitment at 100 percent credit. That is shortfall insurance rather than a saving, because the money left the business either way.
Your first five moves
- Build one model of the whole estate, with every instrument attached to the base it actually applies to.
- Subtract Enterprise Support from total spend rather than leaving it out of the discount arithmetic.
- Separate covered compute from everything else so the 72 percent figure never contaminates the estate average.
- Reclassify marketplace as shortfall protection, not as a saving line.
- Negotiate against the effective rate. The AWS practice builds the model with you.
Frequently asked questions
Why does a 15 percent discount land at 11?
Because the headline applies to one base while the bill covers the whole estate, and Enterprise Support takes up to 10 percent of spend back off the top. Nothing went wrong; the number described a narrower base than the invoice.
What is the actual EDP range?
5 to 20 percent, stepped by commitment threshold. Anything quoted above that is either a different instrument or a percentage of a different base.
What does PPA reach?
5 to 40 percent, but only on specific service lines rather than across the agreement, which is why it cannot be added to an EDP percentage.
Are Savings Plans really up to 72 percent?
Yes, on covered compute. It is a genuine figure and it is a rate on a workload, not on a commitment, so it is not comparable to an EDP number.
What is the commonest comparison error?
Benchmarking an estate average against somebody else covered compute rate, or quoting your own covered rate as an estate figure. Both compare instruments that do not measure the same thing.
How should marketplace be treated?
As shortfall insurance. It burns down EDP commitment at 100 percent credit, which is valuable, but the money left the business anyway so counting it as a saving double counts it.
Why does Enterprise Support matter so much here?
Because it is the only instrument that runs the other way and it is calculated on total spend rather than a subset, so it reduces the effective rate across everything.
Can the instruments be added together?
No. Each applies to a different base, and any model that sums them produces a number the invoice will never match.
What number should we negotiate against?
The effective rate across the whole estate. It is the only figure that reflects what you will actually pay, and it is what makes two offers genuinely comparable.
Is a well negotiated agreement still disappointing?
It can look that way to whoever approved it, if the business case used the headline. Modelling the effective rate up front removes that gap without changing the deal.
