Full narration of the briefing. Click a section heading to jump the player to that moment.
If you search for AWS discount benchmarks you will find half a dozen tables, all confident, all different. At a ten to twenty five million dollar commitment one firm says twelve percent and another says thirty five. That is not a benchmark. That is noise.
So this episode is about what is actually known, what is merely asserted, and how to build a target that survives contact with a deal desk.
Start with the one number that is actually verifiable. AWS publishes no discount schedule at all, but real private pricing agreements get filed with securities regulators as contract exhibits. One of them, from a company committing a million dollars a year over three years, shows a flat nine percent. Nine.
Not thirty. And a transaction dataset of over four hundred AWS deals averaging around two million dollars puts mean savings just under ten percent. When someone shows you a table promising twenty five to thirty five at that size, they are almost certainly counting Savings Plans, which you could have bought yourself off the public rate card. And read every table for what it is measuring.
A number that blends the negotiated discount with Savings Plans is not comparable to one that isolates it, and almost none of them say which they did. Ask any benchmark three questions: what term, what commit size, and does it include rate card discounts you could buy without negotiating at all.
The bands are real, but they are priors. Discount scales with committed volume and term. Single digits at the entry threshold around a million a year. Low to mid teens in the five to twenty five million range.
Twenty percent and above at the largest commitments, and past that for the strategically important logos. Three years is the enterprise default and five years typically buys four to six more points. But price the option you are selling: a shorter term lets you renegotiate into a market that keeps getting cheaper, and the filed contracts show that signing the addendum also suspends your right to terminate for convenience for the whole term.
The breakpoints matter more than the curve. The discount curve is a step function, not a smooth line. Practitioners consistently name thresholds where a small increase in commitment triggers a disproportionate jump in rate. So ask the account team a very specific question: show me the rate at my number, at my number plus a hundred thousand, and at my number plus a quarter million.
If a small step up moves the rate materially, you have found a breakpoint worth reaching. If it does not, you have just proved that committing more buys you nothing, which is equally useful.
Competition is the best evidenced lever there is. Buyers who brought a credible Azure or Google alternative achieved three to eight percentage points better than those who did not. That is more than most concessions you could argue about for a month. But credible is doing the work in that sentence.
Under one percent of cloud customers switch provider in a year, and AWS knows the number. A threat to leave an estate you have spent three years migrating is not a threat. What is credible is a named executive sponsor at the alternative, a completed free assessment from their migration tooling, a defined pilot workload with a date, and a board paper. Present it as evidence, never as a threat.
And note where the leverage actually lives: in what you have not yet moved. Your negotiating position is strongest twelve to eighteen months before the workloads are live and irreversible, and weakest the day after they land. One more correction worth making before you build the target. The effective discount is not the headline.
Marketplace spend retires commitment at zero discount, support is charged on gross spend and never discounted, and professional services and training are excluded outright. Run a fifteen percent headline through a real estate and it lands nearer eleven, and nearer nine once support is counted.
Then set the number you would sign and sleep. Take the floor from your baseline, apply the band for your size and term, and write the target rate, the ramp shape, the shortfall relief, the support treatment and the credit ask on one page before anyone quotes you anything. Because AWS's first offer typically sits five to twelve points below what is achievable, and its opening commitment ask runs fifteen to thirty percent above the customer's own model. Both of those are opening positions, not assessments.
If you have not written your own numbers down first, theirs become the anchor by default.
One last point. At Redress Compliance we benchmark and run these negotiations on pure contingency: twenty five percent of what we save you. Nothing saved, nothing paid. Next episode: the campaign plan, and the AWS approval machine that decides whether your ask ever reaches someone who can say yes.
Redress Compliance works on contingency: our fee is 25 percent of what we save you. Nothing saved, nothing paid. Independent, buyer side only, never vendor funded. Want Redress to contact you? Reach out and we respond the same day.
Talk to a AWS negotiator