Support ran 3 to 10 percent of the AWS bill and almost nobody negotiated it as a line

The commitment conversation absorbs all the attention at renewal, so two costs grow unmanaged in between. Enterprise Support is priced as a percentage of spend that nobody checks, and data transfer is billed per gigabyte against architecture decisions procurement never sees.

Prepared by Redress Compliance · August 15, 2026 · AWS advisory. Based on 35 to 45 AWS commitment and cost engagements, 2024 to 2025.

• Anchor the floor on conservative trailing spend, not the optimistic plan, and negotiate a ramp that grows with verified consumption rather than with forecast.
• Validate the support band against actual spend, and challenge whether every account genuinely needs the top tier, since the percentage declines in bands nobody re-checks.
• Quantify egress and cross region transfer as their own line, and ask for transfer credits tied to the migration ramp rather than discovering the cost after the architecture is fixed.
• Layer a Savings Plan base under predictable load and leave a margin on demand for variability, because a single instrument either over commits or under saves.
• Route eligible third party software through Marketplace to retire commitment you already owe, inside the 25 percent contribution cap.
• Benchmark before every renewal, since the discount is unpublished and without a comparison you negotiate blind against a party that sees thousands of deals.

Every AWS cost conversation converges on the same place: the commitment and the rate. That is rational, because the commitment is the largest single decision and the rate is the number an executive can repeat. It also explains why support and data transfer drift. Both are billed as consequences of decisions taken elsewhere, neither appears as a line anyone is asked to defend, and together they can exceed the value of the discount the negotiation produced.

Support is the cleaner example. It is priced as a percentage of spend that declines across bands, which means it moves automatically as the estate grows and never requires anyone to approve an increase. At 3 to 10 percent of the total bill it is comfortably larger than the difference between two adjacent EDP discount tiers, and it is the only major AWS line where the buyer can change what they consume rather than merely what they pay. Two questions recover most of the value: does the band math match actual spend, and does every account need the top tier. Both take an afternoon, and neither is anyone's job.

Egress behaves differently and is harder to fix late. It is billed per gigabyte and generated by architecture decisions, so by the time the finance team notices 20 to 40 percent annual growth, the traffic pattern is a design rather than a choice. That timing is the whole problem. Transfer credits are negotiable at signature, when AWS is competing for a migration and wants the workload; they are almost impossible to win two years later, when the data is already flowing and the alternative is a re-architecture nobody has funded. The cost is procurement's to negotiate and engineering's to create, which is precisely why it survives.

The correction is structural rather than clever. AWS cost management is a quarterly function that owns four levers, not an annual event that owns one. The commitment floor gets modeled against trailing spend, the compute instruments get layered rather than chosen, the support tier gets re-validated every year, and egress gets priced at the point where it is still negotiable. Run that rhythm and the annual renewal becomes a confirmation of decisions already made, rather than the only moment the estate is examined. The commitment mechanics sit in the EDP negotiation brief, the renewal step in the renewal strategy, and the transfer detail in the egress brief.

Across roughly 35 to 45 AWS commitment and cost engagements in 2024 and 2025, the same two failures repeat:

The patterns: floors modeled on the plan rather than the history, support tiers set once and never revisited, and transfer costs discovered after the architecture made them permanent.

The buyer side move is to own the four levers all year. The wider library sits in the AWS practice.

What does AWS vendor management actually cover?

The continuous control of commitment, consumption and contract terms across the Enterprise Discount Program floor, Savings Plans and Reserved Instances, the support tier, and data transfer. It is a quarterly rhythm rather than an annual renewal event, because the estate changes every month and the controls have to change with it.

Why is Enterprise Support worth negotiating separately?

Because it ran 3 to 10 percent of total AWS spend in the estates we reviewed and was almost never treated as its own line. It is priced as a declining percentage of spend across bands, so both the band math and whether every account genuinely needs the top tier are open questions worth asking each year.

How fast do egress and transfer costs grow?

In data heavy estates we measured 20 to 40 percent year on year growth in egress and cross region transfer. They are billed per gigabyte, rarely modeled at signature, and grow with architecture decisions rather than with procurement decisions, which is why nobody owns them until they are large.

What discount does an EDP actually deliver?

Between 5 and 20 percent, stepped by commitment thresholds rather than negotiated freely: roughly 5 to 10 percent at $1M to $5M of annual commitment, 10 to 15 percent at $5M to $25M, and 15 to 20 percent above $25M. Offers mix credits with rate discounts and the shape changes quarter to quarter.

How should the EDP commitment floor be sized?

Against conservative trailing spend rather than the optimistic plan. More than half the floors we reviewed were scoped 15 to 30 percent above actual trailing spend, and the shortfall is paid in cash, so the arithmetic that produced the headline discount quietly reverses.

Does Marketplace spend help retire the commitment?

Eligible AWS Marketplace purchases can count toward the commitment, which lets third party software you were buying anyway retire commitment you already owe. Note that the marketplace contribution toward commit is capped at 25 percent, so it is a routing mechanism rather than a discount, and it does not rescue an oversized floor on its own.

What cadence should a vendor management function run?

Quarterly. Compare actual spend against the commitment floor and the Savings Plan coverage, adjust the on demand margin before it compounds, and benchmark the discount position before every renewal. Monthly consumption review catches a developing shortfall while it is still cheap to correct.