Forecasting Spend Before Signing an AWS EDP

Forecasting spend before signing an AWS EDP

Forecasting spend before signing an AWS EDP is the work that decides whether the commitment saves you money or costs you a penalty. The Enterprise Discount Program asks you to commit a dollar figure of eligible spend over a one to five year term, and as of June 2026 there is no rollover of unused spend, so a forecast that runs high turns into a shortfall you pay (source: AWS EDP terms, as of June 2026). A forecast that runs low simply means you committed less and still earn the discount on everything above the floor. The asymmetry is the whole reason to forecast conservatively.

Sellers will offer to forecast for you. Their model points up and to the right, because their incentive does. Build your own forecast from your own data, and treat the seller's number as a negotiating position, not a baseline.

Start from real consumption, not a growth story

Pull your actual eligible spend over the trailing twelve to twenty four months. Strip it down to the durable base by removing one time migrations, seasonal peaks, proof of concept environments, and any workload with a known end of life inside the term. What remains is the spend you can rely on, and the only spend that belongs in the committed floor. Our guide on What to exclude from an AWS EDP commitment details exactly what to exclude.

Separate the base from the variable layer

Model spend as two layers. The base is durable, recurring, and predictable. The variable layer is growth, bursts, and projects. Commit the base, leave the variable layer above the line, and you capture the discount on variable spend when it appears without ever owing a penalty if it does not.

Model the ramp, not just the total

A multi year commitment has a shape, not just a size. Migrations land over quarters, not on day one, so a flat commitment from month one can put you in shortfall before your workloads even arrive. Build a ramp that matches when spend genuinely shows up, and resist a back loaded ramp that defers risk into the years you can forecast least well.

• Plot spend quarter by quarter, not as a single annual figure.
• Tie ramp steps to specific, dated migration milestones.
• Stress test the curve against slipped timelines and optimization.
• Reserve headroom in each period rather than committing to the top of your range.

Account for the things that lower spend

Forecasts fail because they ignore the forces that reduce consumption. Reserved Instances and Savings Plans already lower your effective rates, and the EDP discount stacks on top, so your committed dollars buy more covered usage than gross list pricing suggests. Right sizing, autoscaling, and architectural optimization cut consumption further. Every efficiency gain is good for your bill and bad for a commitment that assumed you would keep spending at the old rate.

Factor planned optimization into the forecast. If your FinOps team is about to cut twenty percent of waste, do not commit as though that spend will continue. The discount applies to real usage, and real usage is heading down where your team is doing its job.

Scenario model the base, likely, and stretch cases

A single point forecast is a guess dressed as a plan. Build three cases instead. The base case is the durable spend you are confident recurs. The likely case adds growth you can reasonably defend. The stretch case includes the optimistic outcomes the seller will cite. Commit at or below the base case, and treat the difference between base and stretch as the discounted upside you capture without obligation.

Scenario modeling also arms you for the negotiation. When the seller pushes the stretch number, you can show the base case and the assumptions behind each, and anchor the commitment on evidence. A forecast that survives scrutiny is worth more at the table than a number you cannot explain.

Document assumptions and involve the right people

A forecast is only as good as the assumptions under it, and those assumptions live across the organization. FinOps knows the optimization roadmap, engineering knows the migration timeline, and finance knows the business growth plan. Pull them together before you commit, write the assumptions down, and revisit them as the negotiation evolves. A forecast assembled in a silo by one team is the one most likely to miss the optimization that quietly shrinks the bill.

Documented assumptions also make the commitment defensible later. If consumption diverges from plan, you can see which assumption broke and respond, rather than discovering a shortfall at year end with no idea how you got there.

Turn the forecast into a defensible commitment

The output of good forecasting is a number you can defend under pressure: a durable base, ramped to match real timing, with headroom for error, and the variable layer deliberately left above the floor. That commitment reaches its tier comfortably, avoids the shortfall, and still earns the discount on growth.

This is also your negotiating evidence. A forecast grounded in real consumption lets you push back on the seller's inflated number with data rather than instinct. Overcommitment is the most common mistake in this program, and a rigorous forecast is the cure. Our analysis of Why AWS EDP overcommitment is the most common mistake shows the cost of skipping this step, and our guide on AWS EDP discount benchmarks by commitment size explains the discount your sized commitment should command. This is commercial guidance, not legal advice.