Reserved Capacity Decision Under Uncertain Growth
A SaaS company spends $18,000/month on compute. About 60% of that is a steady baseline that has grown less than 5% per quarter for two years; the rest is spiky batch processing that runs unpredictably. The finance team wants to cut the bill by 30% before the next board meeting in six weeks. The engineering team is mid-migration: two services will move to a different instance family next quarter, and one legacy service may be retired entirely. The company has no commitment-based discounts today. Leadership is tempted to commit the entire footprint for three years to get the deepest discount. There is also a batch workload that is interruption-tolerant and could be restructured. The CFO wants a plan that captures savings quickly without locking in a mistake, and wants to know how the savings will be measured.
1. Recommend how to split the commitment strategy, and explain the risk being managed.
Read the worked answer
Do not commit the entire footprint for three years. The risk is that the discount is real but the commitment is a bet on instance family, region, and duration that the migration may invalidate. A sound split: commit only the steady baseline, and only the portion you are confident survives the migration and retirement. Since 60% of $18,000 is about $10,800/month of stable usage, a commitment covering most but not all of that baseline leaves headroom for the two services changing family and the legacy service that may disappear. Use a flexible commitment type that can apply across instance families and sizes so the migration does not strand the discount, and stagger terms so not everything expires at once. For the spiky, interruption-tolerant batch work, use spot capacity rather than commitments, since it is designed for exactly that profile and does not lock in. For any remaining on-demand usage, right-size first so you are not committing to waste. Measure savings against a documented baseline bill with the same usage, and track utilization of commitments so unused coverage is visible before it becomes a sunk cost.
2. The CFO asks whether the 30% target is realistic in six weeks. How should the team answer?
Read the worked answer
Answer with a decomposed estimate rather than a promise. The baseline portion can capture meaningful savings through commitments, but the achievable percentage depends on how much of the bill is committable and how quickly purchases take effect. If roughly 60% of spend is steady and the commitment discount is in the range typical for that term and payment option, the blended savings will be well below the discount rate itself, because the spiky 40% does not benefit from commitments. The batch portion can contribute separately through spot and restructuring, which may take longer than six weeks to implement safely. So the honest answer is that a large share of the target is achievable quickly from the baseline, and the remainder depends on the batch work and on right-sizing. Present a range with assumptions, identify which levers are immediate versus which need engineering time, and commit to reporting actual savings against the baseline monthly. Do not promise 30% by the board meeting if the arithmetic does not support it; instead show the path and the confidence level.