A commitment is a bet on future usage, so it is only as good as the forecast behind it. Forecasting commitment needs is the work of turning usage history and known business plans into a defensible view of the stable usage you can safely lock in, so you commit to what will actually be there rather than to hope.
How to forecast commitment needs is the process of projecting your future stable cloud usage so you can size reserved instances, savings plans, and committed use discounts to it. A good commitment forecast is not a single optimistic growth number; it is a conservative estimate of the usage you are confident will persist for the full commitment term, built from real usage history and adjusted only for business changes you actually know about. Forecast too high and you create idle commitment; forecast too low and you leave savings on the table. The skill is calibrating to the floor of likely usage, not the ceiling.
This article is part of the complete guide to cloud commitment management. The method below is how we build commitment forecasts across the 500-plus environments we have optimized since 2019.
Pull at least several months, ideally a year, of usage at hourly granularity. The forecast is only as trustworthy as the data, so the usage must be normalized and owned, the output of the See step of our method, before you model anything. Strip out one-off events, completed migrations, and decommissioned workloads so the history reflects the ongoing baseline rather than past noise.
From the cleaned history, find the level of usage present in essentially every hour, the floor described in commitment management for variable workloads. The floor, not the average, is the foundation of the forecast, because it is the usage you can be virtually certain will continue. Forecasting commitment needs is largely the discipline of projecting how that floor will move over the commitment term, not chasing the peaks.
Build the forecast on the usage present in every hour, then apply only growth you can actually defend, such as a signed contract or a planned launch. Treat speculative growth as upside to cover flexibly later, never as a reason to commit deeper today.
Layer on top of the floor only the changes you have concrete evidence for: a signed customer that will add load, a product launch with a firm date, a planned migration that will move or remove workloads, a rightsizing program that will lower the floor. Resist projecting generic growth percentages; a commitment term is long, and optimistic growth that does not arrive becomes idle commitment, the leak detailed in the hidden cost of idle commitments. If growth is uncertain, treat it as upside to be covered flexibly, not as a reason to commit deeper now.
A forecast that ignores your own optimization plans will over-commit. If you intend to rightsize, schedule, or eliminate idle resources over the coming quarters, the committable floor is the usage that survives that cleanup, not today's inflated baseline. This is why forecasting and rightsizing have to be sequenced together, the logic in why you should rightsize before you commit: forecast the estate you are about to have, not the one you are about to fix.
A forecast is a range, not a point, so do not commit the whole projected floor at once. Convert the forecast into a laddered purchase plan, as in how to ladder cloud commitments to reduce risk, committing the most certain portion of the floor first and adding tranches as the forecast proves out. Laddering turns forecast uncertainty from a risk into a series of small, correctable decisions, and the term choice for each tranche follows the reasoning in 1-year vs 3-year commitments.
A commitment forecast is never finished. Revisit it at least quarterly: compare actual usage against the forecast, recalibrate the floor, and feed the variance back into the next purchase. This continuous loop is part of the Run step of our method and connects to the broader practice in continuous rate optimization. A forecast that is checked and corrected every quarter stays accurate; one made once and forgotten drifts into over- or under-commitment within a single term.
We build the forecast from your real usage, strip out the noise, account for your optimization plans, and turn it into a laddered purchase that tracks reality. On the performance model, if we do not save you money, there is no fee.
Get a commitment audit →Forecasting is the analysis that feeds every purchase decision in this cluster. Read the complete guide to cloud commitment management for the full discipline, and download The Commitment Strategy Playbook: RIs, Savings Plans, CUDs for the forecasting and floor-analysis worksheets. When you want the forecast built and maintained for you, see our commitment management service.
New commitment instruments, FOCUS changes, hyperscaler pricing shifts, and the plays that actually move a bill. No schedule, no filler.