Usage-Based Pricing Breaks the Revenue Forecast You Already Trust

A revenue team that’s spent years forecasting fixed-subscription revenue tends to carry that model’s habits into a usage-based pricing shift without fully appreciating how different the underlying behavior actually is. A subscription contract commits a customer to a known dollar amount for a known term, and forecasting it is mostly a renewal and churn exercise. Usage-based revenue commits a customer to nothing beyond a relationship — the actual dollar amount depends on how much they use the product in a given period, which can swing meaningfully based on factors that have nothing to do with the health of the relationship itself.

Applying subscription-era forecasting habits to a usage-based revenue base produces forecasts that look precise and are frequently wrong in ways the old model never had to account for.

Usage Isn’t the Same Signal as Commitment

A subscription customer’s contracted amount is a reasonably stable, forward-looking number, because the customer has committed to it contractually regardless of how much they actually use the product day to day. A usage-based customer’s revenue contribution is only known in arrears, based on actual consumption, which means forecasting it forward requires predicting behavior rather than reading a signed commitment. Two customers who look identical on paper — same contract terms, same onboarding date, same account size — can produce wildly different revenue simply because one uses the product heavily and the other barely touches it, and neither pattern is visible from the contract itself.

The Volatility That Aggregate Numbers Can Hide

Usage-based revenue at the individual account level can be considerably more volatile than subscription revenue, even when the aggregate portfolio looks smooth. A single large customer scaling down usage temporarily — for a seasonal slowdown, a budget freeze, an internal reorganization — can meaningfully affect that month’s revenue in a way a fixed-fee customer simply couldn’t, because the fixed-fee customer’s contribution doesn’t fluctuate month to month regardless of their internal circumstances. Forecasting models that only look at the smoothed aggregate can miss this account-level volatility entirely, right up until a concentration of usage-heavy accounts all soften in the same period.

Comparing the Two Forecasting Foundations

Subscription Revenue	Usage-Based Revenue
Known dollar commitment for the term	Actual dollar amount depends on consumption
Forecast risk concentrated in churn/renewal	Forecast risk concentrated in usage variability
Relatively smooth month to month	Can swing meaningfully with customer behavior
Contract signing is the key forecasting event	Ongoing usage pattern is the key forecasting signal

A revenue model that blends both types, which is increasingly common, needs to forecast each component separately rather than applying one unified method to a mixed base, because the two behave according to fundamentally different mechanics.

Seasonality Hits Usage Revenue Differently Than Subscription Revenue

A subscription business experiences seasonality mostly through its sales and renewal cycle — deals closing faster or slower in certain periods. A usage-based business experiences seasonality directly in its revenue line, because customer usage itself often follows a seasonal pattern tied to the customer’s own business cycle, not the vendor’s. A B2B usage-based product sold heavily into retail customers, for instance, might see usage spike around a retail customer’s own peak season and drop in their slow season, and that pattern shows up directly in monthly revenue regardless of how the sales pipeline is performing. Building customer-specific or segment-specific seasonality into the forecast, rather than assuming a flat run rate, catches this pattern before it produces a surprising miss.

The Temptation to Forecast Off a Recent Average

A common and risky shortcut is forecasting future usage-based revenue by simply extending a recent average forward, smoothing over the fact that the recent average might itself have been an unusually high or low period for reasons that won’t necessarily persist. A customer who ran an unusually large one-time workload last month isn’t necessarily establishing a new baseline, and forecasting forward as if they are sets up an avoidable miss when their usage reverts to a more typical level. Distinguishing genuine trend from a temporary spike requires looking at a longer history and understanding what drove any recent unusual period, not just extrapolating the most recent data point.

Sales Compensation and Forecasting Need to Agree on What Counts

Usage-based pricing also complicates sales compensation in a way that feeds back into forecasting accuracy: if reps are compensated on initial contract value rather than actual realized usage revenue, there’s a structural incentive to oversell expected usage during the sales process to make the deal look larger, which then inflates the pipeline value the forecast is built on. Aligning compensation more closely with realized usage, or at least being explicit internally about the gap between contracted potential and likely actual usage, keeps the forecasting inputs closer to what’s actually likely to materialize.

Building Confidence Intervals Instead of Point Estimates

Because usage-based revenue carries more inherent unpredictability than subscription revenue, presenting it as a single point forecast overstates the precision anyone can honestly claim. A range, built from historical usage volatility for comparable accounts, communicates the real uncertainty far more usefully to leadership making decisions off the forecast, even though it’s a less satisfying single number to put in a board deck. Resisting the pressure to collapse that range into one confident-sounding figure protects the forecast’s credibility when actual results inevitably land somewhere within the range rather than exactly on a single predicted point.

Forecasting the Behavior, Not Just the Contract

The shift to usage-based pricing changes what a revenue forecast actually needs to predict — not just whether a customer renews, but how much and how consistently they’ll actually use the product across the forecast period. Treating this as a minor adjustment to an existing subscription forecasting model, rather than as a genuinely different forecasting problem requiring its own methodology, is where most usage-based revenue forecasts go wrong. Getting it right means building a model that respects how differently usage-based revenue actually behaves, not one that’s been quietly retrofitted from a subscription-era approach.