Forecasting has become more difficult with token and credit models because revenue now depends on how customers use a product, not just what they subscribe to. Unlike traditional seat-based pricing, usage-based and credit-based models introduce fluctuating consumption patterns. This makes revenue, cash flow, and demand harder to predict. As more software companies adopt token and credit pricing, FP&A teams need new forecasting methods, metrics, and closer collaboration with product and customer success teams to improve forecast accuracy and respond to changing customer behavior.
A Disney subscription looks the same every month whether you binge four shows or none at all. Software pricing used to work on that same flat-fee logic. That logic is breaking down fast. Across 2025 and into 2026, one vendor after another swapped seat-based plans for pay-per-use pricing (API calls, compute hours, AI tokens, credits). Figma made the shift. So did HubSpot. So did Salesforce. Industry research now puts the majority of large software companies on some form of consumption-based pricing, and credit-based plans specifically have grown sharply over the past year while pure subscription pricing keeps shrinking.
Customers like it: lower entry price, and you only pay for what you use. It’s a much harder story for whoever has to predict next quarter’s revenue. Forecasting has become more difficult with the token and credit models, and most finance teams are still building forecasts as if nothing changed.
Why Consumption-Based Pricing Is Rewriting Revenue Forecasting
Under seat-based or annual-contract pricing, forecasting was close to mechanical: count seats, apply a renewal-rate assumption, layer in pipeline, land within a reasonable range. Consumption revenue doesn’t behave that way, and the reason sits inside the math itself:
Revenue = Customers × Consumption × Price
That middle variable (consumption) is where most forecasts start to fall apart. A customer might burn through usage fast right after onboarding, then level off. Usage often dips early as new budgets face scrutiny, then jumps months later when a team discovers an unused feature. None of this moves in a straight line, so a forecast built on historical averages blends together several behavior patterns and misrepresents all of them.
Snowflake’s own filings illustrate the shift: net revenue retention fell from 178% in fiscal 2022 to 126% by fiscal 2025. The product wasn’t the problem. Enterprise customers simply reduced spending because usage-based pricing made it easy to scale consumption up or down. The mechanics that fuel outsized growth in strong years make contraction just as fast when budgets tighten, and conventional SaaS metrics were never built to separate those two stories.
Finance Metrics for Consumption Pricing
Most of the standard toolkit (ARR, MRR, net dollar retention, churn) assumes revenue is ratable and predictable across the billing period. Add a consumption layer, and that assumption no longer holds. Teams running hybrid pricing need to rethink their metrics on two levels: what to adapt, and what to build from scratch.
Committed ARR vs Consumption ARR
The first move is splitting ARR into two buckets. Committed ARR covers platform fees and contractual minimums, the stable piece. Consumption ARR captures the variable, usage-driven revenue and trailing credit spend, annualized. Reporting these separately rather than blending them into one ARR figure is becoming standard practice for forecasting consumption-based revenue with any real accuracy.
Credit Burn Rate Forecasting and Utilization
Credit burn rate (how many credits a customer works through per day or week) deserves attention because it moves ahead of the revenue line. Accelerating burn usually signals an overage or upgrade on the way. A slowing burn signals something has changed, and customer success should hear about it before the next business review.
Credit utilization (consumed versus allocated) answers a different question: is the customer getting value from what they bought? An account still sitting at 40% utilization three months in is a warning sign.
A third figure, days of credit exhaustion, divides the remaining balance by daily burn. Once that drops under roughly 20 days, it’s time for a top-up or upgrade conversation. Subscription renewal dates used to create that natural trigger; in a credit model, this metric is the closest replacement.
Revenue Recognition for Prepaid Credits Under ASC 606
ASC 606 for credit models adds another layer of difficulty. When a customer prepays for credits, that cash can’t be booked as revenue until the credits are consumed. It sits on the balance sheet as deferred revenue. Recognition depends on how fast customers burn through their balance, not the calendar, which breaks from subscription accounting, where deferred revenue rolls off on a fixed, predictable schedule.
A few complications follow. Estimating breakage creates additional forecasting challenges because finance teams must predict unused credits before expiration occurs consistently each period ahead. ASC 606 allows finance teams to recognize estimated breakage proportionally as customers redeem remaining credits when historical redemption patterns persist. Mid-cycle top-ups count as contract modifications, triggering a reallocation of the transaction price. Volume discounts or tiered rates introduce variable consideration that must be constrained under the standard.
Rethinking Variance Analysis for Consumption Revenue
Subscription businesses usually trace revenue misses to churn, weak bookings, soft expansion, using net-dollar-retention bridges to isolate each cause accurately.
Credit-based revenue disrupts that structure, forcing finance teams to analyze every miss across four independent dimensions before explaining performance drivers:
- Volume: Did fewer customers buy additional credit packs than expected?
- Intensity: Did customers consume fewer credits per period than the model assumed?
- Price: Did effective credit pricing shift due to discounting or promotional packs?
- Mix: Did the customer base skew toward lower-consumption segments or use cases?
Intensity is the hardest to explain from a spreadsheet alone. It requires product-level usage data, not just finance data. That’s the core adjustment for FP&A under hybrid pricing: a credible forecast can’t be built by extrapolating last quarter’s totals. Someone has to understand why usage moved: a feature launch, a seasonal campaign, one large customer running an unusual workload. Which means asking product and customer-success questions, even though the forecast still lands on finance’s desk.
How FP&A Must Adapt to Usage-Based Pricing
Why Forecasting Cadence Must Change
Quarterly forecasting suits contract-based revenue that barely shifts between updates. Consumption revenue needs a faster rhythm: monthly reforecasting of the variable portion. Plus, a weekly check on credit burn for the accounts most likely to drive the variance. The committed piece can stay on its quarterly cycle, treat the two as separate forecasts that roll up into one number.
FP&A Must Partner Beyond Sales
This is the biggest shift for FP&A teams. Traditional counterparts were sales, for pipeline and bookings, and occasionally HR for headcount. Usage-based pricing forecast work pulls in four more:
- Product, which controls the roadmap and can double credit burn with one feature launch;
- RevOps, which owns billing rules, credit multipliers, and pricing tiers;
- Customer Success, which sees adoption trends before anyone else;
- and Accounting, since breakage and variable consideration require shared assumptions.
What Leadership Needs to See Each Month
Board decks should report Committed ARR and Consumption ARR as separate lines so leadership can see how much of the revenue base is stable versus behavior-dependent. A monthly credit-health summary helps too: average utilization across the base, accounts above 100% utilization (expansion signal), accounts below 50% (churn risk), and the month-over-month shift in overall burn rate.
What FP&A Teams Should Do Next
None of this requires rebuilding an entire revenue model overnight. It requires separating what’s contractually locked in from what depends on customer behavior. Three practical starting points:
- Split ARR into committed and consumption components;
- Work with product teams to get usage data flowing, even if messy at first;
- And build a closer working relationship with customer success and product so the real drivers behind usage (not just historical averages) inform the forecast.
For any company moving toward token or credit pricing, revenue forecasting can no longer lean mainly on contracts, renewals, and pipeline. It also has to account for customer behavior, product usage, pricing mechanics, and revenue recognition rules. That’s a different forecasting discipline than the one most finance teams trained on. And the teams that adapt early will be able to explain not just where revenue came from, but where growth and risk are heading next.
