How to Forecast Subscription, Usage-Based, and Hybrid Revenue
17 MIN READ
|
Updated on October 9, 2026
Most forecasting guides still assume revenue arrives in equal monthly slices. That assumption quietly breaks the moment a share of revenue comes from usage, overages, or prepaid credits. This guide shows finance and RevOps teams how to forecast recurring, usage-based, and hybrid revenue accurately, and it explains why an accurate forecast now depends on the usage data that engineering and product teams own.
What Is Subscription Revenue Forecasting?
Subscription revenue forecasting is the practice of projecting future recurring revenue by combining the size of your subscriber base, average revenue per account, expected churn, and expected expansion. It answers one question for finance and the board: how much revenue will the business recognize over the next several periods, and how confident can we be in that number.
The core inputs every forecast needs are the same across most models. You start with the current base of active subscriptions, apply an average revenue figure, subtract expected churn, and add expected expansion from upgrades and cross-sells. Those four drivers, subscriber count, Monthly Recurring Revenue (MRR) per account, churn, and expansion, produce a baseline recurring revenue line.
The core inputs every forecast needs
A defensible forecast starts with clean values for each driver. Count active paying subscriptions, calculate average revenue per user (ARPU) from booked revenue, measure churn as lost revenue rather than lost logos, and separate expansion revenue so you can model it on its own curve. Pressure-test the recurring base against these four drivers before layering anything variable on top.
What changes when revenue is not flat
The familiar definition is correct, and it is also incomplete, because it assumes revenue is predictable per subscriber. That assumption holds for a single flat plan. It falls apart once a customer’s bill moves with consumption. Today, 51% of recurring-revenue companies combine subscription with usage- or outcome-based pricing, and 75% retain a subscription element. When half your peers earn part of their revenue from consumption, a subscription-only forecast describes a shrinking slice of the real picture.
Why Does Flat MRR Forecasting Break Under Usage-Based and Hybrid Pricing?
Flat MRR forecasting projects revenue as subscribers multiplied by ARPU, minus churn. The model treats every account as a fixed monthly amount. It breaks the moment part of that account’s revenue depends on how much the customer consumes, because a single average hides the volatility that drives the actual number up or down.
Where the “subscribers × ARPU” model fails
The failure is specific. When a customer pays a base fee plus metered usage, ARPU stops being a stable input and becomes an output that changes every month. Averaging across light and heavy users smooths away the peaks that matter most. Finance forecasts a straight line, product knows usage is climbing, and the two teams argue over which usage numbers are real. Pricing is also moving faster than most models assume: 77% of companies changed their pricing model in 2024, and 70% raised prices while 40% failed to align those increases with perceived value. Every one of those changes makes a flat model more fragile.
What it costs when the forecast is wrong the month usage spikes
“Our forecast is always wrong the month usage spikes” is a common complaint from finance leaders, and the cost is real. Under-forecast a usage spike and you miss a board target you could have committed to with confidence. Over-forecast it and you over-hire against revenue that never lands, then face a cash crunch two quarters later. Finance ends up reconciling numbers on weekends at month-end because the model and the billing system disagree. The failure is not a rounding error. It is a credibility problem with the board. For teams moving off flat pricing, our guide on the transition to usage-based pricing covers how the revenue mix shifts in practice.
How Do You Forecast Revenue by Model?
The Forecast-by-Model framework solves the volatility problem directly: forecast each revenue type with its own driver formula, then sum the results into one total. A single blended formula hides the variation that makes usage revenue hard to predict. Separating the models is what makes the forecast defensible line by line, because you can show the board exactly which driver moved.
The four models below each get their own driver formula and a one-line worked example. Chargebee Billing meters usage, overages, and prepaid credits on first-party data, which is what feeds these drivers accurately (full detail is in the section on how Chargebee supports forecasting).
| Model | Driver Formula | Forecast Inputs | Volatility |
|---|---|---|---|
| Flat subscription / MRR | Subscribers × ARPU, adjusted for churn and expansion | Active subscriptions, ARPU, churn rate, expansion rate | Low |
| Usage-based / metered | Forecast units consumed × rate, per tier | Expected units per tier, rate card, expected overage | High |
| Hybrid | Base subscription + forecast overage + prepaid credit drawdown | Base fee, overage forecast, credit balance and burn rate | Medium to high |
| AI-native / agentic | Credit, token, or outcome consumption × rate, with cohort ramp | Consumption per cohort, ramp curve, rate per unit | High |
Flat subscription / MRR revenue
Forecast the recurring base as subscribers multiplied by ARPU, then adjust for churn and expansion. For example, 1,000 accounts at $200 MRR, less 2% monthly churn plus 1% expansion, projects roughly $198,000 in the next month. This line is the most predictable input, so build it first and hold it steady while you model the variable layers separately.
Usage-based / metered revenue
Forecast metered revenue as expected units consumed multiplied by the rate, modeled per usage tier with an expected overage. For example, 500 accounts averaging 12,000 API calls at $0.002 per call projects about $12,000 in usage revenue, before overage. Model a usage range rather than a single point, because the top and bottom of that range are what the board needs to see. The hardest part is operational: the top usage-based pricing challenges reported are explaining the pricing structure to customers (22%) and building and maintaining metering infrastructure (21%).
Hybrid revenue (base + overage + prepaid credits)
Forecast hybrid revenue by summing three components: the base subscription, the forecast overage, and the modeled prepaid credit drawdown. For example, a $1,000 base plus $250 forecast overage plus $300 in credit burn projects $1,550 for the month. Forecast each component on its own curve, then stack them, because they move on different rhythms. The payoff is margin: 67% of companies using a hybrid pricing model expect improved margins, compared with 32% on pure usage-based pricing.
AI-native / agentic revenue (credits, tokens, outcomes)
Forecast AI-native revenue as credit, token, or outcome consumption multiplied by rate, with a cohort-based ramp for new accounts that consume little at first and scale over time. Group new logos into monthly cohorts, apply a ramp curve to each, then sum across cohorts. The commercial case for getting this right is strong: among companies adding AI to their products, those that align pricing with their AI innovation are twice as likely to expect high growth. Chargebee Billing supports usage-based and agentic consumption models natively.
How Do You Build Best-, Base-, and Worst-Case Scenarios?
A single-number forecast gives false confidence. A range shows the board where the risk truly sits. To build three scenarios, take the drivers from the Forecast-by-Model framework and vary the four that move most: new logos, usage per account, churn, and overage rate. Hold the recurring base steady and flex the variable inputs so each scenario tells a clear story.
Setting the drivers for each scenario
Set explicit assumptions for each case rather than nudging a single number up or down. The table below shows how the same four drivers produce three defensible outcomes.
| Driver | Worst Case | Base Case | Best Case |
|---|---|---|---|
| New logos per month | 20 | 35 | 50 |
| Usage per account | Flat | +5% | +12% |
| Monthly churn | 3.5% | 2.0% | 1.2% |
| Overage rate | Low | Expected | High |
Modeling expansion and contraction (NRR and GRR)
Layer expansion and contraction on top using Net Revenue Retention (NRR) and Gross Revenue Retention (GRR). GRR shows how much revenue you keep before expansion, and NRR shows the net effect once upgrades and cross-sells are counted. When existing accounts expand faster than they churn, the base grows without new logos. NRR above 100% correlates with faster growth: the median SaaS company with NRR of 100% or more grew about 48% year over year, roughly twice the rate of companies with lower NRR. Track recognized versus booked revenue with revenue recognition software so expansion shows up on the right schedule.
What Metrics and Benchmarks Should Anchor Your Forecast?
Benchmarks borrowed from flat-subscription businesses mislead once a chunk of revenue is usage-based, because those benchmarks assume revenue is stable per account. Anchor your forecast to the right unit and the right cadence, and treat any external benchmark as a range rather than a target.
MRR vs ARR, and when to use each
MRR and Annual Recurring Revenue (ARR) measure the same recurring revenue on different time horizons. Use MRR to manage month-to-month movement and to catch churn and expansion early. Use ARR for annual planning, board reporting, and valuation conversations. Usage revenue is variable rather than truly recurring, so model it separately and avoid folding it into either figure.
| Metric | Time Horizon | Best Used For | Includes Usage Revenue? |
|---|---|---|---|
| MRR | Monthly | Operational tracking, early churn and expansion signals | No, model usage separately |
| ARR | Annual | Board reporting, annual planning, valuation | No, model usage separately |
How far ahead to forecast and how often to update
A common approach is to forecast recurring revenue on a rolling 12-month horizon that most teams roll forward each month. The guiding principle is that the update cadence should match how fast your revenue drivers change. Refresh more often when a meaningful share of revenue is usage-driven and moves quickly.
Healthy churn and retention benchmarks
Retention deserves a place in the forecast because it is a top concern for the market: 53% of recurring-revenue businesses name customer retention a top concern. A healthy churn rate varies widely by segment and average contract value, so treat any single number with caution and model your own trailing rate first. The metrics that frame this discussion are churn, customer lifetime value (CLV), and customer acquisition cost (CAC), and reading them together shows whether retention is funding growth or draining it.
For a named reference point, SaaS Capital’s 2025 benchmarks report median net revenue retention of 101% and gross revenue retention of 91% for private B2B SaaS above $1M ARR. SaaS Capital treats gross retention of at least 90% as table stakes for peer parity, so model your own trailing rate against that floor.
What Are the Most Common Revenue Forecasting Mistakes?
Each mistake below maps to a specific dollar consequence that finance and product teams already feel. Naming them makes each one easy to check against your own model.
Forecasting usage as if it were flat
The most common mistake is applying a single average to revenue that moves with consumption. A flat assumption looks precise and misses every spike, which is exactly the month the board remembers. Forecast usage as a range per tier instead, and reconcile it against metered actuals each cycle.
Ignoring prepaid credit drawdown timing
Prepaid credits are collected up front and recognized as they are consumed, so the cash and the recognized revenue land in different periods. Teams that forecast the sale rather than the drawdown misstate both revenue and remaining balance. Model the burn rate on each credit balance and forecast when it will be exhausted and repurchased.
Single-scenario forecasting
A one-number forecast hides the risk that matters. Build best, base, and worst cases every cycle so the board sees the range and the assumptions behind it, rather than a single figure that will be wrong in one direction.
Ignoring involuntary churn and payment failure
Involuntary churn from failed payments quietly removes revenue you already earned, and many forecasts never account for it. Smart dunning and recovery reduce involuntary churn: this is a Chargebee Billing baseline, with Chargebee Receivables available as an add-on for deeper recovery.
How Does Chargebee Support Forecasting Across Models?
Disconnected billing, recognition, and retention tools force the reconciliation that makes every forecast late and contested. Chargebee closes that gap by treating subscription as one model on a connected billing layer rather than the whole model, so the same first-party data feeds the forecast across subscription, usage-based, hybrid, and AI-native revenue. Chargebee powers billing and revenue operations for 6,500+ businesses across those models on one platform.
Metering and hybrid billing in Chargebee Billing
Chargebee Billing meters usage, overages, and prepaid credits, and reports MRR, ARR, and churn on first-party transaction data. Because the billing system owns the events, the usage numbers feeding your forecast come from the source that handles the revenue rather than a synced copy. See how Chargebee handles usage-based billing and gen AI billing.
Recognized vs billed revenue with Chargebee RevRec
Chargebee RevRec is a separate product that separates recognized revenue from billed revenue under ASC 606 and IFRS 15 for usage and hybrid contracts. It applies recognition rules to variable consideration, overages, and prepaid credit amortization, so the revenue in your forecast lands on the correct schedule rather than when cash arrives. Explore revenue recognition software for how recognized and booked revenue diverge under variable pricing.
Leadinfo shows the operational payoff. Before Chargebee, the team handled subscription management and recurring billing manually. After standardizing billing on Chargebee, Leadinfo grew 3x faster than its previous rate and lifted ARPU by 25% as it moved upmarket.
Acting on the forecast with the Chargebee Growth suite
A forecast is only as good as the retention and expansion behind it. Chargebee Growth acts on billing data to retain and expand revenue through Plays, with Chargebee Retention available as a use case within the suite. Because Growth runs on Chargebee Billing data, its audiences and interventions use the same records that own the subscription. Learn more about the Chargebee Growth suite.
See how Chargebee meters usage, hybrid, and subscription revenue in one billing layer: explore usage-based billing.
Frequently Asked Questions
How do AI-native companies handle revenue forecasting when they have usage-based and hybrid pricing models?
They forecast by model rather than by a single blended number. Group new accounts into cohorts, apply a consumption ramp to each, model overage and prepaid credit drawdown separately, then sum the results. Companies that align pricing with their AI innovation are twice as likely to expect high growth.
How do I forecast revenue for usage-based or consumption pricing where usage varies month to month?
Forecast expected units multiplied by the rate for each pricing tier, and use a cohort ramp for new accounts that scale over time. Model a usage range rather than a point estimate, then reconcile against metered actuals each cycle. The main operational hurdles are explaining the pricing structure (22%) and building and maintaining metering infrastructure (21%), so clean metering data is the foundation of an accurate usage forecast.
How do I forecast hybrid pricing (base subscription + overages + prepaid credits)?
Sum three components forecast on their own curves: the base subscription, the forecast overage, and the modeled prepaid credit drawdown. Forecast each separately because they move on different rhythms, then stack them into a monthly total. Hybrid models reward the effort: 67% of companies using hybrid pricing expect improved margins, compared with 32% on pure usage-based pricing.
What’s the difference between MRR and ARR in subscription forecasting?
MRR measures recurring revenue monthly and ARR measures it annually. Use MRR to track month-to-month movement and catch churn or expansion early, and use ARR for annual planning, board reporting, and valuation. Usage revenue is variable rather than truly recurring, so model it on its own line and keep it out of both MRR and ARR.
How far ahead should a subscription revenue forecast look, and how often should I update it?
Forecast recurring revenue on a rolling 12-month horizon that most teams roll forward each month. The core principle is that the update cadence should match how fast your revenue drivers change. Refresh more often when a meaningful share of revenue is usage-driven and moves quickly, because variable revenue shifts faster than a flat base.
