Adoption jumped from 30% to 85% in five years. Here is what usage-based pricing actually is, the real tradeoffs, hybrid models, and how Snowflake, Twilio, and OpenAI do it in practice.
Written for SaaS founders and pricing teams deciding whether to move away from flat or per-seat pricing, especially as AI features make consumption costs harder to ignore.
Usage-based pricing charges customers according to how much of a product they actually consume, API calls, compute time, tokens, messages, rather than a flat fee. Adoption has risen from about 30% of SaaS companies in 2019 to 85% by 2024, per Flexprice, driven largely by AI features where compute cost varies enormously per customer.
It is not a universal fix. Pure usage-based pricing creates real unpredictability, 78% of IT leaders report unexpected charges from consumption or AI pricing models, per L.E.K. Consulting, which is why 43% of SaaS companies now run a hybrid model, a base fee plus usage, instead of pure consumption pricing.
Usage-based pricing, sometimes called consumption-based pricing, ties what a customer pays to how much of the product they actually use. Instead of a flat $99 a month regardless of activity, a customer might pay for every API call, every gigabyte processed, every message sent, or every AI generation completed.
The model is not new, utilities have billed this way for a century, but it has become the dominant pattern in SaaS specifically because AI features introduced real, highly variable compute costs per customer that flat pricing simply could not absorb without either overcharging light users or losing money on heavy ones.
In one sentence
Usage-based pricing makes the price move with the value and the cost, so a customer who uses ten times more of the product roughly pays ten times more, instead of paying the same flat fee as someone who barely uses it.
Six numbers that show usage-based pricing moved from a niche experiment to the default expectation.
85%
Of SaaS companies use some form of usage-based pricing today, up from 30% in 2019, per Flexprice
77%
Of the largest software companies now include a usage-based pricing component, per Metronome
64%
Of Forbes Next Billion-Dollar Startups have adopted usage-based pricing
78%
Of companies that adopted usage-based pricing did so within 5 years of founding
43%
Of SaaS companies now use a hybrid model that blends a seat or platform fee with usage
78%
Of IT leaders report unexpected charges from consumption or AI pricing models, per L.E.K. Consulting
Metronome’s State of Usage-Based Pricing research also found that roughly half of companies that adopt usage-based pricing do so within their first two years, and 78% do so within five years, suggesting it is increasingly a launch-time decision rather than a later pivot.
| Model | How It Works | Predictability | Scales With Value | Best For |
|---|---|---|---|---|
| Flat / Fixed Fee | One price regardless of usage | Highest for the buyer | Poor, heavy and light users pay the same | Simple tools with low variance in usage |
| Per-Seat / Per-User | Price scales with number of users | High | Moderate, breaks down when usage varies independent of headcount | Collaboration and productivity tools |
| Usage-Based / Consumption | Price scales directly with a measured unit of consumption | Lowest for the buyer without safeguards | Strong, price tracks value delivered | Infrastructure, APIs, AI features, compute-heavy products |
| Hybrid (Base + Usage) | A platform or seat fee plus a usage or credit component on top | Moderate, base fee anchors the floor | Strong, with revenue predictability for the vendor | Most SaaS products adding AI or compute-heavy features in 2026 |
| Outcome-Based | Price tied to a completed result or task rather than raw consumption | Variable, depends on outcome definition | Strongest alignment, but hardest to implement cleanly | Agentic AI features priced per completed task |
Neither list is theoretical, both come directly from 2026 adoption and complaint data.
43% of SaaS companies now blend a base fee with a usage component, making hybrid pricing the practical default for new launches rather than the exception. Here are the three most common patterns.
A fixed monthly or annual fee covers access and a baseline usage allowance, with additional usage billed on top or drawn from a prepaid credit pool. This is the most common hybrid pattern in 2026.
Core seats are billed per user as before, while newer AI features are metered separately, often by tokens, generations, or completed tasks, so AI compute cost does not silently erode the base plan margin.
Customers pick a tier with an included usage allotment, and usage beyond that allotment bills at a per-unit overage rate, giving predictability up to a point and then scaling with real consumption.
| Industry / Product Type | Fit | Why |
|---|---|---|
| Infrastructure and Dev Tools | Excellent | Compute, storage, and bandwidth costs vary enormously per customer, making usage-based pricing the natural default, as seen with Snowflake. |
| Communications APIs | Excellent | Every message, call, or notification has a near-identical marginal cost, making per-unit pricing simple and fair, as seen with Twilio. |
| AI and LLM-Powered Features | Excellent | Token and compute costs scale directly with usage intensity, and flat pricing risks losing money on heavy users, as seen with OpenAI. |
| Project Management and Collaboration Tools | Moderate | Usage does not always track value cleanly, per-seat or hybrid models often fit better than pure consumption pricing. |
| Vertical SaaS with Stable, Predictable Workflows | Weak | When usage barely varies month to month, a flat or per-seat model is simpler for both vendor and customer with little downside. |
Snowflake bills compute usage by the second, with a 60-second minimum per query, according to Snowflake’s own cost documentation. Customers pay in credits that convert to compute time, meaning a heavier analytical workload costs proportionally more without a separate plan change.
Twilio publishes exact per-message rates for SMS and MMS, for example $0.0083 per outbound or inbound US SMS message, $0.022 per outbound MMS, and $0.0165 per inbound MMS. Every message sent through the platform is billed individually, making cost directly proportional to volume.
OpenAI’s API prices per million tokens, with separate rates for input and output tokens across its model lineup. The Batch API offers roughly a 50% discount for asynchronous, non-urgent workloads, and prompt caching can cut costs by up to 75% on repeated context, both according to OpenAI’s published pricing documentation.
Whatever pricing model you land on, buyers still have to find you first. Tools like MediaFast help SaaS founders show up in the Reddit threads where people are already comparing pricing models like these, before a competitor gets there first.
MediaFast finds the Reddit threads where SaaS buyers are actively debating pricing models, so you can join the conversation before a competitor does.
Seven steps, roughly in order, from picking a metric to reviewing it after launch.
Pick a pricing metric that tracks customer value, not just vendor cost
The best usage metric is something the customer already understands as valuable, API calls, active seats plus usage, GB processed, not an internal cost proxy that feels arbitrary to the buyer.
Build real, auditable metering before you price anything
Usage-based pricing lives or dies on billing accuracy. Customers who feel a usage bill is wrong lose trust immediately, so metering has to be built and tested before the pricing model goes live.
Add a base fee or minimum commitment for revenue predictability
Pure usage-based pricing with zero floor makes vendor revenue volatile. A base fee, minimum commitment, or credit pack gives you a predictable revenue floor while keeping upside tied to usage.
Give customers visibility and control over their own spend
Real-time usage dashboards, spend alerts, and hard or soft spend caps directly address the unexpected-charges objection that 78% of IT leaders report today.
Model a handful of real customer usage patterns before launch
Run your lightest, median, and heaviest existing customers through the new pricing model before shipping it, to catch cases where the new price would be wildly higher or lower than what they pay today.
Plan the migration path for existing customers carefully
Grandfathering, transition credits, or a parallel-run period reduce churn risk when moving an existing customer base from flat or per-seat pricing to a usage-based or hybrid model.
Revisit the pricing metric as the product evolves
The right usage metric at launch is not always the right one two years later. Companies that review pricing regularly grow roughly 25% faster than companies that set a price once and leave it, per Zylos Research.
Once your usage-based or hybrid model is set, the harder question is often what number to actually charge. The how much should I charge for my SaaS guide covers that framework in detail.
| Ask | Good Metric Example | Bad Metric Example |
|---|---|---|
| Does the customer already understand this unit as valuable? | API calls, GB stored, active users | Internal compute cost, server hours the customer never sees |
| Does it scale predictably with the value the customer gets? | Emails sent for an email tool, leads found for a lead-gen tool | A metric that can spike due to a bug or retry loop unrelated to value |
| Can you measure it accurately and cheaply at scale? | Events already logged in your system | A metric requiring expensive new instrumentation to track reliably |
| Is it hard for the customer to game or avoid? | Usage tied to real outcomes | A metric customers can trivially reduce without reducing actual usage |
The single biggest driver of the "unexpected charges" objection is unclear pricing page copy. This template addresses cost, predictability, and control up front instead of burying them in a support article.
[Base Plan] $X/month Includes [Y units] of [metric] per month Additional usage: $Z per [unit] beyond your included allotment What this means for you: - Track your usage in real time from your dashboard - Set a spend alert or hard cap, no surprise bills - Overage only applies past your included allotment - Downgrade or upgrade anytime as your usage changes
It is not automatically the right model for every SaaS product.
Infrastructure, compute, and AI-heavy products have real, variable costs that flat pricing cannot absorb fairly across light and heavy users.
API calls, messages sent, or records processed map naturally to the value a customer receives, making the price feel fair rather than arbitrary.
Usage-based entry points let customers start small and grow into higher spend naturally, rather than facing a large fixed cost on day one.
Your product delivers roughly constant value regardless of usage volume, where flat or per-seat pricing is simpler and fairer
You lack the engineering resources to build accurate, real-time metering and billing infrastructure
Your buyers are budget-constrained enterprises that need to lock in a fixed, predictable annual cost for procurement
Each of these is a leading cause of churn or revenue volatility after a usage-based launch.
If customers cannot reasonably estimate their bill in advance, procurement teams will push back hard, and existing customers will churn out of frustration even if the product delivers value.
Billing bugs discovered after customers are already being charged destroy trust fast. Run the new metering silently against real usage before turning on billing.
The single biggest driver of the "unexpected charges" objection is a lack of real-time visibility, not the pricing model itself. A dashboard and alert system solves most of this problem cheaply.
A metric that makes sense for a compute-heavy infrastructure product may not fit a lighter-weight SaaS tool at all. Usage-based pricing has to be derived from your own value delivery, not borrowed wholesale.
Pure usage pricing with no floor makes forecasting extremely difficult for the vendor and often for the customer’s own budgeting too. Most successful 2026 implementations are hybrid for this reason.
Moving existing customers to usage-based pricing without a transition plan, grandfathering, or credits is one of the fastest ways to spike churn right when the new model launches.
A usage-based account can shrink its spend for months before formally canceling. Treating declining usage as a routine metric rather than an early churn warning misses the chance to intervene.
Not automatically. Without a base fee, a quiet month for a customer means a quiet month of revenue for you too. Most successful 2026 implementations pair usage pricing with a base fee specifically to avoid this.
Usage-based components now appear in a wide range of SaaS categories, especially anywhere an AI feature has been added, not just classic infrastructure products.
The complaint data is really about unpredictability and lack of visibility, not the model itself. Customers with clear dashboards and spend caps generally accept usage-based pricing as fair, since it means they are not overpaying for capacity they do not use.
The concrete boxes to check before flipping usage-based billing on for real customers.
Metering is built, tested, and validated against real usage before billing goes live
A base fee or minimum commitment gives revenue a predictable floor
Customers can see real-time usage and projected spend in a dashboard
Spend alerts or hard caps exist so no customer gets a surprise bill
Existing customers have a clear migration path with grandfathering or transition credits
The pricing metric has been checked against your lightest, median, and heaviest customers
A review cadence is scheduled to revisit the metric as the product evolves
A pricing model where the amount a customer pays scales directly with a measured unit of consumption, such as API calls, compute time, or completed tasks.
The underlying system that tracks and records exactly how much of a usage metric each customer consumes, the foundation any usage-based pricing model depends on.
A model combining a fixed base fee, often for seats or platform access, with a usage or credit-based component on top.
Usage beyond an included allotment in a tiered plan, typically billed at a per-unit rate once the allotment is exceeded.
A pricing model where the customer pays only when a defined result or completed task occurs, rather than for raw consumption.
A hard or soft limit a customer can set on their own usage-based bill, used to address the unpredictability objection to consumption pricing.
A prepaid balance of usage credits a customer draws down as they consume the product, common in AI and API-based pricing models.
The specific unit a usage-based model charges for, such as tokens, seats plus usage, GB processed, or completed workflows.
The percentage of recurring revenue retained from existing customers over a period, including expansion and contraction, a key health metric for usage-based pricing models.
A reduced usage rate offered in exchange for a customer committing to a minimum spend or usage level over a set period, common in cloud and infrastructure pricing.
The primary sources behind the adoption stats, real-world pricing examples, and framework on this page.
Flexprice
The source for the 30% to 85% usage-based pricing adoption shift between 2019 and 2024.
Snowflake Documentation
Snowflake’s official documentation on per-second, 60-second-minimum credit billing for compute.
Twilio
Twilio’s published per-message rates, a real-world example of granular usage-based billing.
OpenAI
OpenAI’s per-token pricing model, including batch and prompt-caching discounts.
Three illustrative composites showing how hybrid usage-based pricing typically plays out across different customer sizes, not case studies of specific named companies.
Pays close to the base fee alone, with usage staying inside the included allotment most months. Low barrier to entry keeps this account from churning over cost, even though the usage-based ceiling has never been tested.
Regularly exceeds the included allotment during busy periods and pays overage charges, but the real-time dashboard and spend alerts prevent the charges from ever feeling like a surprise.
Usage volume justifies negotiating a committed-use discount, trading a minimum spend commitment for a reduced per-unit rate, a common pattern once an account’s usage becomes predictable at scale.
Usage-based pricing changes the shape of churn risk rather than removing it. A customer on a flat plan either stays or leaves, but a customer on usage-based pricing can quietly shrink their usage, and therefore their bill, long before they formally cancel. That slow shrinkage is often a leading indicator that traditional logo-churn metrics miss entirely.
The practical fix is to track net revenue retention and usage trend per account, not just logo counts. A shrinking usage trend on an account is an early warning sign worth a proactive customer success conversation, well before the account reaches zero usage and formally churns.
Usage-based pricing went from a niche infrastructure pattern to the default expectation for SaaS in five years, 30% adoption in 2019 to 85% by 2024. The companies that get it right pair it with a base fee for revenue predictability and real spend visibility for the customer, since 78% of IT leaders have already been burned by unexpected consumption charges.
Pure usage-based pricing is not the safe default, hybrid is, at 43% adoption and rising. Pick a metric your customers already understand as valuable, build metering before you bill anyone, and revisit the model as your product changes.
Pricing is only half the growth equation, distribution is the other half.
The questions pricing and growth teams ask most before switching to consumption-based billing.
Usage-based pricing, also called consumption-based pricing, charges customers based on how much of a product they actually use, API calls, compute time, messages sent, tokens processed, rather than a flat monthly fee or a per-seat charge. The price scales up or down with real consumption instead of staying fixed regardless of usage.
Very fast. Usage-based pricing adoption among SaaS companies rose from about 30% in 2019 to 85% by 2024, per Flexprice, and Metronome’s State of Usage-Based Pricing research found 77% of the largest software companies and 64% of Forbes Next Billion-Dollar Startups now include a usage-based component in their pricing.
Unpredictability, for both sides. Vendors lose some revenue forecasting certainty without a base fee, and customers report unexpected charges as a top objection, 78% of IT leaders have experienced this with consumption or AI pricing models, per L.E.K. Consulting. Real-time spend dashboards and spend caps are the standard fix.
Most successful 2026 implementations are hybrid: a base platform or seat fee for revenue predictability, plus a usage or credit component on top for value alignment. 43% of SaaS companies now use this blended approach, making it the practical default rather than pure consumption pricing for most products.
Snowflake bills compute by the second with a 60-second minimum per Snowflake’s own documentation, Twilio charges per message sent with published per-unit SMS and MMS rates, and OpenAI’s API prices per token with separate rates for input and output plus discounts for batch processing and prompt caching.
Pick a metric the customer already understands as valuable, not an internal cost proxy. It should scale predictably with the value delivered, be measurable accurately at low cost, and be hard for customers to game or avoid without actually reducing their real usage of the product.
