A pricing model where customers pay only for the resources, services, or units they actually consume — no fixed fees, no minimum commitments, no paying for idle capacity. Revenue scales linearly with customer usage, aligning the vendor's economics directly with the value the customer extracts.
Also called: Consumption-based pricing, Metered billing, Pay-per-use
Section 1
How It Works
The usage-based model inverts the traditional pricing contract. Instead of asking a customer to commit to a fixed price for a fixed period — and hoping they use enough to feel it was worth it — the vendor says: use what you need, and we'll bill you for exactly that. The unit of consumption varies by industry: compute hours, API calls, kilowatt-hours, miles driven, gigabytes stored, messages sent, transactions processed. But the principle is universal: cost tracks value.
The critical insight is that this model eliminates the customer's upfront risk. There is no large purchase order to justify, no shelfware to regret, no capacity planning to get wrong. The barrier to adoption drops to near zero, which is why usage-based pricing has become the default for cloud infrastructure and is rapidly spreading into SaaS, fintech, communications, and logistics. AWS didn't win the cloud market by being the cheapest — it won by letting a two-person startup pay $14 in month one and $14 million in month forty without ever renegotiating a contract.
Monetization mechanics vary. The simplest form is pure metering: you consume X units, you pay X × unit price. More sophisticated implementations use tiered pricing (the per-unit cost decreases at higher volumes), committed-use discounts (pre-purchase a baseline at a lower rate, pay on-demand above it), or hybrid models that combine a small platform fee with metered usage on top. The best implementations make the meter visible to the customer in real time — AWS's billing dashboard, Twilio's usage logs, Datadog's per-host counters — because transparency builds trust and reduces bill shock.
VendorInfrastructure / ServiceCompute, APIs, storage, bandwidth, units
Provisions→
Metering LayerUsage TrackingReal-time measurement, billing, dashboards
Consumes→
CustomerEnd User / DeveloperPays only for actual consumption
↑Revenue = Units consumed × Per-unit price (often tiered)
The central tension in the model is revenue predictability versus growth efficiency. Usage-based companies grow faster because adoption friction is low, but their revenue is inherently volatile — tied to customer activity, seasonality, and macroeconomic conditions. A recession doesn't just slow new bookings; it shrinks existing revenue as customers consume less. This is the tradeoff every usage-based company must manage, and it explains why the most successful ones eventually layer in committed contracts or minimum spend agreements to stabilize the base.
Section 2
When It Makes Sense
Usage-based pricing is not universally superior to subscriptions or licenses. It works brilliantly under specific conditions and poorly under others. The model rewards companies that can meter consumption cleanly, serve customers with highly variable demand, and build products where usage naturally expands as the customer succeeds.
✓
Conditions for Usage-Based Success
| Condition | Why it matters |
|---|
| Variable and unpredictable demand | Customers don't know how much they'll need in advance. Fixed pricing forces them to over-provision (waste) or under-provision (risk). Usage-based pricing eliminates this guesswork. |
| Clear, measurable unit of value | The consumption metric must be intuitive and directly correlated with the value the customer receives. API calls, compute hours, and messages sent are clean. "Active users" is murkier. |
| Usage grows with customer success | The model's magic is built-in net revenue retention: as the customer's business grows, their consumption — and your revenue — grows automatically. No upsell motion required. |
| Low marginal cost of delivery | Each incremental unit consumed should cost the vendor significantly less than the price charged. Cloud infrastructure, APIs, and digital services have near-zero marginal costs at scale. |
| Developer or technical buyer | Technical buyers are comfortable with metered pricing because they understand the units. Enterprise procurement teams often prefer predictable budgets — usage-based pricing can create friction with CFOs. |
| Land-and-expand sales motion | The model excels when you can enter an organization through a small team or use case and expand organically as usage spreads. No need for top-down enterprise sales to capture initial revenue. |
| High elasticity of demand | Lower prices or zero-commitment entry should meaningfully increase the number of customers who adopt. If demand is inelastic (customers will pay any price), fixed pricing captures more surplus. |
The underlying logic is alignment: when the customer pays only for what they use, the vendor's incentive is to make the product so good that customers use more of it. This creates a virtuous cycle — better product drives more consumption, more consumption drives more revenue, more revenue funds more product investment. The companies that execute this cycle well achieve net revenue retention rates above 130%, meaning their existing customer base generates 30%+ more revenue each year without any new logos.
Section 3
When It Breaks Down
The usage-based model's greatest strength — revenue that scales with consumption — is also its greatest vulnerability. When customers consume less, revenue contracts instantly. There is no contractual floor to cushion the fall.
| Failure mode | What happens | Example |
|---|
| Revenue volatility | Revenue swings with customer activity, seasons, and macro conditions. Forecasting becomes unreliable. Public markets punish the unpredictability with lower multiples. | Snowflake's stock dropped ~15% in a single day in March 2023 after guiding for slower consumption growth amid enterprise cost optimization. |
| Bill shock and churn | Customers receive unexpectedly large bills, lose trust, and either churn or aggressively optimize usage downward. The vendor's revenue spikes then craters. | Early AWS customers regularly reported surprise bills from misconfigured services or forgotten instances running overnight. |
| Optimization as headwind | Sophisticated customers actively engineer their usage downward — better caching, query optimization, workload scheduling — shrinking the vendor's revenue without reducing the value received. | Cloud cost optimization tools like Vantage and CloudZero exist specifically to reduce AWS, GCP, and Azure spend. |
| Misaligned consumption metric | The metered unit doesn't correlate with customer value. Customers feel they're paying for inputs (API calls) rather than outcomes (problems solved). Resentment builds. |
The most dangerous failure mode is revenue volatility compounded by macro sensitivity. In a subscription model, a recession slows new bookings but existing ARR is contractually protected. In a usage-based model, a recession hits both new and existing revenue simultaneously — customers consume less across the board. This is why the 2022–2023 "cloud optimization" wave hit usage-based companies like Snowflake, MongoDB, and Datadog harder than seat-based SaaS companies. The model's elasticity, which is a feature during growth, becomes a liability during contraction.
Section 4
Key Metrics & Unit Economics
Usage-based businesses require a different measurement framework than subscription companies. The standard SaaS metrics — ARR, monthly churn rate, seat expansion — don't capture the dynamics of a consumption model. You need metrics that track the velocity and efficiency of usage, not just the presence of a contract.
Net Revenue Retention (NRR)
(Starting Revenue + Expansion − Contraction − [Churn](/mental-models/churn)) ÷ Starting Revenue
The single most important metric. Best-in-class usage-based companies achieve NRR of 130–170%, meaning existing customers generate 30–70% more revenue each year. Snowflake reported 158% NRR at its IPO. Below 110% signals weak product-market fit or a consumption ceiling.
Dollar-Based Consumption Growth
Current period consumption ÷ Prior period consumption (same cohort)
Tracks how much more (or less) an existing customer cohort consumes over time. Unlike NRR, this isolates pure usage growth from pricing changes. The leading indicator of future revenue.
Gross Margin per Unit
(Unit Price − Marginal [Cost](/mental-models/cost) of Delivery) ÷ Unit Price
Critical because the model only works if each incremental unit consumed is profitable. AWS reportedly operates at 60–70% gross margins on compute. If your marginal cost per unit is high, the model breaks at scale.
Revenue per Customer (Avg)
Total Revenue ÷ Active Customers
Tracks whether the customer base is deepening or diluting. Rising average revenue per customer signals healthy expansion. Flat or declining signals you're adding low-value customers faster than existing ones grow.
Core Revenue FormulaRevenue = Σ (Active Customers × Units Consumed per Customer × Price per Unit)
Growth = New Customer Revenue + Existing Customer Expansion − Existing Customer Contraction − Churn
Unit Economics: LTV = (Avg Monthly Revenue per Customer × Gross Margin) ÷ Monthly Revenue Churn Rate
The key levers are adoption breadth (how many customers start using the product), consumption depth (how much each customer uses over time), and unit pricing (how much you charge per unit). The best usage-based companies optimize for adoption breadth first — get as many customers as possible onto the platform at low or zero cost — then let consumption depth do the heavy lifting. Twilio's developer-first strategy exemplifies this: make it trivially easy to send your first SMS via API, then watch as that developer's company scales from 100 messages a month to 100 million.
Section 5
Competitive Dynamics
Usage-based businesses build competitive advantage differently than subscription companies. The moat is not the contract — there is no contract to lock anyone in. The moat is the integration depth, data gravity, and switching cost that accumulates as a customer's usage grows.
Consider AWS. A startup that spins up a single EC2 instance can leave tomorrow. But a company running 500 microservices across EC2, S3, Lambda, RDS, and DynamoDB — with years of operational data, custom IAM policies, and engineering workflows built around AWS primitives — faces a migration cost measured in millions of dollars and months of engineering time. The usage-based model creates switching costs through accumulated complexity, even though it promises no lock-in. This is the model's competitive paradox: the absence of contractual commitment makes adoption easy, and easy adoption leads to deep integration, and deep integration creates the strongest lock-in of all.
The model tends toward oligopoly in infrastructure markets (AWS, Azure, GCP in cloud; Twilio, Vonage, Bandwidth in communications) and fragmentation in application-layer markets where switching costs are lower. The determining factor is whether the consumption unit is commoditized (storage, compute) or differentiated (proprietary algorithms, unique data sets). Commoditized units invite price competition; differentiated units support pricing power.
Competitors typically respond to a usage-based incumbent in three ways. First, price undercutting — offering the same unit at a lower price, which works only if the challenger has structural cost advantages (Google's custom TPU chips vs. AWS's general-purpose instances). Second, bundling — packaging the usage-based service into a broader platform at a fixed price, which appeals to customers who want budget predictability (Microsoft bundling Azure credits into Enterprise Agreements). Third, vertical specialization — building a usage-based product optimized for a specific workload or industry where the generalist's pricing is inefficient (Snowflake specializing in analytics workloads vs. AWS Redshift).
Over time, the strongest usage-based companies deepen their moat by expanding the surface area of consumption — more products, more units to meter, more reasons for the customer to route workloads through the platform. AWS launched with a handful of services in 2006; it now offers over 200, each generating its own usage-based revenue stream. The platform becomes an ecosystem, and the ecosystem becomes the moat.
Section 6
Industry Variations
The usage-based model is one of the oldest pricing mechanisms in commerce — utilities have charged per kilowatt-hour for over a century — but its application has expanded dramatically with digitization. The model manifests differently depending on what's being consumed, who's consuming it, and how easily consumption can be metered.
◎
Usage-Based Variations by Industry
| Industry | Consumption unit | Key dynamics |
|---|
| Cloud infrastructure | Compute hours, GB stored, data transferred | Massive scale advantages. Gross margins 60–70%. Winner-take-most dynamics. Committed-use discounts (1–3 year reservations) hybridize the model for enterprise. AWS reportedly generated ~$91B in 2023 revenue. |
| Communications APIs | Messages sent, minutes used, API calls | Developer-first adoption. Extremely low time-to-first-dollar. Twilio charges ~$0.0079 per SMS. High volume, low unit price. Revenue concentration risk from large customers. |
| Data & analytics | Queries run, compute credits, data scanned | Usage correlates directly with business value (more queries = more insights). Snowflake's credit-based model lets customers scale compute independently of storage. NRR often exceeds 130%. |
| Utilities (electricity, water, gas) | kWh, gallons, therms | The original usage-based model. Regulated pricing. Natural monopoly dynamics. Smart meters enable real-time pricing and demand response. Low margins, high capital intensity. |
Section 7
Transition Patterns
Usage-based pricing rarely emerges in isolation. Companies typically arrive at it from simpler models and evolve toward hybrid structures as they mature and encounter the model's inherent limitations.
Evolves fromFreemiumLicensingAccess over ownership / Rental
→
Current modelUsage-based / Pay-as-you-go
→
Evolves intoSubscriptionOutcome-based / Pay-for-performanceSwitching costs / Ecosystem lock-in
Coming from: Many usage-based companies start with a freemium tier that gives away a limited amount of consumption for free — Twilio's free trial credits, AWS's
Free Tier, Datadog's free plan for up to 5 hosts. The free tier solves the adoption problem; the usage-based pricing captures value as consumption grows. Others transition from traditional licensing: Adobe moved from perpetual licenses to Creative Cloud subscriptions, and many infrastructure companies have moved from annual license fees to consumption-based pricing as cloud delivery made metering feasible.
Going to: As usage-based companies scale into the enterprise, they almost always hybridize. Snowflake offers capacity commitments. Databricks sells committed-use agreements. AWS offers Reserved Instances and Savings Plans. The pure usage-based model evolves toward a subscription floor with usage-based upside — guaranteed minimum revenue for the vendor, budget predictability for the customer. The most ambitious companies evolve further toward outcome-based pricing, where the customer pays for results (transactions processed, revenue generated) rather than inputs consumed.
Adjacent models: The closest neighbors are subscription (fixed recurring revenue), Product-as-a-Service (physical assets priced on consumption), and AI as a Service (the newest and fastest-growing usage-based category). The boundary between usage-based and subscription is increasingly blurred, with most modern SaaS companies offering some form of hybrid.
Section 8
Company Examples
Section 9
Analyst's Take
Faster Than Normal — Editorial ViewUsage-based pricing is having its moment, and for good reason. The alignment between customer value and vendor revenue is elegant. The land-and-expand motion is capital-efficient. The net revenue retention numbers can be spectacular. But I think the market is overindexing on the model's strengths and underweighting its structural weaknesses.
The dirty secret of usage-based pricing is that it transfers forecasting risk from the customer to the vendor. In a subscription model, the customer bears the risk of over-provisioning — they pay for 100 seats whether they use 60 or 100. In a usage-based model, the vendor bears the risk of under-consumption — they've built the infrastructure to serve 100 units but the customer only uses 60. This risk transfer is why customers love the model and why public-market investors are increasingly skeptical of it. Snowflake trades at a lower revenue multiple than many seat-based SaaS companies despite superior growth, largely because Wall Street can't forecast its revenue with the same confidence.
The founders I see succeeding with this model share one trait: they obsess over the consumption metric. The choice of what to meter is not a pricing decision — it's a product strategy decision. Meter the wrong thing and you create perverse incentives (customers optimizing away from your product's best features). Meter the right thing and you create a flywheel where the customer's success automatically generates your revenue. Datadog's per-host pricing works because more hosts means more infrastructure, which means the customer's business is growing. OpenAI's per-token pricing is more ambiguous — a customer who gets a better answer in fewer tokens is succeeding, but generating less revenue.
My honest read: pure usage-based pricing is a transitional state, not an end state. Almost every successful usage-based company eventually hybridizes — adding platform fees, minimum commitments, or reserved capacity pricing. The pure model is a brilliant customer acquisition strategy, but it's an incomplete business model. The companies that will win the next decade are the ones that use usage-based pricing as the wedge and then build contractual revenue on top of it, capturing the best of both worlds: the adoption velocity of pay-as-you-go and the predictability of committed contracts.
The one exception is infrastructure at true hyperscale — AWS, Azure, GCP — where the sheer volume of consumption and the diversity of the customer base create natural revenue stability even without contracts. But there are perhaps three companies on earth that operate at that scale. For everyone else, the hybrid is the destination.
Section 10
Top 5 Resources
01BookThe definitive account of Amazon's evolution, including the creation of AWS and the strategic logic behind usage-based cloud pricing. Stone details how Andy Jassy's team studied utility pricing models to design AWS's pay-per-use structure. Essential context for understanding how the most successful usage-based business in history was built.
02BookWritten by two long-tenured Amazon executives, this book explains the internal mechanisms — including the "two-pizza team" structure and PR/FAQ process — that enabled AWS to iterate on usage-based pricing at speed. The chapters on AWS's launch reveal how Amazon tested consumption pricing before it had a name for the model.
03BookTzuo, founder of Zuora (the billing platform that powers many usage-based companies), makes the case for recurring revenue models broadly — but the most valuable chapters dissect the spectrum from pure subscription to pure usage-based and the hybrid models in between. Read this for the billing and revenue recognition complexities that usage-based companies must solve.
04EssayJanz's framework for building a $100M business — hunting elephants, deer, rabbits, mice, or flies — is directly applicable to usage-based pricing strategy. The model's natural fit is the "mice" and "rabbits" segments: massive numbers of small-to-medium customers whose individual consumption is modest but whose aggregate volume is enormous. A clarifying lens for go-to-market design.
05BookWritten before AWS existed, this book by two Berkeley economists laid the theoretical groundwork for why information goods should be priced on consumption rather than production cost. Shapiro and Varian's analysis of versioning, bundling, and marginal-cost pricing remains the best intellectual framework for understanding why usage-based pricing dominates digital markets. Prescient and still relevant.
