Choosing your usage-based value metric: the layer cake pricing model

Pricing done well is about finding a shared language of value between a business and its customers. Usage-based billing gives that value a quantifiable anchor: a metric that connects what customers recognize as valuable with the economics of delivering the product. Where you start depends on how you define that value.

Many companies stall early in a pricing transformation because the shift can feel dramatic and the pressure to get the model exactly right is high. In practice, pricing is an iterative process. A pricing model matures as the company learns from customer behavior, cost data, and willingness to pay.

This guide maps several entry points into usage-based billing and offers a framework for deciding where to begin, how to learn, and where the model can develop over time.

First, understand your why

To understand where to start, it helps to identify where the pressure is coming from. Companies typically pursue usage-based billing for one of two broad reasons: to cover costs more directly as usage scales, or to find new opportunities to capture and communicate value.

Cover variable costs

LLM inference, compute, storage, data processing, and infrastructure all carry costs that can increase with product activity. Consumption pricing connects revenue more directly to the cost of delivering the service and can help protect margins as usage grows.

When a company is struggling to keep up with the infrastructure costs of rapidly scaling customers, this is often the most defensible starting point.

Monetize product expansion

A traditional seat or license fee often bundles much of a platform’s value into one price. Usage-based billing creates another path for monetization: companies can launch new capabilities, measure adoption, and allow revenue to grow with usage without creating a new SKU for every feature.

When expansion is stalling or product launches are not translating into revenue growth, monetization opportunities may be the more urgent priority.

Communicate value more clearly

Seats can be an indirect proxy for value. Usage-based pricing gives customers a more visible connection between product activity and what they pay for. It also gives companies better evidence about which capabilities customers use, value, and expand.

When sellers struggle to justify pricing or customers resist a usage-based model, clearer value communication may be the primary goal.

Where is the dominant pressure in your business coming from?

Value is part event, and part abstraction

A value metric is the unit a company uses to express what customers receive from a product. In usage-based pricing, that value metric can be equated with different kinds of measurable events.

But the events that most closely track increased usage don't always reflect how your customers understand value. For instance, a customer support team may consume tokens in order to run a chatbot that resolves customer complaints. But what they value isn't tokens, it's satisfaction.

The price you set could be based on that consumption metric, or it could involve some translating in order to balance cost control with ease of selling. There are three potential levels to this:

• A consumption metric tracks the resources required to deliver the product, such as tokens, storage, data movement, or compute time.
• An action metric tracks a meaningful product capability, such as an agent run, workflow, or document processed.
• An outcome metric tracks a result that can be evaluated against an agreed success criterion, such as an accepted document, resolved ticket, or satisfactory customer conversation.

Each metric introduces a different level of abstraction between the event being measured, the cost of delivering the product, and the value recognized by the customer. Greater abstraction can create more strategic upside, but it also requires stronger instrumentation, customer research, and evidence.

So where should you start?

Getting unblocked

Companies often get blocked on choosing a value metric because they’re trying to tackle all three goals at once. While there is a progression we often see, exactly where a company starts depends on its specific need.

Company stage

Earlier-stage companies often experience cost pressure as an existential constraint. When infrastructure costs rise faster than revenue, a consumption metric can provide an immediate way to protect margins and understand the economics of serving each customer. Larger companies generally have more latitude to test different models, manage a transition over time, and absorb short-term uncertainty while they gather evidence.

Product maturity

Product maturity determines how much evidence exists for defining value beyond consumption. A mature product may have a clear history of which capabilities drive adoption, retention, and expansion. A newer company or product line may lack that behavioral data, making a consumption metric the most credible starting point by default. The metric creates a foundation for learning while the product and customer relationship develop.

Audience

The audience also shapes which metrics customers can understand and manage. Technical, AI-savvy, or self-serve customers may be comfortable paying for tokens or credits and may actively optimize their own consumption. Less technical teams or buyers with lower tolerance for billing complexity may place greater value on predictable pricing and reliable outcomes. In those settings, an action or outcome metric may be easier to sell than a raw resource measure.

Product type

For infrastructure products that sell storage, compute, throughput, or processing capacity, the cost metric may also be the value metric. The resource being consumed is close to what the customer is purchasing. Applications and workflow products often require more interpretation because the underlying costs support a higher-level capability or outcome.

Taken together, these factors identify the most credible entry point. When the starting point is unclear, a consumption metric is usually a sound way to get off the ground. It is measurable, auditable, and capable of generating the usage data needed to support more abstract pricing later.

Applying the framework

The factors above point toward three practical entry points. The right starting point depends on the problem the business is trying to solve and the evidence available to support a metric.

Start with consumption when cost exposure is the priority

Consumption is usually the easiest place to begin because usage events and cost inputs are often already available. An AI infrastructure company may use tokens processed. A data platform may use storage, data movement, or compute time. These metrics are measurable, auditable, and closely connected to the cost of service.

For infrastructure businesses, the consumption layer may remain the primary pricing model for a long time. The resource itself is often what customers are buying, and the relationship between consumption and value is relatively direct. More abstract metrics can be introduced later for managed workflows, premium capabilities, or higher-level services.

A cost-aligned metric creates a way to recover variable costs while establishing a reliable source of usage data. It also gives customers a clear explanation of what drives the bill, provided the company defines the metric and its relationship to the product experience.

Snowflake illustrates this starting point. Its credit system is tied primarily to compute usage, while storage is priced separately. The credit acts as a proxy for infrastructure spend, and the same currency can support a growing set of capabilities without requiring a new SKU for each feature. The model prioritizes cost recovery and elasticity, which is appropriate for a product where the consumed resource is close to the value being purchased.

Start with an action metric when product expansion is the priority

An action metric may be the better starting point when the primary opportunity is incremental revenue from new capabilities.

An AI application could measure agent runs, workflows executed, or documents processed. These activities may not map perfectly to COGS because their underlying cost varies by model, context, or number of steps. They can still represent meaningful product adoption and provide a natural basis for expansion.

A company may meter the feature independently, include it in a tier, or translate several capabilities into credits. The choice depends on how customers understand the product and how much flexibility the commercial model requires.

This approach is useful when the product is expanding faster than its SKU structure. A company can launch a capability, observe adoption, and create a revenue path without assigning a separate package to every feature.

Octave provides an example of action-based pricing alongside a conventional subscription. Its primary structure is a flat annual platform subscription, while usage is metered through credits consumed when agents run. Each agent consumes a different number of credits per run based on its configuration and model settings. This structure preserves predictable platform access while creating incremental monetization for capabilities that customers adopt at different rates.

Start with outcomes when success is clear and trust is the priority

Outcome-based pricing can be an appropriate starting point when customers and suppliers share a clear definition of success.

Customer support technology provides a useful example. A company may be able to define a successful interaction through criteria such as a ticket being resolved or a customer conversation receiving a satisfactory rating. Pricing around those outcomes can reduce the perceived risk of an AI system, particularly in a market where customers remain cautious about whether automation will work reliably.

In this context, outcome-based pricing can serve as a trust mechanism. Customers pay for a result that has been accepted against an agreed standard rather than for an opaque amount of underlying model activity.

This model requires more preparation than a simple consumption metric. The success criteria must be measurable, consistent across customers, and defensible when an outcome is disputed. When those conditions are already present, outcome-based pricing may be a stronger starting point than a lower-level usage metric.

Intercom Fin demonstrates this approach in customer support. Its pricing includes charges for outcomes such as resolutions, handoffs, and disqualifications, connecting payment to a measurable result rather than to seats or raw model activity. The model is viable because support teams can evaluate whether an interaction met an operational criterion, and the outcome is meaningful to a buyer that may not want to manage the underlying model economics.

What about hybrid models?

Hybrid models remain useful when part of the value is difficult to meter directly. A seat or license fee, often paired with tiers, can capture the baseline value of access, breadth of functionality, and product improvements that do not correlate neatly with usage. Product teams invest in features, reliability, controls, reporting, and workflow improvements that may benefit customers even when consumption stays flat. Placing these capabilities in higher tiers can help capture a return on that investment without forcing every source of value into a usage event.

Usage-based layers can then price the parts of the product that are variable or readily observable. A tiered platform fee can cover access and ongoing product development, while consumption, action, or outcome metrics capture the parts of the experience that scale with customer activity. This structure also gives customers a degree of predictability while preserving room for expansion.

The right balance depends on which forms of value are easiest to measure, which costs vary with usage, and how customers prefer to buy. For many products, a hybrid model is the most practical way to connect pricing to both the product itself and the activity it enables.

The layer cake model

More and more, companies are mix and matching these metric types to build a pricing "menu" that allows them to balance all of these goals across product lines.

A company may use a platform fee for predictable access, consumption pricing for variable infrastructure costs, action-based pricing for new capabilities, and outcome-based pricing for a small set of proven workflows. These layers can apply to different products, customer segments, or parts of the same customer experience.

Hybrid pricing models can also include a seat or license fee, often organized into tiers. This fixed component captures value that is important to the customer but difficult to meter consistently, such as access, administration, collaboration, or a broad set of bundled capabilities. Usage-based layers can then price the parts of the product that vary more directly with consumption, adoption, or outcomes.

Consider an AI customer-support platform. A fixed platform fee could cover access and administration. Credits could cover general model consumption across several capabilities. Action-based charges could apply to selected workflows, while an outcome-based charge could apply when a customer conversation meets an agreed satisfaction criterion. Each component serves a different commercial purpose, and the model can evolve as the company gathers evidence about cost, adoption, and customer value.

The layer cake is therefore a portfolio structure rather than a required launch sequence. Its usefulness comes from separating the jobs that pricing needs to perform:

• The foundation manages cost exposure.
• The growth layer captures adoption of new capabilities.
• The value layer communicates outcomes customers recognize and trust.

The layer cake in practice

Few public pricing models expose all three metrics in a perfectly discrete form. More often, the layers are distributed across the product surface, customer segments, or commercial model.

HubSpot Breeze: contacts, actions, and outcomes in one credit system

HubSpot Breeze combines paid platform access with HubSpot Credits for a broad set of AI capabilities. Beneath the credit system sits a raw consumption scale based on the number of contacts. Different capabilities then draw down that shared currency at different rates: Data Agent runs represent an action-based charge, while Customer Agent conversations resolved represent an outcome-based charge. Customer Agent moved from a per-conversation charge to a lower price per resolved conversation in April 2026.

This structure shows how a layer cake can operate through one credit system. The underlying contact volume establishes a consumption baseline, while action and outcome rates create more abstract ways to price individual capabilities. HubSpot can therefore preserve a common currency while expressing different relationships between product activity, cost, and customer value.

Vercel: a varied layer cake across the product

Vercel provides an example of a layer cake that combines predictable access with usage-based expansion. Its primary structure is a tiered monthly fee, which provides a stable foundation for platform access. Selected capabilities are then metered according to the output they generate. Bot Identification, for example, is priced by the number of bot-identification results returned.

The model allows different parts of the product to carry different pricing jobs. The monthly plan supports predictable access and baseline revenue, while output-based charges create room to monetize capabilities that customers adopt at different rates. Vercel’s pricing shows how a company can add usage-based layers without replacing its core subscription structure.

Designing each layer

The following considerations help determine whether a metric is clear, defensible, and appropriate for the product.

Consumption

• Best suited to margin protection and products where the resource being consumed is close to the value being purchased.
• Common examples include tokens, storage, data movement, and compute time.
• Requires reliable measurement, clear definitions, and billing that customers can trace back to product activity.

Action

• Useful when product expansion is the priority and a capability such as an agent run or workflow is easier to recognize than the underlying resource.
• May still function as a cost proxy because actions often consume variable compute.
• Works best when metering is selective and tied to meaningful adoption rather than every countable activity.

Outcome

• Appropriate when the product has a shared, measurable success criterion, such as an accepted document, resolved ticket, or satisfactory customer conversation.
  Can communicate value and build trust, especially when customers are skeptical of AI or do not want to manage model economics.
  Requires strong instrumentation, customer research, consistent criteria, and a process for handling disputed outcomes.

Choosing a starting point and growing deliberately

The monetization maturity curve is best understood as a framework for choosing among different starting points. Consumption is often the most accessible path to margin protection. Action metrics create opportunities to monetize product expansion. Outcome metrics can communicate trusted value when success is measurable.

A company does not need to adopt every layer. It needs to understand the economic purpose of each one, the evidence required to support it, and the role it should play across products and segments.

Over time, the model can become more sophisticated. A cost metric can support new feature pricing. Feature adoption data can reveal which outcomes are worth measuring. Proven outcomes can then sit alongside consumption and credits in a hybrid commercial model.

Pricing maturity is therefore a combination of hierarchy and composition. The hierarchy reflects increasing complexity and customer understanding. The layer cake reflects the ability to combine pricing approaches when different parts of the business have different economic needs.