.png)
Frameworks, core principles and top case studies for SaaS pricing, learnt and refined over 28+ years of SaaS-monetization experience.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
SaaS Pricing Models the Complete 2025 Guide
Every SaaS pricing conversation eventually reaches the same hard question: what should the customer pay for? The apparent choices - seats, usage, credits, capacity, transactions, outcomes - are not interchangeable ways to collect revenue. Each tells the buyer what the product is, determines how finance forecasts spend, and sets the limits on expansion.
The issue became sharper in 2025 as AI moved from embedded assistance toward software that can complete parts of a job. Salesforce entered 2025 with Agentforce priced from $2 per conversation, while HubSpot moved Breeze Customer Agent to credits for eligible Pro and Enterprise customers beginning June 2, 2025. Those moves mattered because they separated the price of the human-facing platform from the cost and value of automated work.
Monetizely’s position is clear: every SaaS offer needs one primary pricing metric that reflects how customers receive value. Seats should remain the default where people remain central to the work; consumption should govern measurable resource use; and outcome pricing belongs only where an agent performs a defined job with limited human intervention and auditable results.
A pricing model succeeds when the buyer can connect the invoice to a real unit of value
A pricing model has three parts. The package defines what the customer receives. The metric defines what the customer is charged for. The rate defines how much that unit costs.
Leaders often start with the rate because it is visible and easy to debate. Yet moving a price from $50 to $65 does not fix a model that charges per user for a product whose cost and value grow with data volume, transactions, or automated resolutions. Snowflake’s fiscal 2025 investor materials make the point in a mature form: its platform charges for compute, storage, and data transfer consumption, while customers typically commit and pay in advance.
The comparison below separates the four structures that buyers most often encounter.
| Primary structure | What the customer pays for | Best fit | Main risk | B2B SaaS signal |
|---|---|---|---|---|
| Per seat | A named or active user | A person uses the product repeatedly to do a job | Shelfware when many licensed users rarely engage | Cursor’s tiers distinguish individual, team, and enterprise needs through collaboration, administration, and security features. |
| Usage-based | A measurable resource such as hosts, data, API calls, storage, or credits | More use creates more value and more vendor cost | Surprise invoices if usage is hard to forecast | Snowflake prices around compute, storage, and data transfer. |
| Outcome-based | A completed, defined result such as a conversation resolved or claim processed | Software performs a bounded task with a verifiable result | Commercial disputes if the result is ambiguous | Salesforce announced Agentforce pricing from $2 per conversation in October 2024, with further capabilities scheduled for February 2025. |
| Platform fee plus one variable metric | Access, control, and a single value-linked unit | Enterprise software where governance is valuable but automated work varies | Confusion when the fixed fee and variable fee each try to explain the product | HubSpot paired seat-based subscriptions with credits for Breeze Customer Agent in 2025. |
The table points to a practical rule: a product can have more than one charge, but it cannot have more than one commercial story.
Monetizely’s 5-Step Pricing Framework addresses the common failure of treating pricing as a rate-card project. The sequence starts with goals and segmentation, then moves to packaging, pricing metric, price points, and operationalization. Each step narrows the next. A company first decides which customers it intends to win and why; then builds offers for those buyers; then selects the unit that best matches value and economics; then sets rates; and finally makes billing, reporting, sales compensation, and renewals work in practice. The logic is developed further in Monetizing Agentic AI.
The sequence matters because segmentation is not merely a marketing exercise. A solo developer, a 30-person engineering team, and a global bank may all want faster code, but they do not buy the same controls, support, or contract terms. Cursor’s public example illustrates the point: its core job remains coding faster, while the package expands from individual use toward organization-level administration and enterprise governance.
The framework shows why “usage versus seats” is the wrong opening question. The more useful question is: which customer segment receives value through which repeatable unit?
Per-seat pricing is often dismissed as legacy SaaS. That is a mistake. Seats are highly effective when access to the product is the source of value, the user is identifiable, and the buyer already organizes budgets around employee roles.
Consider a sales platform used daily by account executives. The value is not the number of records viewed or workflow steps clicked. It comes from each seller having reliable access to customer history, forecasting, pipeline management, and collaboration. A seat is simple to approve, easy to forecast, and aligned with the customer’s staffing plan.
The same logic applies to much of AI-assisted software. A coding assistant that helps an engineer write, test, and understand code still has a human at the center of the job. Cursor’s package design reflects that reality: the product can be tiered around team administration and enterprise controls without turning every model interaction into a customer invoice.
Seat pricing should therefore remain the primary meter when three conditions hold:
- A person remains responsible for the work and uses the product regularly.
- Added users create a clear increase in customer value.
- Heavy use does not create an uncontrolled cost burden for the vendor.
The third condition now deserves more scrutiny. AI inference can make an unlimited seat economically unsound for a small group of extreme users. The answer is not to abandon the seat automatically. A better architecture keeps the seat as the primary metric, includes a reasonable allowance, and charges a transparent overage only when use crosses a clear economic threshold.
Usage pricing works when the thing being measured is close to the thing the customer values. Snowflake is the cleanest large-scale example. A customer that runs more compute, stores more data, and transfers more data is using more of the platform. Snowflake’s fiscal 2025 materials explicitly tie its pricing model to compute, storage, and data transfer resources, while noting that customers generally pay annually in advance for capacity contracts.
Datadog offers a second useful pattern. Its public pricing page, accessed September 7, 2026, prices Infrastructure Pro per monitored host per month and separately prices other units such as containers, metrics, events, and tests. Its billing documentation also explains a committed minimum with hourly overage for host-based products, which suits changing environments such as autoscaling fleets.
Neither company charges “for value” in the abstract. Each charges for a unit the customer can measure, manage, and connect to its own operating footprint.
| If the product’s value rises with… | Primary metric to test first | Why the buyer can accept it | Watch for |
|---|---|---|---|
| Number of employees using a workflow | Seat | Headcount is visible in annual planning | Inactive licenses and low adoption |
| Data processed, stored, or transferred | Consumption | More data activity uses more platform resources | Unplanned spikes and unclear cost controls |
| Hosts, containers, or service instances | Resource usage | The customer operates and budgets around infrastructure | Double charging across linked products |
| Completed automated customer interactions | Outcome or credit | A defined business task has been performed | Disagreement over what counts as completion |
| A broad enterprise platform plus variable automated work | Platform fee plus one variable metric | The fixed fee covers access and control; the variable unit reflects scalable activity | Layering several usage metrics into one opaque invoice |
The lesson is straightforward: consumption pricing is not inherently modern or customer-friendly. It is useful only when the customer can actively influence the meter and see why usage rose.
Agentic software changes the analysis because the buyer may no longer be paying for a person’s access to a tool. The buyer may be paying for the work completed without that person.
The Agentic Monetization Spectrum, or AMS, provides a disciplined way to judge that shift. It assesses an agent across three dimensions: zero-human ability, meaning how much human work remains; operational domain, meaning whether the agent handles one task, one function, or several functions; and output/cost ratio, meaning whether the value created grows faster than the cost to run the system. Higher scores move the pricing center of gravity away from seats and toward usage or outcomes.
For practical use, we score each dimension from 1 to 3: small, medium, and large. The total is not a substitute for judgment. It is a forcing mechanism against claiming that every chatbot, copilot, or workflow automation deserves outcome pricing.
| Product or archetype | Zero-human ability | Operational domain | Output/cost ratio | AMS score | Primary meter indicated |
|---|---|---|---|---|---|
| Cursor-style coding assistant | 1 | 2 | 2 | 5 | Seat, with a capacity guardrail for extreme use |
| Devin-style coding agent | 2 | 2 | 1 | 5 | Credits or compute-linked usage while results still need frequent review |
| HubSpot Breeze Customer Agent | 2 | 2 | 2 | 6 | Credits tied to automated service activity |
| Salesforce Agentforce service agent | 2 | 2 | 2 | 6 | Conversation or defined-action pricing |
| Sierra-style enterprise service agent | 3 | 3 | 2 | 8 | Outcome pricing, with a platform commitment for enterprise controls |
The score explains why similar-looking products need different commercial designs. Cursor and Devin may both involve AI coding, but their pricing should diverge if one primarily assists a developer while the other takes on tickets that require compute and later review. Monetizely’s analysis of their packages reaches the same strategic distinction: a low-friction plan needs enough capacity for a real evaluation, while enterprise packages need to reflect deployment, security, and workflow depth.
HubSpot’s 2025 decision is especially instructive. The company provided Pro customers with 3,000 monthly credits and Enterprise customers with 5,000 credits, then offered additional capacity from $10 per 1,000 credits for Breeze Customer Agent. The structure preserves a predictable subscription while making high-volume automated activity visible and monetizable.
Enterprise buyers often need stable budgets, governance, security, integrations, implementation support, and auditability. Those needs justify a platform fee. They do not justify using a platform fee to hide an otherwise weak pricing metric.
The disciplined architecture is a platform commitment plus one primary variable unit. In an autonomous customer-service product, the primary unit may be a resolved interaction. In a data platform, it may be consumption. In an AI coding product, it may remain a seat, with capacity limits protecting margin.
The following scenario shows how a buyer should evaluate an outcome-based offer over three years. The key question is not the initial annual commitment. It is what the buyer will actually pay as automated volume rises.
| Annual resolved interactions | Contract structure | Annual spend | Three-year spend if volume stays constant |
|---|---|---|---|
| 750,000 | 1 million committed resolutions at $1.10 each | $1.10 million | $3.30 million |
| 1,000,000 | 1 million committed resolutions at $1.10 each | $1.10 million | $3.30 million |
| 1,500,000 | 1 million committed resolutions at $1.10, then $1.30 per overage resolution | $1.75 million | $5.25 million |
| 2,000,000 | 1 million committed resolutions at $1.10, then $1.30 per overage resolution | $2.40 million | $7.20 million |
The table reveals what sales presentations often obscure: the buyer’s downside begins below the commitment, while the vendor’s upside begins above it. A responsible price design makes both visible before signature.
A strong metric cannot survive if product telemetry, billing, sales quoting, and customer reporting count different things. Monetizely’s view is that operationalization is not back-office cleanup. It is part of the pricing decision itself.
Salesforce’s pricing evolution underscores the challenge. The company initially announced Agentforce from $2 per conversation, then in June 2025 announced per-user SKUs with unlimited actions for employee-facing agents. That is not inconsistency. It reflects different products, users, and value patterns: employee-facing assistance can remain tied to the user, while customer-facing automation can be tied to interactions.
Before release, operators should require the following controls.
| Control | What must be true | Customer-facing proof |
|---|---|---|
| Meter definition | Product, finance, and legal use the same definition of a billable unit | A one-page definition with examples and exclusions |
| Usage visibility | Customers can see consumption before the invoice arrives | In-product dashboard, alerts, and downloadable activity logs |
| Overage rules | Thresholds, rates, and approval paths are explicit | Spend caps, notification settings, and committed-use options |
| Sales compensation | Reps are paid for durable recurring revenue, not just large commitments | Compensation that recognizes adoption and healthy expansion |
| Renewal review | The vendor can explain change in spend with product evidence | A usage-to-value review 90 to 120 days before renewal |
The operational standard should be demanding. If a customer cannot predict, monitor, and explain a charge to its CFO, the company has not created a sophisticated model. It has created friction.
The market does not need more companies claiming to sell outcomes while billing for opaque credits. Nor does it need reflexive per-seat models that ignore AI cost and autonomous work. The durable answer is more exacting: select the metric customers already recognize as the unit of value, then build packages and contract terms around that unit.
For most conventional SaaS, the answer remains seats because people still perform the work. For infrastructure and data software, usage should lead because resource consumption is the product’s natural unit. For agents, the threshold is higher: outcome pricing is earned only when the agent does a defined job with limited human intervention and a result both parties can verify.
What operators should do next:
Build a three-year price architecture before publishing a list price. Model low, expected, and high adoption cases for each target segment, including committed spend, overage, discounts, and services.
Create separate economics for new customers and migrations. Existing customers should not be forced into a new meter without a clear comparison of prior spend, expected spend, and the timing of any protection period.
Run shadow quotes on the last 20 closed-won and 20 closed-lost deals. Calculate what each account would have paid under the proposed model, then inspect where the model changes the sales motion or creates buyer resistance.
Make the pricing owner accountable for adoption quality, not booked ARR alone. Track active seats, credit use, overage incidence, discount depth, gross margin, and renewal outcomes by segment.
Treat every new meter as a product launch. Give customers an estimator, spend alerts, invoice explanations, and a named path to resolve disputes before the first bill lands.
Assumptions: The three-year outcome-pricing exhibit is a modeled contract example in U.S. dollars, excluding taxes, implementation fees, and negotiated discounts. It is intended to show spend sensitivity, not market pricing benchmarks.
Sources
Monetizing Agentic AI: https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
Monetizely, “Goals and Segmentation,” “Packaging,” “Choosing the Right Pricing Metric,” “Finding the Right Price Points,” “Operationalizing Agentic AI Pricing,” and related company examples, accessed September 7, 2026: https://www.getmonetizely.com/monetizing-agentic-ai-book-saas/step-1-goals-and-segmentation https://www.getmonetizely.com/monetizing-agentic-ai-book-saas/step-2-packaging-designing-offers-that-fit https://www.getmonetizely.com/monetizing-agentic-ai-book-saas/step-3-choosing-the-right-pricing-metric https://www.getmonetizely.com/monetizing-agentic-ai-book-saas/step-4-finding-the-right-price-points https://www.getmonetizely.com/monetizing-agentic-ai-book-saas/step-5-operationalizing-agentic-ai-pricing https://www.getmonetizely.com/monetizing-agentic-ai-book-saas/the-agentic-monetization-spectrum https://www.getmonetizely.com/monetizing-agentic-ai-book-saas/cursor-segments-understood-capabilities-mapped-well https://www.getmonetizely.com/monetizing-agentic-ai-book-saas/devin-right-segments-wrong-sized-packages https://www.getmonetizely.com/monetizing-agentic-ai-book-saas/sierra-ai-three-segments-one-served
HubSpot, “Expanding Access to Breeze Customer Agent with HubSpot Credits,” May 8, 2025: https://ir.hubspot.com/news-releases/news-release-details/hubspot-credits
Salesforce, “Salesforce Unveils Agentforce,” September 12, 2024, and “Salesforce Launches Agentforce 3,” June 11, 2025: https://investor.salesforce.com/news/news-details/2024/Salesforce-Unveils-AgentforceWhat-AI-Was-Meant-to-Be/default.aspx https://investor.salesforce.com/news/news-details/2025/Salesforce-Launches-Agentforce-3-to-Solve-the-Biggest-Blockers-to-Scaling-AI-Agents-Visibility-and-Control/default.aspx
Snowflake, Q4 Fiscal 2025 Investor Presentation, February 26, 2025: https://investors.snowflake.com/files/docfinancials/2025/q4/Q4-FY2025-Investor-PresentationvFF.pdf
Datadog, “Pricing Comparison” and “Pricing” billing documentation, accessed September 7, 2026: https://www.datadoghq.com/pricing/list/ https://docs.datadoghq.com/account_management/billing/pricing/