.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 Benchmark Study 2025 Insights from 100 Companies
Pricing appears simple from a distance: select a metric, set a number, publish a page. Our 2025 benchmark of 103 B2B SaaS companies found a very different reality. The median entry plan sat at $29 per user per month, usage-based pricing appeared in 43% of the companies reviewed, and 61% used some form of multi-part price. Enterprise ACV grew 23% on average, even as buyers demanded more clarity about what drove that growth.
The stakes are high because a pricing model does more than produce revenue. It tells customers what the product is worth, directs product investment, shapes sales behavior, and determines whether expansion feels earned or imposed. A company that charges for the wrong thing can grow bookings while building resentment, discount pressure, and renewal risk beneath the surface.
Monetizely's position is clear: SaaS companies should choose one primary meter that buyers can connect to value, package that meter for distinct customer segments, and use seats, platform fees, commitments, or limits only to support that logic. The 2025 evidence does not support adding usage charges because they are fashionable. It supports charging more precisely when the product creates more measurable value.
The market adopted variable pricing without abandoning familiar buying habits
The central signal from the benchmark is often misread. Seats did not disappear in 2025. Instead, the seat became less likely to stand alone. SaaS companies increasingly combined it with a second factor: contacts, transactions, data volume, resolutions, credits, or platform access.
That shift matters because a seat is an effective meter only when the buyer experiences value through an individual user. Jira is a clear example. A project manager, engineer, or program lead uses the product directly, so a per-user model remains intuitive. Yet even Jira's broader pricing logic ties higher plans to more automation and AI allowances, recognizing that activity and capability rise beyond headcount alone.
The benchmark data makes the pattern visible.
Exhibit 1. The 2025 benchmark shows more pricing layers, not a wholesale move away from subscriptions
| Benchmark signal | 2025 result | Strategic meaning |
|---|---|---|
| Median entry-level price | $29 per user per month | Low-friction entry remains important for self-serve acquisition |
| Average annual price increase | 8% to 12% | Vendors sought revenue gains without relying only on new logos |
| Usage-based pricing adoption | 43% | More companies linked expansion to measurable consumption |
| Hybrid pricing adoption | 61% | More contracts combined a predictable base with a variable element |
| Enterprise ACV growth | 23% | Larger accounts accepted higher spend when packaging matched broader needs |
| Freemium adoption | 38% | Free access remained a selective acquisition tool, not a universal default |
Source: Monetizely’s 103-company benchmark, published December 22, 2025.
The table does not say that every company needs several meters. It says that the market is moving away from one-size-fits-all pricing. Buyers will accept complexity when each charge answers a distinct question: access, scale, or delivered work.
Snowflake offers a useful contrast to classic seat-based SaaS. As of January 31, 2025, the company priced its platform around consumption of compute, storage, and data transfer, while typically billing capacity contracts annually in advance. The customer receives budget visibility through a commitment, but the core meter remains resource use.
Datadog follows a related logic. Its February 18, 2026 Form 10-K described pricing tailored to customers’ usage needs and identified increased usage, broader deployment, and adoption of more products as expansion paths across its 32,700 customers as of December 31, 2025.
The lesson is not “move to consumption.” Snowflake and Datadog sell software whose economic value rises as customers process more data, monitor more infrastructure, and run more workloads. Their meter works because it follows the work.
Segment design determines whether price complexity creates value or friction
Monetizely's 5-Step Pricing Framework, developed in Monetizing Agentic AI, starts with Goals and Segmentation, moves through Packaging, Pricing Metric, Rate Setting, and Operationalization. The order matters. Goals establish whether the company is pursuing adoption, ARR growth, margin protection, or a move upmarket. Segmentation identifies groups with different needs and willingness to pay. Packaging turns those differences into offers. Only then should leaders select the metric, set rates, and build the systems that meter, bill, and report on the model.
Many teams reverse that sequence. They begin with a question such as, “Should we charge per seat or per transaction?” The better question is, “Which buyer gets value in a different way, and what offer would make that difference visible?”
The benchmark shows wide gaps across customer segments.
Exhibit 2. List-price benchmarks rise sharply as buying complexity increases
| Customer focus | Entry or team benchmark | Higher-tier benchmark | What customers are buying |
|---|---|---|---|
| SMB SaaS | $15 per user per month for Starter | $65 per user per month for Business | Core functionality, simple collaboration, low commitment |
| Mid-market SaaS | $49 per user per month for Team | $89 per user per month for Business | More automation, controls, integrations, and support |
| Enterprise SaaS | $47,000 median ACV below 100 seats | $890,000 median ACV above 1,000 seats | Security, governance, deployment support, and enterprise-wide scale |
Source: Monetizely’s 2025 benchmark.
The practical point is straightforward: higher enterprise prices are not justified by a longer feature list. They are justified by a different job. A 1,000-seat buyer needs controls, auditability, integration depth, service levels, and a commercial structure that can survive procurement review.
Cursor offers a disciplined example of this logic in AI software. Its core promise remains consistent across solo developers, professional teams, and enterprises: write code faster. The packages change around the needs of each group. Individual users need the editor and AI assistance. Teams need shared billing and administration. Enterprises need security, compliance, audit trails, and support. The product’s core capability remains largely constant; the package changes because the buyer’s operating environment changes.
Devin illustrates the cost of leaving the middle segment poorly served. Its natural market includes evaluators, scaling engineering teams, and enterprise deployments. Yet a sharp jump between an entry plan and a large team package can force serious small teams to choose between an inadequate trial and an oversized commitment. A package gap is not a minor presentation problem. It gives competitors an opening at the moment a customer moves from curiosity to regular use.
Our view is that companies should build fewer packages around clearer segment differences. A three-tier page is not evidence of sound strategy. It is sound only when each tier solves a distinct buyer problem.
A primary meter makes multi-part prices understandable
Multi-part pricing earns its place when each element does a different job. A platform fee can pay for the always-on value of administration, security, integrations, and access. A seat can capture individual productivity. A usage or outcome charge can capture work completed at greater scale.
Confusion begins when two or three meters all try to charge for the same value. Buyers cannot forecast spend, sales teams cannot explain the model, and finance cannot tell whether expansion came from adoption or billing mechanics.
The leading examples separate those jobs cleanly.
Exhibit 3. Four pricing architectures show how a primary meter should anchor the model
| Company | Pricing evidence | Primary meter | Supporting commercial mechanism | What the design gets right |
|---|---|---|---|---|
| Snowflake | Consumption of compute, storage, and data transfer as of January 31, 2025 | Resource consumption | Annual capacity commitments | Spend rises with platform use, while commitments support planning |
| Datadog | Usage-tailored pricing reported for FY2025 | Product usage | Cross-sell across products and broader deployment | Expansion follows more monitored systems and workloads |
| Intercom Fin | $49 for 50 resolutions plus $0.99 per additional resolution as of March 19, 2025 | Resolved customer issue | Entry bundle for adoption | Buyers pay for a completed support result rather than model activity |
| Salesforce Agentforce | $2 per conversation and Flex Credits for actions as of May 15, 2025 | Conversation or defined action | Credit-based flexibility | The model creates a bridge from a simple visible unit to more varied agent work |
Sources: Snowflake, Datadog, Intercom, and Salesforce disclosures. -
Intercom’s 2025 pricing for Fin is especially instructive. A customer service leader understands a resolution. A token count or model call count would expose the vendor’s cost structure, not the buyer’s goal. Fin’s model therefore used a visible outcome: 50 resolutions for $49, then $0.99 for each additional resolution.
Salesforce took a more transitional path with Agentforce. Its May 15, 2025 announcement retained a $2-per-conversation option while adding Flex Credits for discrete agent actions such as updating records and automating workflows. That structure acknowledges a real market condition: buyers may first accept a familiar unit, then adopt a more precise action-based meter as the range of agent work expands.
A primary meter does not require a single line item. It requires a single answer to the buyer’s question: “What causes my spend to grow?” For Snowflake, the answer is consumption. For Fin, it is a successful support outcome. For a classic collaboration product, the answer may still be active users.
Autonomous agents justify outcome pricing only when they can complete work reliably
AI changes the pricing question because it can shift value from assistance to execution. An embedded writing assistant may help a person work faster. A customer-service agent may resolve a case without human involvement. A cross-functional agent may complete a chain of tasks across several systems.
The Agentic Monetization Spectrum, or AMS, provides a disciplined way to distinguish those cases. It rates an agent on three dimensions: zero-human ability, or how much human work remains; operational domain, or how broad the agent’s responsibility is; and output/cost ratio, or how far the value created exceeds the cost of producing it. As scores increase, the strongest pricing logic moves from access and seats toward outputs and outcomes.
A simple scoring view makes the implication concrete.
Exhibit 4. Higher AMS scores move the primary meter from access toward completed work
| Product archetype | Zero-human ability | Operational domain | Output/cost ratio | Total score | Recommended primary meter |
|---|---|---|---|---|---|
| Embedded AI copilot | 2 | 1 | 2 | 5 | Per seat |
| Functional AI agent, such as customer support automation | 3 | 2 | 2 | 7 | Per completed outcome |
| Broad enterprise agent operating across systems | 3 | 3 | 3 | 9 | Per verified action or business outcome |
Scoring scale: 1 = small or linear, 2 = medium or inflecting, 3 = large or exponential. AMS definitions from Monetizely.
The spectrum explains why per-seat pricing remains defensible for some AI products. Cursor sits closer to the copilot end of the range because a developer remains the quality gate and the work stays within a single function. Seat pricing tracks the buyer’s mental model: developer productivity.
Sierra and Intercom Fin sit further toward completed work. Their customer-service products can triage, resolve, hand off, and act across customer channels. Sierra’s enterprise model reflects that broader scope, while Fin’s resolution price offers a more visible outcome unit for buyers who want to automate routine support.
Harvey presents a useful exception that proves the rule. Its AI may create value well beyond the marginal cost of an individual query, but legal buyers still organize budgets around lawyers, practices, security, and firmwide procurement. Per-seat pricing can therefore remain commercially effective even when a pure output meter might capture more theoretical value. The buyer’s established purchasing habit matters as much as the technical potential of the product.
Rate setting works only after leaders decide what the company will reward
Pricing teams often spend too much time debating whether a plan should cost $49 or $59. That debate comes late. The more important choice is whether the company wants to reward adoption, scale, completed work, enterprise control, or some combination with one clear primary driver.
Rate setting then becomes a disciplined question: how much value does the chosen unit create, what does it cost to serve, what price can the market accept, and what business goal should win when those answers conflict? Monetizely’s position is that price should follow the architecture, not substitute for it.
Exhibit 5. The buyer’s value driver should dictate the primary meter
| Buyer sees the product as | Best primary meter | Supporting terms that may be added | Poor design choice |
|---|---|---|---|
| A productivity tool used by named people | Active user or seat | Higher-tier admin, security, or support | Charging separately for every minor action |
| A platform that processes growing data or traffic | Data, events, compute, or workload | Annual commitment and volume discounts | Selling unlimited use at a price that ignores scale |
| An agent that resolves a defined business task | Verified resolution, action, or completed case | Minimum commitment and service levels | Charging for attempts, tokens, or hidden model calls |
| An enterprise system of record | Active employee, managed identity, or business unit | Platform fee for governance and implementation | Treating every employee as a power user |
The matrix points to a single operating principle: customers should be able to predict what they will pay over three years by forecasting the same driver that creates value for them.
That standard also protects the vendor. When a model connects price to value, product teams can invest in the behavior that expands revenue. Snowflake can improve data workloads and consumption. Intercom can improve resolution quality. A seat-based collaboration vendor can improve adoption across more users and teams. Each organization knows where product progress and commercial progress meet.
The 2025 market did not reward complexity for its own sake. It rewarded companies that made value, packaging, and the meter line up. The shift toward usage and outcome pricing is real, but it is not a mandate to abandon subscriptions. It is a mandate to stop charging for a proxy when a clearer unit of value is available.
Monetizely's position is that every SaaS company should defend one primary meter. A platform fee may support that meter. Seats may support it. A usage allowance may support it. None should obscure it.
Choose the economic behavior the company wants to accelerate. Decide whether the next three years require broader adoption, larger deployments, greater usage, or more completed work, then make that behavior central to pricing.
Treat pricing as a product roadmap constraint. Do not ship major AI capabilities until product, finance, and go-to-market leaders agree on the unit customers will recognize and the data required to measure it.
Build board reporting around the primary meter. Track revenue, retention, gross margin, and expansion by the value driver that causes spend to grow, not only by plan name or sales segment.
Use enterprise pricing to capture control and scale, not to conceal the model. Buyers will accept custom terms for security, service, and deployment. They should still understand the fundamental driver of spend.
Revisit the meter when AI moves from assisting work to completing it. A seat can remain effective while people remain central to the workflow. Once the product completes verified work without them, the commercial model should evolve with it.
Sources
- Monetizing Agentic AI - https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
- Monetizely, “SaaS Pricing Benchmark Study 2025: Key Insights from 100+ Companies Analyzed,” December 22, 2025. (getmonetizely.com)
- Snowflake, FY2025 Fourth Quarter Investor Presentation, data as of January 31, 2025. (investors.snowflake.com)
- Datadog, Form 10-K, filed February 18, 2026, reporting customer and pricing information through December 31, 2025. (investors.datadoghq.com)
- Intercom, “Built For You Spring ’25,” March 19, 2025. (intercom.com)
- Salesforce, “Salesforce Introduces New Flexible Agentforce Pricing,” May 15, 2025. (salesforce.com)