.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.
A CFOs Guide to SaaS Pricing Balancing Revenue Growth and Margins 8E51F
Pricing has become a finance decision again. For much of the subscription era, a company could add users, hold a seat price, and rely on scale to improve gross margin. That logic weakens when more customer activity also means more cloud spend, model inference, support load, or data processing. A fast-growing account can now be the least profitable one in the portfolio.
The CFO’s question is not whether to pursue growth or margins. It is how to design a price architecture that lets revenue rise with customer value while making the cost tail visible before it damages gross profit. Buyers also need an answer they can budget for: what they will actually pay over three years, what causes the bill to rise, and what controls they retain.
Monetizely’s position is clear: SaaS companies should price on the unit that represents the customer’s real source of value, not on the unit that is easiest to invoice. Seats remain the right primary meter when a human user is the product’s anchor; usage belongs where the product scales with workloads; measurable outcomes belong where an autonomous agent completes valuable work. In every case, finance must set a committed minimum and a hard cost guardrail around that primary meter.
A visible unit of value resolves the growth-versus-margin conflict
A pricing model fails when sales sees a simple offer, product sees adoption, and finance sees an uncapped obligation. The failure often starts with a fixed subscription that offers unlimited use of an expensive service. Revenue stays flat while cloud or inference costs climb.
Public SaaS companies show why the economics need close attention, even though no single company’s margin is a direct benchmark for another. Datadog reported $3.43 billion in revenue and $687 million in cost of revenue for fiscal year 2025, equal to a GAAP gross margin of about 80.0%. Snowflake reported $4.68 billion in revenue and $1.54 billion in cost of revenue for the fiscal year ended January 31, 2026, equal to about 67.2%. Their product mix, contracts, and infrastructure needs differ, but both figures show that cost of revenue is a strategic line item rather than an operational afterthought.
The practical implication is straightforward. Finance should not ask, “Should we move to usage pricing?” The better question is, “What customer event creates value, what does it cost us to deliver, and can the buyer understand and control it?”
The five pricing decisions must occur in sequence, not in a discount meeting
Monetizely’s 5-Step Pricing Framework gives CFOs an order of operations that prevents rate cards from becoming a substitute for strategy. The five steps are Goals and Segmentation, Packaging, Pricing Metric, Price Points, and Operationalizing Pricing. Goals and Segmentation establishes whether the company needs penetration, higher expansion, margin protection, or a more efficient sales motion, and identifies the buyer groups that matter. Packaging then creates offers suited to those groups. Only after those choices should the team select the pricing metric, set price points, and build the billing, entitlement, reporting, and sales processes that make the design real. As Monetizing Agentic AI argues, the order matters because a price point cannot repair a package built for the wrong customer or a meter that tracks cost but not value.
CFOs should own the first and fifth steps with the CEO, CRO, and product leader. The first step sets the economic objective. The fifth determines whether that objective survives contact with usage spikes, sales exceptions, renewal negotiations, and invoices.
Three questions should be answered before a new price is approved:
- Which customer segment should expand fastest over the next 24 months?
- Which product behavior creates value that the customer recognizes and can measure?
- Which customer behavior creates a cost exposure that finance cannot leave unlimited?
A company selling project management software to 20-person teams and global enterprises should not assume that the same package can serve both. The small team may value fast setup and a predictable per-user bill. The enterprise may pay more for identity controls, audit logs, support, and administrative reporting. Those are different needs, not merely different discount bands.
Packaging is the bridge between segmentation and price. A package should make a buyer say, “This was designed for the way we operate,” rather than, “We are buying a list of features we may never use.”
Cursor offers a useful current example. As of September 8, 2026, its Teams plans combine paid seats with included usage and on-demand consumption. Its Enterprise offer adds pooled usage, invoicing, SCIM, audit controls, and advanced administration. The design recognizes that a solo developer, a collaborative engineering team, and a governed enterprise do not buy the same thing, even when all want faster coding.
The same principle appears in conventional SaaS. Jira’s public pricing page, accessed September 8, 2026, lists Standard at $7.91 per user per month and Premium at $14.54 per user per month. Paid tiers include different amounts of pooled Rovo AI credits, while higher tiers add planning, support, administrative, and security capabilities. Atlassian is not simply charging more for more software. It is monetizing the management needs that grow with organizational complexity.
Exhibit 1: Four public pricing designs show how the meter follows the work performed, September 8, 2026
| Company | Public commercial design | Primary value signal | Cost-control mechanism | Source |
|---|---|---|---|---|
| Atlassian Jira | Per-user tiers, with pooled AI credits included by plan | Coordinated work by named users and teams | Tier limits and pooled monthly credits | |
| Datadog | Per-host pricing plus separate usage units for containers, metrics, logs, and spans | Monitored infrastructure and data volume | Commitments, hourly overage logic, product-specific units | |
| Snowflake | Consumption-based model for compute and storage | Data workloads processed and stored | Credits, usage monitoring, and prepaid capacity options | |
| Salesforce Agentforce | User access, Flex Credits, conversations, and resolution-based options | Agent actions and customer interactions | Credit pools, action multipliers, and digital-wallet visibility |
The common lesson is not that every SaaS company should copy consumption pricing. Each company identifies the work that creates customer value, then gives finance a way to observe the activity that drives delivery cost.
The primary meter is the unit that drives most expansion revenue. It should be the center of the commercial model, not an afterthought attached as an overage clause. A secondary charge can support it, but the buyer must always know which unit matters most.
A CFO can assess the candidates through four tests: whether the meter tracks customer value, whether it makes delivery cost visible, whether it is easy to budget, and whether the company can invoice and defend it without manual exceptions.
Exhibit 2: The primary meter should match the customer’s source of value
| Product condition | Recommended primary meter | Value alignment | Budget clarity | Cost visibility | CFO judgment |
|---|---|---|---|---|---|
| Human employee uses software to do better work | Active seat | High | High | Medium | Keep the seat as the anchor |
| Customer adds systems, workloads, or data | Host, transaction, stored volume, or processed volume | High | Medium | High | Use a commitment with measured overage |
| AI assistant supports a human reviewer | Seat with included usage | High | High | Medium | Include enough usage for normal work, then meter heavy use |
| Agent completes a defined business task | Completed action or verified outcome | Very high | Medium | High | Charge for the completed unit, with a minimum commitment |
The table points to a disciplined choice: use the meter that makes the buyer’s gain legible, then add contractual limits that prevent unpriced cost from accumulating.
Datadog’s model demonstrates the logic in infrastructure software. Its public list price for Infrastructure Pro is $15 per host per month when billed annually, while related services use other units such as containers, custom metrics, indexed spans, and ingested data. Datadog’s billing documentation also distinguishes between stable environments and more variable fleets through different commitment and hourly billing structures.
Snowflake reaches the same destination through a different product. Its pricing materials state that customers pay for consumption, with total cost based on credits used and the applicable price per credit, while storage is calculated on average monthly use after compression. A data platform earns more as customers run more valuable workloads, and the buyer can manage that spend through consumption monitoring and capacity commitments.
Unlimited expensive use turns growth into a gross-margin leak
A fixed annual subscription is not inherently dangerous. It becomes dangerous when the product’s marginal cost rises sharply with a customer behavior that the contract does not price, cap, or govern.
Consider a SaaS offer with $1 million in annual revenue and $360,000 in cost of revenue. Gross margin is 64%, producing $640,000 in gross profit. If customer activity doubles and raises cost of revenue to $540,000 while the fixed contract remains at $1 million, gross margin falls to 46%, and gross profit declines to $460,000.
Exhibit 3: A value-linked usage layer preserves gross profit as costly activity rises
| Scenario | Annual revenue | Cost of revenue | Gross profit | Gross margin |
|---|---|---|---|---|
| Baseline committed subscription | $1,000,000 | $360,000 | $640,000 | 64.0% |
| Usage doubles under an unlimited contract | $1,000,000 | $540,000 | $460,000 | 46.0% |
| Usage doubles and revenue rises to $1.3 million | $1,300,000 | $540,000 | $760,000 | 58.5% |
The difference is not a billing trick. A well-chosen variable layer gives the customer room to expand while ensuring that a higher-cost deployment generates more gross profit, not less.
The guardrail should not be tokens if customers cannot connect tokens to any business result. Customers buy processed claims, analyzed contracts, monitored hosts, resolved support requests, and productive employee time. Tokens may be useful inside the cost model, but they rarely make a good front-door price unless the buyer is already an infrastructure or developer audience that understands them.
Agentic AI raises the stakes because the product may perform work that previously required an employee, while its costs vary with model choice, reasoning depth, tool calls, and workflow length. A seat can still be useful for access, but it loses force when the human is no longer the central unit of production.
The Agentic Monetization Spectrum, or AMS, helps locate the correct meter for these offers. It scores an agent on three dimensions: zero-human ability, meaning how little human work remains; operational domain, meaning whether the agent handles a task, a function, or work across functions; and the output/cost ratio, meaning how quickly delivered value rises relative to compute cost. As autonomy, domain breadth, and value relative to cost increase, the pricing center should move away from a human seat and toward output or outcomes.
Exhibit 4: AMS places agents on a path from seat pricing to outcome pricing
| Product | Zero-human ability | Operational domain | Output/cost ratio | Recommended primary meter |
|---|---|---|---|---|
| Cursor | Medium | Medium | Inflecting | Seat with included and on-demand usage |
| Devin | Large | Medium | Inflecting to exponential | Usage tied to completed agent work |
| Harvey AI | Medium | Large | Exponential | Seat remains workable, with a heavy-use layer |
| 11x Alice | Large | Medium | Inflecting | Base commitment plus qualified-output charge |
| Sierra AI | Large | Large | Exponential | Verified customer-service outcome |
The score explains why a single answer for AI pricing is inadequate, while preserving a firm rule: pricing must follow the work the buyer believes it is buying. Sierra’s per-resolution approach fits a product that resolves customer interactions. Cursor’s seat-led model fits an assistant whose work remains anchored to an individual developer.
Salesforce’s Agentforce price architecture shows how an established SaaS vendor can support several levels of autonomy without treating them as one product. As of September 8, 2026, Salesforce lists Flex Credits at $500 per 100,000 credits, states that standard Agentforce actions consume 20 credits, and offers $2 per conversation and $2 per help-agent resolution options. The company also lists an Agentforce user license at $5 per user per month, which preserves a familiar access charge while usage and outcome units capture agent activity.
Pricing only balances growth and margins if the company runs it as a control system. A static annual price review is too slow for products whose use and costs can change weekly.
Finance should review a compact set of measures every month:
Exhibit 5: Five measures reveal whether the model is funding profitable expansion
| Measure | What it reveals | Action when it moves the wrong way |
|---|---|---|
| Gross margin by package | Whether a tier is priced above its delivery cost | Change included usage, add a surcharge, or raise the floor |
| Gross margin by customer cohort | Whether new customers are less profitable than the installed base | Tighten qualification or redesign entry packages |
| Expansion revenue by meter | Whether the primary meter grows with realized value | Remove friction from the meter or revisit the value unit |
| Overage disputes and credits | Whether customers understand and accept the bill | Improve usage visibility, alerts, and contract language |
| Discounted recurring revenue | Whether sales is bypassing the architecture | Set approval limits and measure discount impact by segment |
The operating discipline matters as much as the rate. Monetizely’s research on pricing operations notes that billing, metering, entitlement logic, usage reporting, and customer-readable invoices must work together; otherwise, even a strong metric becomes a source of revenue leakage and customer conflict.
The board should not receive a pricing update limited to average selling price and discount rate. Those measures describe the result of commercial activity, not whether the company has priced the economics of its product correctly.
Monetizely’s position is that the CFO should insist on one named primary meter for each major offer. That meter should be visible in product telemetry, the customer contract, the invoice, the renewal plan, and the board dashboard. A company can combine a minimum annual commitment with a value-linked growth unit, but the primary meter must remain clear.
The payoff is greater than margin protection. A coherent meter makes sales qualification sharper, product investment more focused, customer spend more predictable, and expansion more defensible. Growth then becomes a financial asset rather than a volume target purchased with hidden delivery cost.
Set a formal margin floor by package and customer cohort. Require executive review when any new deal, renewal, or product change is expected to push a cohort below that threshold.
Choose one primary meter for every major offer. Put its definition in product telemetry, quoting tools, invoices, and board reporting so every function works from the same unit.
Fund a pricing operations roadmap alongside the product roadmap. Treat metering, entitlement controls, customer usage alerts, and invoice clarity as revenue infrastructure.
Move pricing authority upstream. Have finance, product, and sales leadership approve package architecture before field teams set discount guidance or launch a new AI feature.
Report expansion quality, not expansion alone. Track whether incremental revenue from usage, seats, or outcomes produces incremental gross profit after cloud, model, and support costs.
Footnotes
- Monetizing Agentic AI: https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
- Datadog official pricing and billing documentation, accessed September 8, 2026; Datadog 2025 Form 10-K, filed February 2026. (docs.datadoghq.com)
- Snowflake official pricing materials, accessed September 8, 2026; Snowflake fiscal 2026 Form 10-K, filed March 20, 2026. (snowflake.com)
- Atlassian Jira Cloud pricing page, accessed September 8, 2026. (atlassian.com)
- Salesforce Agentforce official pricing and Flex Credits rate card, accessed September 8, 2026; Cursor official pricing and team-pricing documentation, accessed September 8, 2026. (salesforce.com)