.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.
What Is a Billing Model a Complete Guide to Understanding SaaS Billing Types
A billing model looks simple from the outside. A customer pays monthly, annually, or as they use a product. Yet the choice shapes far more than an invoice. It decides which customers can buy without fear, which product behaviors a company rewards, how much revenue is predictable, and whether heavy users create profit or losses.
The stakes have risen because B2B SaaS companies increasingly sell several forms of value at once. HubSpot combines platform subscriptions, seats, and credits for AI actions. Twilio charges by communications activity. Snowflake charges for consumption of compute and storage. Intercom charges for both human-agent seats and Fin AI Agent outcomes. A finance leader therefore needs more than a monthly price list. They need a clear answer to a harder question: what, exactly, should the customer pay for?
Monetizely's position is direct: every SaaS company should choose one primary meter that tracks repeatable customer value and that buyers can forecast. Base fees, included allowances, and overages should support that primary meter, not conceal a weak one.
A billing model turns a product promise into a repeatable commercial system
In everyday SaaS language, “billing model” often refers to everything from per-seat pricing to annual invoicing. That shortcut creates confusion. A complete billing model is the set of rules that converts a product offer into a charge, an invoice, and a payment over time.
Five decisions sit inside that system:
Monetizely's 5-Step Pricing Framework places those decisions in order: goals and segmentation; packaging; pricing metric; price points; and operationalization. The sequence matters because a company cannot sensibly set a price before it knows which buyer it serves, what that buyer receives, and which unit best represents the value delivered. The framework also recognizes a practical truth: billing operations come last, but they determine whether the first four decisions survive contact with finance systems and customer scrutiny. The logic is developed more fully in Monetizing Agentic AI.
Exhibit 1: The five decisions behind a working billing model
| Decision | Core question | Output | Common failure |
|---|---|---|---|
| Goals and segmentation | Which customers matter most, and what must pricing accomplish? | Defined buyer groups and commercial goals | One offer forced onto startups, mid-market firms, and enterprises |
| Packaging | What does each buyer receive? | Plans, bundles, add-ons, services, and contract terms | Feature tiers built around product modules rather than buyer needs |
| Pricing metric | What does the customer pay for? | Seat, host, message, credit, transaction, outcome, or account | Charging for an internal cost driver that buyers cannot understand |
| Price points | What is the rate and discount logic? | List prices, commitments, volume breaks, and overages | Setting a number before choosing the meter |
| Operationalization | Can the company measure, rate, invoice, and explain the charge? | Usage records, billing rules, invoices, and dispute processes | A metric that sounds elegant but cannot be audited |
The table shows why monthly versus annual billing is not the central choice. Invoice timing matters for cash flow and customer budgeting, but the meter determines what customers believe they are buying.
A billing type is best understood as the unit that anchors the commercial relationship. Each model distributes uncertainty differently between vendor and customer.
A flat subscription places most uncertainty with the vendor. The customer receives broad access for a known price, while the vendor accepts the risk that usage could rise. Consumption pricing shifts more uncertainty toward the customer, who pays when activity rises. Outcome pricing shifts the discussion further, because the buyer pays for a defined result rather than access or effort.
Exhibit 2: The major SaaS billing types and the jobs they perform
| Billing type | Primary meter | What the buyer can predict | Best fit | Current B2B SaaS example |
|---|---|---|---|---|
| Flat subscription | Account, workspace, or platform | A fixed recurring bill | Products valued for ongoing availability rather than frequent activity | HubSpot Professional starts at $1,300 per month on annual billing and includes six seats, as accessed September 8, 2026. (hubspot.com) |
| Per-seat | Named or active user | Spend based on team size | Collaborative software where people remain the main users and value anchors | HubSpot Starter begins at $7 per seat per month on annual billing, as accessed September 8, 2026. (hubspot.com) |
| Resource-based | Host, device, database, or other managed asset | Spend based on deployed environment | Infrastructure products tied to a visible technical footprint | Datadog Infrastructure Pro lists at $15 per infrastructure host per month on annual billing, as accessed September 8, 2026. (datadoghq.com) |
| Usage-based | Message, API call, GB, minute, or transaction | Spend based on operating activity | Products where each unit of activity has clear customer and delivery value | Twilio U.S. SMS starts at $0.0083 per outbound message, plus applicable carrier fees, as accessed September 8, 2026. (twilio.com) |
| Credit-based consumption | Standardized credit tied to compute or service use | Spend based on prepaid or postpaid consumption | Platforms with several technical services that need one common unit | Snowflake states that credits are consumed while customers use compute resources, as accessed September 8, 2026. (snowflake.com) |
| Outcome-based | Verified business result | Spend based on completed work | Autonomous products that complete a defined task with limited human labor | Intercom Fin charges $0.99 for a resolution, procedure handoff, or disqualification, and $9.99 for a qualified lead, under terms published July 30, 2026. (intercom.com) |
The lesson is not that one type is universally superior. The lesson is stricter: the product’s operating role should determine the primary meter.
A sales workspace used by 40 account executives has a human-centered value story. Seats make sense because the buyer already manages headcount, access, and accountability by person. A communications API creates value each time a message or call occurs. Charging by activity is more credible than charging for abstract access. A support agent that resolves a customer issue without a human should move toward the completed resolution, because the customer is buying completed work.
Snowflake offers a useful contrast. Its credit system does not claim to price every business outcome. It translates technical consumption into a common unit across a broad data platform. That choice works because compute activity is both measurable and material to Snowflake’s delivery cost.
Rate mechanics change cost behavior but do not create customer value
Teams often call tiered pricing, volume pricing, and overage pricing separate billing models. They are not. They are rate mechanics layered onto a primary meter.
A company can charge per seat and use volume discounts. It can charge per message with graduated rates. It can charge a platform subscription with included credits and overages. The meter remains the core commercial decision.
Exhibit 3: Rate mechanics determine how spend changes after usage rises
| Rate mechanic | How it works | Customer effect | Main risk |
|---|---|---|---|
| Flat rate | One recurring charge regardless of usage within the contract | Maximum budget certainty | Heavy usage can erode margin |
| Tiered rate | Price changes when the customer enters a new band | Simple to explain | A threshold can create a sudden bill increase |
| Volume rate | All units receive a lower rate after a threshold is reached | Rewards scale | Can reduce revenue faster than expected |
| Graduated rate | Each band receives its own rate | More proportional pricing | Harder for buyers to calculate manually |
| Included allowance plus overage | A base fee includes a set amount, then extra units are charged | Predictable starting cost with expansion path | Included units can obscure the real meter |
| Commitment plus true-up | Customer commits to a minimum level, then reconciles later | Budget control for buyer and revenue visibility for vendor | Renewal disputes if measurement is unclear |
The practical implication is clear: companies should debate the meter before debating the discount curve. A poor meter with sophisticated tiers remains a poor meter.
Monetizely recommends testing any proposed meter against five questions:
A meter that fails three of those tests should not reach the rate-card stage.
Packaging and billing are related, but they solve different problems. Packaging answers, “What does this customer receive?” Billing answers, “What causes the customer to pay more?”
Confusing them creates familiar problems. A company may place every advanced feature in an enterprise tier, then learn that enterprise buyers want only two of those features and demand a steep discount. Another company may sell low-cost entry plans with generous usage, then discover that its most active customers consume far more infrastructure than the subscription covers.
HubSpot’s current structure illustrates a coherent layered design. Its subscription plans establish platform access and seats, while HubSpot Credits pay for certain AI-powered actions. The company does not present credits as a replacement for the customer platform. Credits extend the model where variable agent activity creates a separate source of value and cost.
Exhibit 4: A coherent offer keeps one primary meter visible
| Product situation | Primary meter | Supporting commercial elements | Current example |
|---|---|---|---|
| Collaborative business software with AI assistance | Seat | Plan tier, included AI credits, optional add-ons | HubSpot combines seats with included and purchasable credits for agent actions. (hubspot.com) |
| Infrastructure monitoring | Managed host or device | Data retention, event ingestion, container overages | Datadog charges infrastructure monitoring per host and separately prices several usage-heavy services. (datadoghq.com) |
| Communications platform | Message, minute, or active user | Phone number fees, carrier pass-through fees, volume discounts | Twilio charges by message activity and applies separate number and carrier charges. (twilio.com) |
| Autonomous customer-service agent | Successful resolution or completed procedure | Minimum commitment, service terms, optional human-seat software | Intercom Fin charges for defined outcomes and can also sit alongside Intercom seat plans. (intercom.com) |
The table points to a disciplined architecture: use a secondary charge only when it covers a distinct source of value, cost, or service level. Do not add meters simply because a competitor has more lines on its pricing page.
AI changes billing because it can reduce, and sometimes remove, the human user as the value anchor. A writing assistant used by a marketer can still fit a seat model. An agent that resolves support issues, qualifies leads, or completes coding tasks begins to look less like software access and more like delegated labor.
The Agentic Monetization Spectrum, or AMS, helps make that distinction. It rates an agent across three dimensions: zero-human ability, or how little human work remains; operational domain, or whether the agent handles one task, a functional workflow, or work across functions; and output/cost ratio, or whether delivered value rises faster than the cost to produce it. Small, medium, and large positions on those dimensions indicate how far a company should move from per-seat pricing toward output or outcome pricing.
Intercom Fin provides a useful case. Its published rules define an outcome as a resolution, a completed procedure handoff, a qualification, or a disqualification. That clarity matters. The company can identify the event, rate it, show it to the customer, and avoid charging for failed attempts.
Exhibit 5: AMS assessment for an autonomous support-agent archetype
| AMS dimension | Assessment | Score | Billing implication |
|---|---|---|---|
| Zero-human ability | Medium - the agent can resolve defined issues, but humans still manage exceptions, escalations, and service design | 2 of 3 | Seats should not be the primary charge for autonomous resolutions |
| Operational domain | Medium - the agent works across an end-to-end customer-support workflow within one function | 2 of 3 | A resolution or completed workflow is more credible than a token count |
| Output/cost ratio | Inflecting - a completed customer issue can be worth far more than the compute used, but performance must remain provable | 2 of 3 | Outcome pricing can work, with clear definitions and reporting |
| Total | Agent performs meaningful work but remains bounded by a single operating function | 6 of 9 | Outcome should be the primary meter; commitments can provide budget control |
The score supports a specific conclusion. An autonomous support agent should charge primarily for successful work, not for a human seat that becomes less relevant as the agent resolves more conversations. A platform fee may still fund onboarding, integrations, governance, or service levels, but it should remain secondary to the completed outcome.
The fifth step in Monetizely's 5-Step Pricing Framework is operationalization because even an excellent metric fails when the customer cannot verify it. Usage data must move from product event to billing record without losing account identity, timestamp, price version, entitlement status, or evidence of completion.
Before launch, operators should establish five practical controls:
Intercom’s outcome rules show the standard to aim for. The company specifies that only one outcome is charged per conversation and that unsuccessful attempts do not create a charge. Clear boundaries reduce the chance that finance teams and customer-success teams must renegotiate every invoice manually.
A complete billing model is not a menu of ways to charge. It is a deliberate link between customer value, product behavior, unit economics, and an invoice that a buyer can defend internally.
Monetizely's position is that companies should resist the temptation to copy consumption pricing, outcome pricing, or elaborate credit systems because those models appear modern. A collaborative product should begin with the person using it. An infrastructure product should begin with the resource or activity consuming capacity. An autonomous agent should begin with the work it completes. Each choice produces a clearer contract, a more credible renewal story, and a stronger basis for margin management.
Assign one executive owner for the primary meter. Product, finance, sales, and customer success should contribute, but one leader must resolve conflicts between adoption, revenue, and margin goals.
Treat a billing-model change as a product migration, not a price update. Create separate plans for new customers, renewals, and existing contracts rather than forcing every account through the same conversion date.
Review forecast error at the customer level after each billing cycle. Compare expected spend with actual invoices, then investigate accounts with large gaps before those gaps become renewal objections.
Measure the model on four board-level outcomes: conversion, expansion, gross margin, and invoice disputes. ARR alone can make an unsustainable billing model look successful.
Retire secondary meters that do not change customer behavior or protect economics. Every extra line item adds sales friction, product complexity, and support cost.
Footnotes
- Monetizing Agentic AI. https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
- HubSpot, “Customer Platform Pricing,” accessed September 8, 2026. (hubspot.com)
- Twilio, “Messaging Pricing” and “SMS Pricing in United States,” accessed September 8, 2026. (twilio.com)
- Snowflake, “Pricing” and “Pricing Calculator: Feedback and FAQs,” accessed September 8, 2026. (snowflake.com)
- Datadog, “Pricing,” accessed September 8, 2026. (datadoghq.com)
- Intercom, “Pricing” and “Fin AI Agent Outcomes,” accessed September 8, 2026; outcome guidance published July 30, 2026. (intercom.com)