SaaS pricing models: every type, real examples and how to choose

A SaaS pricing model is the rule that turns what a customer uses into what they pay. Per-seat, tiered, usage-based, credit-based and hybrid are all SaaS pricing models. Each one answers the same question differently: when a customer gets more out of your product, what makes their bill go up?

Of everything in your pricing, the model matters most. Pick the wrong one and nothing else you do will fix it. Better packaging won’t, and neither will sharper price points or a stronger sales team. Pick the right one and you barely notice it’s there. Customers who succeed with your product pay more, and nobody feels punished.

Most guides to SaaS pricing models are written by billing companies. They list the same seven models, give each a pros and cons line, and tell you to start by checking what your competitors charge. That last piece of advice is the most common mistake I see. This guide covers the twelve models you’ll actually run into, names companies using each one today, and gives you a way to choose that starts from your own product rather than someone else’s.

What is a SaaS pricing model?

A pricing model does two jobs. It calculates what each customer pays, and it makes that amount differ between customers. A flat fee does the first job perfectly and the second not at all: a five-person agency and a 5,000-person enterprise pay the same. Most of the complexity in real SaaS pricing exists to do the second job better.

The model is one layer of a larger pricing architecture. I split that architecture into six layers, from the bottom up: capability, fencing, packaging, pricing model, monetization mechanics (discounts, trials, billing terms) and price points. Price points are the tip of the iceberg and the part everyone argues about. Everything underneath them is your pricing structure, and the pricing model is the most important piece of it.

Inside the model sits the value metric, the unit the bill scales with. Seats, API calls, resolved tickets, dollars processed. The metric is the most important part of the model, because once you have a good one, the model mostly picks itself.

One more distinction clears up a lot of confusion. Models differ by when the customer commits to a quantity:

• License: bought up front. Seats, hosts, locations. Predictable for both sides, and customers limit their own use by deciding how much to buy.
• Consumption: measured after the fact. API calls, messages, active users. The customer only pays for what they use, and neither side knows the bill until the period closes.
• Credits: consumption, estimated and paid for up front. A hybrid of the two.

“Subscription” is sometimes called a pricing model. It’s not. It’s really the billing rhythm the other models run on. A per-seat plan billed monthly is a subscription, and so is a monthly credit allowance.

SaaS pricing models compared

Freemium isn’t in the table on purpose. It’s an acquisition choice that sits in front of a pricing model, covered under “Where freemium fits” below.

12 SaaS pricing models, with real examples

Prices and plan details below are as of September 2026. Pricing pages change often, so treat any number as a snapshot.

Flat-rate pricing

One price, everything included, however many users or however much usage. The customer never has to guess the next invoice, and the sales process is a link to the checkout page.

The cost is the second job of a pricing model. A flat rate can’t tell customers apart, so small customers overpay and large ones get a bargain. Basecamp is the example every guide uses, and even Basecamp now sells a per-user plan next to its flat Pro Unlimited plan.

• Fits: a narrow product sold to customers of similar size, or an MVP you want in front of buyers fast.
• Breaks: once your largest customer gets ten times the value of your smallest, and pays the same.

Tiered pricing

Several fixed packages at rising prices, each unlocking more features, more capacity or better support. PagerDuty runs a free plan and a ladder of paid tiers, each adding incident response features on top of the last. Tiers segment your market: small teams take the entry plan, larger ones pay for the tier with the features they can’t live without.

Tiers alone rarely carry the whole model. Most tiered SaaS products also charge per seat or per unit inside each tier, which makes them hybrids in practice.

• Fits: self-serve products with distinct customer groups who value different features.
• Breaks: when the tier boundaries don’t match how customers grow, so they sit in a tier that’s too small and resent the jump to the next one.

Per-seat pricing

A price per named user. Figma sells editor seats. Harvey sells legal seats at roughly $1,200 a year. At Toggl we priced by seats because that’s what made sense: each user got their own value from the product.

That is the test for seats. If every person using your product gets value from it independently, a seat tracks value and customers expect to pay that way. If one person gets the value and five others just look at the output, seats punish the customer for sharing it, and they’ll share a login instead.

HubSpot’s 2024 move to paid core seats and free view-only seats shows the fix. Charge for the people who work in the tool and let everyone else look for free.

• Fits: collaboration and productivity tools, and anything where users do their own work in the product.
• Breaks: when the software replaces the work rather than helping someone do it. Headcount falls exactly when your value rises.

Per-active-user pricing

A price per user who was actually active in the billing period. Slack’s Fair Billing Policy tracks activity and credits back seats that went unused, so a company can give everyone access without paying for people who never open the app.

It solves the biggest problem with seats in company-wide rollouts: a login nobody uses is worth nothing to the buyer. The price is a harder forecast for both sides, and you need a definition of “active” that holds up at the invoice. Does a service account count? Someone who only received a notification?

• Fits: tools rolled out to a whole company where usage is uneven.
• Breaks: when your telemetry can’t separate a human doing work from a system event. It’s also sometimes considered too pro-customer: any quiet month means a smaller invoice for you.

Per-unit pricing

A price per thing the customer has under management. Datadog charges per host, CrowdStrike per endpoint, Gusto per employee on payroll. The unit is something the customer runs rather than someone who logs in.

Per-unit is easy to measure and easy for finance to forecast. The trade-off is that it tracks the size of the customer’s organization, not how well your product is doing for them.

• Fits: infrastructure, security, HR and payroll, where the count of managed things is a fair proxy for the work you do.
• Breaks: when a customer’s unit count stays flat while their value from you grows.

Usage-based pricing

A price per unit consumed, billed in arrears. Twilio charges per message segment. AWS charges for the compute and storage you actually used. The barrier to entry is close to zero, and revenue grows automatically as the customer grows.

OpenView’s benchmarks put usage-based companies at 125% median net revenue retention against 115% for subscription peers. That gap is real in infrastructure and developer platforms and narrows in broader B2B samples, because automatic expansion also runs automatically in reverse when a customer’s business slows.

• Fits: APIs and infrastructure, where value scales evenly with volume and buyers expect utility-style billing.
• Breaks: when customers can’t forecast their own usage. Procurement pushes back, and the first bill shock costs you trust. Ship caps and spend alerts before you ship the meter.

Committed usage and drawdown

The customer commits to a spend up front, usually annual and usually at a discount, and draws usage down against it. Snowflake sells capacity this way. Supabase’s Pro plan works on the same principle at a smaller scale: the base fee includes a monthly compute credit, and compute beyond it bills by the hour.

A commitment turns usage into a number finance can budget. You get cash up front and a forecastable floor, and the customer still pays for what they actually use.

• Fits: infrastructure sold to larger accounts that want usage pricing with a budget line.
• Breaks: at renewal, if a large share of the commitment went unused and the customer feels they overpaid.

Credit-based pricing

The customer buys credits in advance, and each action burns a set number of them. Credits are usage-based pricing paid up front. They sit between a license, where the vendor gets the cash and the customer carries the risk, and pure usage, where the customer carries little risk and the vendor waits for its money.

There are two kinds:

• Single-use credits: one credit buys one thing. Audible’s monthly credit buys one audiobook.
• Multi-use credits: a currency spent across different actions at different rates. Clay, Replit and Lovable work this way. Clay overhauled its model in March 2026 to bill platform actions separately from data credits passed through at cost, cutting data pricing by 50% to 90%.

Credits do things other models can’t. Customers spend more freely because the money already feels spent. A leftover balance keeps them from churning, where a license buyer who overpaid just downgrades. You can change how many credits an action costs without renegotiating a contract. And a customer burning through credits halfway through the term is an upgrade signal you didn’t have to go looking for.

A few rules I’d hold to. Credits should expire, and a 24-month life is a sane default. Never refund credits as cash, or you’ve become a bank. Keep the math simple, give the customer a live usage dashboard and a forecast, and let admins set spend controls. At renewal, let actual spend in the last period set the commitment for the next one. I call that a transposed commitment, and it removes the renewal argument.

• Fits: AI products and tools where usage is bursty and different actions cost very different amounts to serve.
• Breaks: when nobody can work out what a credit buys. Procurement starts treating credits as a markup.

Take rate pricing

A percentage of the money flowing through the product. Stripe charges a percentage plus a fixed fee per card payment. Shopify pairs a plan fee with fees on the transactions its merchants process.

A dollar processed is the most consistent unit in software. Every dollar is worth exactly what the next one is, customers never try to process less, and everyone already expects to pay this way for payments.

• Fits: payments, commerce, marketplaces and anything that sits directly in the flow of the customer’s revenue.
• Breaks: everywhere else. If you don’t touch the money, you can’t take a cut of it.

Output-based and outcome-based pricing

These two get mixed up constantly, so it helps to place them on a chain. Every product sits on a chain that runs from inputs, through activities and outputs, to outcomes and finally impact. Usage-based pricing bills on inputs. Agent and task-based pricing bills on activities. Output-based pricing bills on completed work. Outcome-based pricing bills on a verified change in the customer’s business.

Nearly everything sold as outcome-based pricing is output-based. Intercom’s Fin charges $0.99 per resolved support conversation. Zendesk charges per verified resolution. Salesforce meters Agentforce in Flex Credits at $500 per 100,000 credits, where a digital action costs 20 credits. Those are all outputs: something the software finished, not money landing on the customer’s P&L.

True outcome pricing exists, but mostly where a third party decides the result and real cash moves. Chargeflow takes 25% of the chargebacks it recovers, and Visa and Mastercard adjudicate the win. Riskified charges a percentage of approved orders and reimburses the merchant for any that turn out to be fraud. I cover where the line sits, and how close you can get to it, in outcome-based pricing and why it’s mostly output-based.

• Fits: AI agents with a programmatic definition of “done”, like a resolved ticket.
• Breaks: when “done” needs a judgment call. Every invoice becomes a dispute.

Hybrid pricing

A base fee or seats, plus a usage, credit or output layer on top. Almost every mature SaaS company runs one, whatever its pricing page calls it.

The reason is a trade-off with three corners, and you only get two of them:

• Value alignment: the customer pays more as they succeed.
• Predictability: procurement can budget the line a year out.
• Margin protection: your infrastructure and inference costs are always covered.

A hybrid lets you hold two and part of the third. Intercom sells seats at $29 to $139 a month plus $0.99 per resolution, with a 50-outcome monthly minimum. Cursor sells a $20 seat with a pool of requests, then bills overage. Gusto charges a base fee plus a price per employee.

Adding AI features to an established SaaS product is the most common reason companies end up here right now. A seat model that worked for years stops working once some users run up inference costs far higher than others, and the model has to change. A lot of my work at Potio right now is exactly this.

• Fits: most post-PMF SaaS and AI products.
• Breaks: when the layers stack up faster than a buyer can follow them.

Where freemium fits

Freemium gives away a limited version for free and charges for the rest. Zoom caps free meetings at 40 minutes. Dropbox gives free users a small amount of storage.

It isn’t a pricing model. Free users don’t pay, and when they convert, they move onto one of the models above, usually tiered seats. Freemium is a decision about acquisition. It works when free users cost you almost nothing to serve and bring in paying ones, and it fails when free users use up support and infrastructure without ever converting.

How to create a SaaS pricing model in five steps

Start from the value metric, always. If you find a good value metric, or more than one, it leads you to the model. I’ve written about the method in depth in how to find your SaaS value metric, so here is the short version.

1. Map your customer’s value chain. Ask what the customer is trying to achieve. Then ask what has to happen before that, and before that, until you reach sign-up. For Calendly the chain is: set up an account and connect calendars, share booking links, get meetings booked, and have the meetings happen.
2. Find the value metric. Put candidate metrics under each step of the chain, then screen them on two questions: can you measure and attribute it, and does the customer happily pay more as it grows? That screen removes about 80% of candidates. Put the survivors through three filters: operational fit, customer perception and economic logic.
3. Let the metric’s shape pick the model. A metric counting people points to seats. Volume with even value per unit points to metered usage. Volume where some units are worth far more than others points to credits. Money flowing through the product points to a take rate. A completed job points to output pricing. Things under management point to per-unit pricing.
4. Decide what sits around the metric. A base fee sets a minimum and makes the per-unit price fall as customers grow. Add-ons carry features only some customers want. Cost-covering fees, like storage or premium support, charge for things customers already expect to pay for. Together with packaging and fences, this is your pricing structure.
5. Set price points last, then test the model. When a customer succeeds, does their bill go up? When they struggle, does it stay flat or go down? Would anyone try to use less of the thing you charge for? If a customer would work to shrink your metric, you’re charging for a cost, not for value.

Which SaaS pricing model fits your product?

B2B SaaS pricing models: what changes as the buyer gets bigger

Simple pricing is right when you sell to one type and size of customer. The bigger the deal, the more the model has to carry.

Self-serve products should stay close to one metric and a few tiers. A buyer on a credit card won’t read a rate card. Enterprise deals are different. Large customers negotiate hard against the core metric, so the model needs other parts that hold their price: a base or platform fee, charges for costs the buyer already expects to cover (support, storage, higher SLAs), and volume breaks shown up front so the buyer can see their size is already priced in. A price sheet that shows the discount they’re getting leaves much less room to push for more.

Volume discounts need watching. In one repricing I ran for a $2M ARR automotive tech SaaS, small shops paid roughly two dollars an order and the largest accounts around fifty cents. The company had never lost a large account on price. We moved them to one plan priced only on orders, with users and locations unlimited, and ARR grew 80% in the twelve months after launch, attributed to the pricing change, with no change in the cancellation rate.

Why you shouldn’t choose a pricing model because it’s trending

The biggest mistake I see is companies fitting themselves to a model. Someone recommended it, or a product they admire uses it, and they try to reshape their pricing around it. Never do that. Pricing models aren’t a trend to follow.

Right now the trend is credits and usage. Most founders who come to me already want one or the other. And for some of them, even today, seats are the better answer, because each user gets their own value from the product and a seat tracks that value better than any meter would.

The only thing a trending model does is raise your customers’ expectation to pay. People accept a pricing model more easily when it’s how they already pay for similar things. That’s a real benefit if the model also fits your value metric. It’s worthless if it doesn’t.

The adoption numbers are also less settled than they look, because most of them come from billing vendors. Metronome, which sells usage-based billing, reported in 2025 that 77% of the largest software companies have some form of usage-based pricing. Chargebee’s 2025 survey of 473 finance, product and GTM leaders found subscription still features in 75% of pricing strategies. Read together, usage is spreading mostly as a layer on top of subscriptions, and pure usage is still mostly an infrastructure model.

How AI is changing SaaS pricing models

AI breaks the economics seat pricing was built on. A traditional SaaS user costs roughly the same to serve as the next one. An AI user can cost many times more than the next one, depending on how hard they push the product.

Replit is the clearest public case. Its AI coding agents took revenue from $300M to $525M ARR while gross margin went from 36% to negative 14% across 2025. It moved from flat per-run fees to effort-based agent pricing to stop the bleeding, and took heavy criticism from users who couldn’t predict what a task would cost. Cursor introduced usage limits after a customer ran up a $7,225 bill that had to be refunded.

That’s why so many AI products have moved to credits or to seats plus credits. How you sell the extra capacity matters as much as the model itself:

• Bursty usage (solo users, prototypes, the occasional big job) suits top-ups. You’re selling flexibility.
• Sustained usage (teams using the product every day) should push customers to the next tier. You’re selling predictability. Selling top-ups to steady users swaps good recurring revenue for one-off cash.

Credits spreading through AI products does change what your customers expect. It doesn’t mean credits fit your product. Run the same value metric test as everyone else.

Pricing models vs SaaS pricing strategies

A pricing model sets how the bill is calculated. A SaaS pricing strategy is how you decide what level to charge and how to position it. The two get blended together on most lists, which is why “the 7 types of pricing strategies” rarely agree with each other.

The strategies that come up most:

• Value-based: price from what the product is worth to the customer.
• Cost-plus: price from your cost to serve, plus a margin.
• Competitor-based: price relative to the alternatives.
• Penetration: price low to win share, then raise.
• Skimming: price high while you’re one of few options, then lower.

A value-based strategy can run on a per-seat model, and a cost-plus strategy can run on usage. The strategy decides whether a seat costs $10 or $50. The model decides that you charge per seat at all. For SaaS the strategy should lean on value, with cost setting a floor, especially for AI products where cost to serve moves around.

How to change your pricing model without a revolt

When a pricing model change fails, the rollout is usually to blame more than the model.

Unity is the cautionary tale. In September 2023 it announced a runtime fee charged per game install, applied to games already in the market. Developers revolted, and in September 2024 Unity canceled the fee and went back to seat pricing with a price increase. HubSpot’s 2024 move to paid core seats went the other way: existing customers could stay on their legacy pricing, and view-only seats became free, so nobody paid for colleagues who only needed to look.

The rules I follow when moving a customer base onto a new model:

1. New customers first. Launch to new signups while existing customers stay on legacy pricing. It shows you where the model is wrong before it touches revenue you already have.
2. Fix what the pilot shows. In the automotive tech repricing, the pilot exposed a dead zone between two plans where customers got punished on overage, so we added a tier.
3. Cap the biggest jumps. Some legacy accounts faced a 3x increase and we capped it at 2x. A customer who leaves at 3x can’t be moved up later.
4. Migrate in cohorts. Roughly 10% of the base first, then everyone else once the first wave shows no rise in cancellations.
5. Never apply a change retroactively. Charging for something a customer already did is the fastest way to lose their trust.

Choosing your SaaS pricing model

There are twelve SaaS pricing models worth knowing, and most real companies run a hybrid of two or three. The list is the easy part. The choice comes from your value metric: find the unit that grows as your customer succeeds, and the model mostly follows.

Test it with one question. When your best customers get more value, do they pay more without anyone feeling punished? If yes, the model is doing its job, and you’ll hardly notice it. If you’re arguing about the model every quarter, you’re usually arguing about a value metric nobody has named.

Frequently asked questions

What are the four main SaaS pricing models?

There’s no official list of four. The four most often grouped together are flat-rate, tiered, per-seat and usage-based pricing. In practice most SaaS companies combine two of them, typically tiers with seats inside each tier, or a seat or base fee with a usage layer on top.

What are the 7 types of pricing strategies?

The lists that circulate usually mix pricing strategies with pricing models. The strategies proper are value-based, cost-plus, competitor-based, penetration and skimming pricing. Items like freemium, tiered or usage-based pricing that often round a list out to seven are acquisition choices or pricing models, not strategies.

What is the most common SaaS pricing model?

Tiered pricing with a per-seat charge inside each tier is still the most common setup in SaaS, and it’s what I see most often in companies coming to me. Hybrid models that add usage or credits on top of seats are growing fastest, driven mostly by AI features with variable costs to serve.

Is per-seat pricing dead?

No, but it’s no longer the default. Seats still work where each user gets their own value from the product, like design, productivity and internal tools. They break where the software replaces work instead of helping someone do it, because the customer’s headcount falls as your value grows. That’s why most AI products now pair seats with a usage, credit or output layer.

What is the difference between usage-based and credit-based pricing?

Usage-based pricing measures consumption and bills for it afterwards. Credit-based pricing sells consumption in advance: the customer buys credits up front and each action burns some. Credits give the vendor cash earlier and the customer a budget they control, and they let the vendor change what an action costs without rewriting contracts.

How often should you change your SaaS pricing model?

Rarely, and only for a structural reason: the product’s scope changed, your cost to serve changed, or customers are working to use less of what you charge for. A model change touches contracts, billing, sales compensation and every forecast in the company. Price points can move every year. A pricing model should hold for years.