The four common models
Most SaaS companies use one of four pricing models: freemium, usage-based, seat-based, or tiered. The right model is not a matter of taste — it is determined by the *value metric*: the unit of value your product delivers that scales with what the customer cares about.
Freemium
Freemium offers a permanently free tier alongside paid plans. It works when (1) the marginal cost of a free user is near zero, (2) free users help paid users (network effects, viral loops), and (3) there is a clear, painful upgrade trigger. It fails when free users cost real money to serve and never convert. Notion and Figma made it work; thousands of forgotten SaaS tools did not.
Usage-based
Usage-based pricing (also called consumption pricing) charges per unit of work — per API call, per processed message, per GB stored. It aligns price with the value the customer extracts and lowers the barrier to start, but introduces revenue volatility and forecasting risk. Default to usage-based if your costs are themselves usage-based (AI APIs, infrastructure resale).
Seat-based
Seat-based pricing charges per user. It is simple to communicate, easy to forecast, and works whenever the product becomes more valuable as more people in the company use it. The classic risk: customers stuff one shared login on five seats. Build account-level limits or shared-resource quotas to make sharing painful enough to discourage it.
Tiered
Tiered pricing bundles features into named plans (Starter, Pro, Business). It is the easiest model to set up but the hardest to get right — bad tiering either gives away too much in the lowest plan or pushes customers into "I don't fit any of these" purgatory. Aim for three tiers: free or starter, the plan most customers should land on (highlight it), and a higher tier with workflow- or volume-based gates.
Picking the model
Run our [SaaS Pricing Architect](/free-tools/pricing-architect) for an AI-generated recommendation tailored to your product, and the [LTV Calculator](/free-tools/ltv-calculator) to model the lifetime value implied by each candidate model. For the upstream question of *which audience* to price for, see [How to validate a startup idea in 7 days](/blog/validate-startup-idea-in-7-days).