Usage-based pricing vs subscription is the wrong question to ask before you know who is actually paying you. The real answer: usage-based pricing works when your customer is another piece of software, and subscription pricing works when your customer is a human being who does not want to watch a meter. That is not a slogan. It is the rule of thumb Andreessen Horowitz's growth team landed on after pricing hundreds of portfolio companies, and it holds up better than any pricing template you will find. If you are a seed or Series A founder staring at a pricing page with three columns and a "contact us" button, this decision matters more than your logo, your landing page copy, or your next feature. Get it wrong and you will spend a year fighting your own pricing model instead of your competitors.
What usage-based pricing actually means
Usage-based pricing means customers pay in proportion to how much of your product they consume, measured in a unit like API calls, storage, seats, or compute time, instead of a fixed monthly fee.
Subscription pricing means customers pay a fixed amount on a fixed schedule, regardless of whether they use the product once or a thousand times that month.
Founders usually discover the difference the hard way. On a Hacker News thread with over 100 upvotes, a founder asking whether to launch with subscription or usage-based pricing got the same answer from a dozen operators: start with a subscription, because it is easier to sell, easier to explain, and easier for a first-time founder to manage without billing infrastructure. Usage-based pricing sounds more "aligned with value," but it adds real technical and sales complexity before you have proven anyone will pay you at all. If you want the fuller picture on setting a price at all, see how we approach B2B SaaS pricing strategy as a starting point.
The framework that actually decides it
Usage-based pricing generally works best when your end user is other software. Subscription pricing generally works best when your end user is a human being.
That distinction, from a16z's growth partners, comes down to friction. Software-to-software products, like infrastructure, data pipelines, or dev tools, tend to have usage that scales exponentially and customers who want to pay only for what they consume. Software-to-human products, like a CRM or a project management tool, have a natural ceiling on usage. No sales rep wants to get charged every time they log an opportunity.
Two quick checks for where you land:
- Lean subscription if: your buyer is a person who logs in to get work done, usage per seat is roughly predictable, and your sales motion depends on a clear, comparable price a buyer can approve without finance dragging out the deal.
- Lean usage-based if: your product powers another system rather than a person, usage can grow 10x or 100x without a human deciding to "use it more," and you already have solid telemetry to measure and bill that usage accurately.
Why most early-stage founders get this wrong
Founders copy the pricing model of the company they admire, not the company they actually are. Seeing Snowflake or Twilio price by usage does not mean a seed-stage product with three enterprise pilots should do the same.
The real cost of jumping to usage-based pricing too early is not customer confusion. It is revenue unpredictability at the exact moment you need predictable revenue most, when you are trying to hit a growth number for your next raise. A subscription gives you a number you can forecast. Usage-based pricing gives you a number that depends on what your customers decide to do this month, and early on, you do not have enough customers for that to average out.
There is a second, quieter cost: billing infrastructure. Usage-based pricing requires accurate metering, invoicing that handles overages, and a sales team that can explain a variable bill, a gap Orb's research on usage-based revenue flags as the top reason companies stall mid-migration. Most pre-seed and seed teams do not have that infrastructure, and building it before product-market fit is a distraction from the thing that actually determines survival, which is finding customers who will pay you anything at all. If enterprise buyers are already asking you for custom terms, pricing your first enterprise deal is the more urgent problem to solve first.
The hybrid model most companies land on eventually
Most companies that start with one pure model end up somewhere in the middle. Chargebee's 2025 State of Subscriptions research put hybrid adoption at 43 percent of SaaS companies, and multiple billing platforms project that figure climbing past 60 percent by the end of 2026 as more products add a usage layer on top of a subscription base.
The sequence that works for most early-stage teams looks like this:
- Start with a flat subscription. Pick two or three tiers based on outcomes your buyer cares about, not feature counts. This is the easiest thing to sell when you have no case studies yet.
- Instrument usage data from day one, even if you are not billing on it. You cannot design a usage-based tier later without months of real consumption data to price against.
- Add a usage component only when a trigger appears: a customer whose usage is clearly outgrowing their plan, an enterprise deal that specifically asks for consumption pricing, or a cost structure where your marginal cost per unit of usage is now material to your margins.
- Keep a subscription floor under any usage component. A minimum monthly commitment protects your revenue forecast while still letting usage capture upside from your biggest accounts.
How Snowflake and Twilio actually price it
Snowflake bills by the terabyte stored and by compute time per second, with no per-seat charge at all, because its buyer is a data platform, not an individual. That model helped push its net revenue retention to 158 percent at scale, well above what most subscription-only SaaS companies report.
Twilio runs the opposite sequence from what most founders expect: it started usage-first, billing per SMS and API call, and only later layered in monthly subscription floors and committed-use contracts with volume discounts for its largest customers. The subscription element was added to give predictability to buyers who wanted it, not because usage-based pricing had failed.
The pattern across both: usage-based pricing companies average roughly 10 percentage points higher net dollar retention than subscription-only peers, because the pricing model itself creates built-in expansion revenue as customers grow. That number is a reason to plan for usage-based pricing eventually. It is not a reason to start there before you have the volume or telemetry to make it work.
What to do in your first 90 days
If you are pre-seed or seed stage with a product used by people, default to a simple subscription with two or three tiers, and stop there. Do not add a usage meter until a specific customer conversation forces the question.
If your product is infrastructure, an API, or anything one piece of software calls on behalf of another, start instrumenting usage now even if you launch with a flat price, because you will need six months of consumption data before any usage-based tier can be priced correctly.
Either path, write down the one metric you would bill on if you ever moved to usage-based pricing. Deciding that in a calm moment beats deciding it under pressure during a renewal negotiation.
Frequently asked questions
Is usage-based pricing better than subscription pricing for SaaS?
Neither is universally better. Usage-based pricing tends to produce higher net revenue retention and works best for software-to-software products. Subscription pricing is easier to sell and forecast, and works best when your buyer is an individual human user.
When should a startup switch to usage-based pricing?
Switch when you have reliable usage telemetry, a clear cost driver tied to consumption, and specific customer demand for consumption-based billing, not before. Most companies add it years after launch, not at day one.
What is hybrid pricing in SaaS?
Hybrid pricing combines a fixed subscription floor with a variable usage component, so revenue stays predictable while still capturing expansion revenue as usage grows. Roughly 43 percent of SaaS companies already use some form of it.
Does usage-based pricing increase net revenue retention?
On average yes. Companies using usage-based pricing report about 10 percentage points higher net dollar retention than subscription-only peers, largely because expansion happens automatically as customer usage grows.
Should an early-stage SaaS startup use usage-based pricing?
Only if your product is infrastructure or API-like and you already have usage telemetry. Most early-stage teams with human end users are better served starting with a flat subscription and adding usage pricing later.
What metric should I bill on if I move to usage-based pricing?
Pick the unit that most directly tracks the value your customer gets, such as API calls, compute time, or data processed, not an arbitrary technical metric that is easy to track but disconnected from value.
Pricing is a decision you get to revisit, not a decision you have to get perfect on day one. Pick the model that matches who is actually using your product today, instrument everything, and let the data tell you when it is time to change. For more on how we work through decisions like this with founders, see our process or apply to work with us.