Startup Revenue Models Explained: Picking How You Get Paid
Subscription, usage-based, transaction fee, marketplace take rate, licensing and freemium - what each does to your cash flow, your metrics and your fundraising story.
Founders spend months on the product and an afternoon on how it gets paid for. The revenue model decides your cash flow shape, which metrics matter, how you are valued, and how much capital you need before you are self-sustaining. It deserves more than an afternoon.
Subscription
Customers pay a recurring fee for continued access. The dominant model in software, and the one investors understand best.
What it does well. Revenue is predictable, which makes planning and forecasting tractable. Retention compounds - this year's base is next year's starting point. Valuation multiples for durable recurring revenue tend to be the most generous available.
What it demands. Continuous delivered value, because the customer re-decides every renewal. Churn becomes the number that governs everything: at high churn, you are refilling a leaking bucket and growth costs escalate indefinitely.
Metrics that matter. Monthly or annual recurring revenue, gross and net churn, expansion revenue, payback period.
Usage-based
Customers pay for what they consume - per API call, per transaction processed, per gigabyte.
What it does well. Pricing aligns with value received, so the entry point is low and adoption is easy. Revenue grows with the customer without a renegotiation.
What it demands. Accurate metering, which is a real engineering commitment, and tolerance for lumpy revenue. A customer's bad quarter is your bad quarter. Forecasting is materially harder than subscription.
Common hybrid. A committed base subscription with usage above it. This is where much of infrastructure software has landed, because it gives the vendor a floor and the customer alignment.
Transaction fee
You take a fee per transaction facilitated - a fixed amount, a percentage, or both.
What it does well. Scales directly with customer activity and requires no purchasing decision after the first one.
What it demands. Volume, and defensibility. Percentage fees invite competitive undercutting unless you own something the customer cannot easily route around. Frequency is everything: a business built on an annual transaction has almost no compounding.
Marketplace take rate
A percentage of the value exchanged between two parties you connect.
What it does well. Revenue grows with marketplace liquidity without you carrying inventory or delivery risk.
What it demands. Solving the cold start problem - no supply without demand, no demand without supply - and then defending against disintermediation. Once a buyer and seller have transacted through you, they have every incentive to transact around you next time. Your take rate has to be justified by something that keeps happening: payments, trust, discovery, dispute resolution, guarantees.
Licensing
A fee for the right to use your technology, often perpetual or multi-year, sometimes with maintenance attached.
What it does well. Large upfront payments, strong cash flow early, and a fit with enterprise and government buyers whose procurement is built around capital purchases rather than subscriptions.
What it demands. Long sales cycles, heavy customisation pressure, and lumpy revenue that makes forecasting difficult. Investors typically value it below recurring revenue because renewal is less certain.
Freemium
Strictly an acquisition strategy on top of another model, usually subscription.
When it works. The free tier delivers real value, the constraint that triggers upgrade arrives naturally with usage, serving a free user is cheap, and free users generate distribution - inviting colleagues, creating public content, referring others.
When it fails. Free users cost meaningful money to serve, conversion is low, and support load scales with the unpaid base. Then it is a marketing expense wearing a product's clothes.
Choosing
Three questions do most of the work.
How does the customer receive value - continuously, or at a moment? Continuous value supports subscription. Moment-of-value supports transaction.
What do your costs scale with? If serving a customer costs more as they use more, flat subscription pricing will erode your margin as your best customers grow. Price on the dimension that drives your cost.
How does the buyer prefer to buy? Enterprise and government procurement often cannot process a per-usage bill. Consumers rarely tolerate a licence. Fighting your buyer's purchasing process is expensive and unnecessary.
The mistake to avoid
Copying a model because a company you admire uses it. Their model fits their cost structure, buyer and value delivery. If yours differ on any of those, you have imported a constraint without the reason for it.
Pick the model that matches how your customer actually gets value and how your costs actually behave. Then get very good at measuring the two or three numbers that model lives or dies by.
This is general guidance, not financial advice. Revenue model choice has tax and accounting consequences - take professional advice.
Frequently asked questions
- Can a startup use more than one revenue model?
- Yes, and mature companies usually do - a subscription base with usage overages, or a marketplace take rate plus subscription tooling for suppliers. But at early stage, running two models at once usually means neither is understood well enough to optimise. Get one working, then layer.
- Why do investors prefer subscription revenue?
- Predictability. Recurring revenue with low churn can be forecast, which makes the business easier to value and easier to fund. That preference is real but frequently over-applied: a transaction business with strong frequency and retention can be an excellent company, and forcing a subscription onto a product nobody wants to subscribe to produces churn rather than predictability.
- Is freemium a revenue model?
- Strictly it is an acquisition strategy layered on top of one - usually subscription. It works when the free tier is genuinely useful, the paid trigger is a natural consequence of getting value, and serving free users is cheap. It fails when free users cost real money to serve and convert at a rate too low to cover them.
- When should we change our revenue model?
- When customer behaviour tells you to - you keep losing deals on structure rather than price, or customers keep asking to buy differently, or your costs scale with a dimension your pricing ignores. Changing the model is disruptive and should be deliberate. Changing the price within a model is much easier and is usually tried too late.
