Two companies can sell the identical thing and run on completely different logic. Sell a software licence once and your team lives for the next deal; rent that same software monthly and the team lives for the next renewal. The product didn't change. The revenue model did, and with it, almost everything about how the business behaves.

The quick version

  • A revenue model is how you charge, not how much. It decides what your whole organisation ends up optimising for.
  • The six common patterns, subscription, transactional, marketplace, freemium, licensing, ads, each reward a different behaviour: keeping customers, closing sales, growing both sides, converting free users, protecting IP, or selling attention.
  • Most durable businesses run two or three of these together (a free tier that feeds a subscription; a marketplace that also runs ads). The art is matching the model to how your customer actually wants to buy.
  • Pick the model first, then the price. Pricing is a separate, downstream decision, see pricing strategies and pricing mechanisms.

The idea in depth

In the Business Model Canvas, Alexander Osterwalder and Yves Pigneur's now-standard framework from Business Model Generation (2010), "revenue streams" is one of nine building blocks, and it sits directly opposite the cost structure. That placement is the whole point. Your revenue model is the mechanism that converts a value proposition into money, and it has to fit the customer's willingness to pay on the other side. Get it right and the rest of the canvas pulls in one direction. Get it wrong and you can have a brilliant product nobody pays for the right way.

Here are the six patterns, and, more useful, what each one actually rewards.

flowchart TD
  Q(["How does the customer get value?"]) --> A(["Ongoing access over time"])
  Q --> B(["A discrete purchase or transaction"])
  Q --> C(["Connecting them to someone else"])
  A --> S(["Subscription, recurring fee"])
  A --> L(["Licensing, pay to use the IP"])
  B --> T(["Transactional, pay per unit/use"])
  B --> F(["Freemium, free core, paid upgrade"])
  C --> M(["Marketplace, take a cut of the deal"])
  C --> AD(["Advertising, sell the attention"])
A rough decision tree: the model usually follows how the customer experiences value. Leaders Loop

Subscription, transactional and licensing: charging for your own product

Subscription charges a recurring fee for continued access, Netflix, your gym, most SaaS. Its defining trait is that revenue compounds: you start each month with last month's customers, add new ones, and subtract the ones who leave. Tien Tzuo, who built Zuora around this shift, argues in Subscribed (Portfolio, 2018) that it reorients the entire company away from one-off transactions and toward an ongoing relationship. So the move is: if you go subscription, retention is the business. The classic Bain finding by Frederick Reichheld and Earl Sasser, "Zero Defections," Harvard Business Review, 1990, is that lifting customer retention by five percentage points can raise profits by 25% to 95%, because the cost of acquiring a customer is amortised over a longer life. In a subscription model that maths isn't a nice-to-have; it's the engine.

A revenue model is a set of incentives wearing the costume of a billing method.

Transactional charges per purchase or per use, a coffee, a flight, an API call, pay-as-you-go cloud. It rewards volume and conversion: every sale stands on its own, so the team optimises the moment of purchase rather than the relationship. So the move is: watch for lumpiness. Transactional revenue is honest (customers pay when they get value) but volatile, which is exactly why so many transactional businesses bolt on a subscription tier to smooth the curve.

Licensing charges for the right to use intellectual property, patents, a brand, software you install rather than rent, ARM's chip designs, a franchise. It's capital-light and scales without you doing the production, but you give up control of the customer relationship and the experience. So the move is: licensing is strongest when your IP is genuinely hard to replicate and someone else is better placed to deliver it. Where it's weak, the licensee eventually asks why they're paying you at all.

An honest limitation: none of these three tells you what to charge, and a strong model paired with bad pricing still fails. The model decides the shape of the money; unit economics decide whether that shape is profitable.

Marketplace, freemium and ads: charging for access to others

Marketplace connects buyers and sellers and takes a cut, the "take rate" or "rake." a16z, in its "13 Metrics for Marketplace Companies," defines take rate as the share of gross merchandise value the platform keeps, and treats it as a read on how much value the marketplace itself adds. The temptation is always to take more. Bill Gurley's essay "A Rake Too Far" (Above the Crowd, 2013) is the corrective: because your rake is baked into the price the buyer sees, a high rake makes everything on your platform look expensive and "create[s] a natural impetus for suppliers to look elsewhere." So the move is: set the take rate by how much of the transaction you genuinely handle, payments, trust, logistics, matching, not by how much you can get away with. Marketplaces that do a lot of that work sustain take rates in the 10–30% range; ones that mostly introduce two parties cannot.

Freemium gives the core away free and charges for upgrades, Spotify, Dropbox, Slack. The free tier is a marketing cost that buys reach; a small slice converts to paid and funds the rest. The hard truth is how small that slice is: across SaaS, free-to-paid conversion typically runs in the low single digits, commonly around 2–5% (First Page Sage's 2026 benchmark report). So the move is: freemium only works when free users are cheap to serve at scale and the paid upgrade solves a problem the free tier deliberately leaves open. If serving free users is expensive, or the free tier is too generous, you've built a charity.

Advertising makes the product free and sells the audience's attention to a third party, Google, Meta, free-to-air TV. It rewards engagement above all, which is its quiet hazard: the user becomes the inventory, and the incentive to maximise time-on-site can pull against the user's actual interests. So the move is: ads work at scale and where attention is abundant; below real scale, the per-user revenue is so thin that most teams are better off charging the user directly.

flowchart LR
  V(["Free core product"]) --> U(["Large free user base"])
  U --> C(["~2–5% convert"])
  C --> P(["Paying subscribers"])
  U --> W(["~95%+ stay free"])
  P --> R(["Recurring revenue funds the free tier"])
  R --> V
Why freemium is really a subscription business with a wide funnel: a thin conversion layer carries the whole model. Leaders Loop

A worked example

Picture a small company that has built a good invoicing tool for tradespeople. (All figures below are illustrative.) They could run it four ways. As a one-off licence at £200, 1,000 customers gives £200,000 once, then they have to find 1,000 more people next year just to stand still. As a subscription at £15/month, those same 1,000 customers are worth £180,000 a year and, if 90% renew, most of next year's revenue is already in the building before sales does anything. As freemium, they give the basic tool away, reach 20,000 users, and convert 4% to a £20/month plan, 800 payers, £192,000 a year, plus a pipeline of 19,200 who might upgrade when their business grows. As a marketplace, they connect tradespeople to customers and take 12% of each job; the model now lives or dies on transaction volume, not seat count.

Same product, four businesses. The subscription version makes retention the job. The freemium version makes onboarding and the free-to-paid nudge the job. The marketplace version makes liquidity, enough buyers and sellers to keep deals flowing, the job. Choosing the revenue model is choosing which problem your team will spend the next five years solving.

Frequently asked questions

What's the difference between a revenue model and a pricing strategy?

The revenue model is the structure, subscription vs transactional vs marketplace. Pricing is the number and method within that structure, value-based, cost-plus, tiered, and so on. You decide the model first, then the price. See pricing strategies and monetisation & packaging.

Can I use more than one revenue model?

Most strong businesses do. Freemium feeds a subscription. A marketplace runs ads on the side. A SaaS tool licenses an enterprise on-prem version. The risk isn't mixing models, it's mixing them without noticing they reward conflicting behaviours, so your team gets pulled in two directions at once.

Why is subscription so popular now?

Predictable recurring revenue is easier to forecast, finance and value, and digital delivery made it cheap to bill monthly. But it isn't free money: it only works if customers keep getting enough value to stay, which is why churn is the metric that haunts every subscription business.

Is advertising a "bad" revenue model?

No, it funds an enormous share of the free internet. The caution is structural: when the advertiser pays, the user's interests and the business's interests can quietly diverge. That's a trade-off to manage deliberately, not a reason to avoid it.

How do I actually choose?

Start with how your customer experiences value, ongoing access, a one-off purchase, or a connection to someone else, and let that point you at the model (the first diagram above). Then pressure-test it against your unit economics: can you serve customers profitably at the price the model implies?

Related in the Toolkit

Where to go next