Revenue models: how businesses make money

0:00 Revenue models

Two companies can sell exactly the same product and end up with completely different businesses, just because of how they charge. One sells once and hunts for new customers every month. The other earns every month from the same customers. In this lecture, you'll learn the main revenue models, how to choose one that matches how customers buy and experience value, and how AI products are changing pricing, with credits, usage and even outcome-based fees. You'll also learn a quick stress test that shows whether your model survives your heaviest users.

0:40 Why it matters

Why does this matter? Because your revenue model shapes almost everything else: cash flow, how you sell, how you support customers, and what your business is worth. Recurring revenue is prized because it's predictable, but only if customers keep getting value. Churn compounds quietly. If you lose five percent of customers every month, after a year you've lost nearly half of the original group. Here's the key idea. The best revenue model charges in the same rhythm and unit in which the customer experiences value.

1:17 The transport analogy

Here's an analogy. Think about how you pay for transport. A taxi charges per trip. A monthly travel pass charges a flat fee for unlimited rides. A car lease charges a fixed amount plus a fee if you exceed the mileage. None is right or wrong. Each fits a different pattern of use. Business revenue models are the same. One-off sales are the taxi. Subscriptions are the travel pass. Hybrid models, a platform fee plus usage above an allowance, are the car lease. And there are others: commission on transactions for marketplaces, freemium, advertising, licensing, and product plus service.

2:00 Why AI changes pricing

Now, why are AI products changing this? Because the cost of serving a customer depends on usage. With traditional software, one more login costs almost nothing. With an AI product, every task calls a model, and every call costs money, often priced per token, which is roughly a piece of a word. So flat per-seat pricing can quietly lose money on heavy users. That's why you see tiered plans with credits, pure usage pricing, hybrids with an allowance and overage, and outcome-based pricing, such as paying per resolved support ticket. Each trades predictability for fairness in a different way.

2:43 Worked example 1: a Dubai studio (illustrative)

A simple worked example. A fitness studio in Dubai, illustrative. Option one, pay per class: flexible for customers, but revenue is unpredictable and marketing must constantly refill classes. Option two, monthly membership: builds habit and predictable income, but churn has to be managed. Option three, class packs of ten, valid for three months: upfront cash, and some classes go unused. The studio chose a membership as the core, with class packs for occasional visitors. Why? Because regulars experience value every week, which matches a monthly rhythm, while occasional visitors experience it per class. Two segments, two rhythms, two models.

3:26 Worked example 2: stress test (illustrative)

Now the realistic AI scenario, all numbers illustrative. Maya's startup in London offers an AI tool that drafts replies to customer emails, for twenty-nine pounds a month per user. Let's stress-test it. A light user sends fifty tasks a month. At two pence of model cost per task, that's one pound. Add two pounds of other variable costs and about eighty-seven pence of payment fees, and contribution is about twenty-five pounds. Healthy. A typical user sends three hundred tasks. Contribution drops to about twenty pounds. Still fine. But a heavy user sends two thousand tasks. That's forty pounds of model cost alone. Contribution is minus fourteen pounds. Every heavy user loses her money.

4:15 Watch me: fixing the model

Watch me fix it. I open the stress-test sheet from the lesson. Option one, add an allowance: five hundred tasks included, then a small fee per extra hundred tasks. I recalculate the heavy user and the contribution turns positive. Option two, add a higher tier for power users at a higher monthly price with a bigger allowance. Option three, an engineering change: route simple tasks to a cheaper, smaller model, and keep the expensive model for complex ones. I model each option in a separate column. Then, and this matters, I don't just pick the best-looking spreadsheet. I test with customers which option they accept, because a model they hate won't save anyone.

5:04 Five questions to choose a model

So how do you choose? Ask five questions. One, how does this customer prefer to buy? Small businesses often like predictable monthly fees, larger companies prefer annual contracts, and consumers often want pay as you go. Two, how do they experience value? Continuously, per event, or in proportion to use? Three, what cash flow do you need? Annual prepayment brings cash in early, which can fund growth, while monthly billing lowers the barrier to trying you. Four, what do the alternatives charge, and how? Sometimes a different model is itself the differentiation. And five, what does each customer cost you to serve, and how much does that vary? That fifth question is where AI products most often get caught out.

5:56 Monthly vs annual

One more design choice deserves its own moment: monthly versus annual billing. Annual prepayment, often with a modest discount, improves cash flow and usually reduces churn, because customers commit for longer. But it raises the barrier to buying, and it creates a refund and renewal conversation a year later. Monthly billing is easier to start and lets customers leave easily, which keeps you honest about value. Many businesses offer both and let customers choose, then watch which segments pick which. Just be careful with the maths. Revenue collected upfront isn't earned yet. You still owe the customer the service for the rest of the year, so don't spend it as if it were profit.

6:46 Check your payment rails

Before you choose, check your payment rails. Your model has to work with payment methods you can actually use and your customers prefer. In the UK and US, cards and direct debits dominate recurring billing. In the UAE and Saudi Arabia, cards and digital wallets are widespread, and buy-now-pay-later is common in e-commerce. In Pakistan, bank transfers, cash on delivery and mobile wallets matter, alongside the Raast instant payment system. And processors differ by country. For example, at the time of writing, Stripe supports businesses in the UK and UAE but doesn't list Pakistan as a supported country. Always check the provider's current availability page.

7:32 Common mistakes

Let's cover the common mistakes. Copying a model from another market without checking how customers there like to buy. Freemium with no clear route to paid. Flat pricing that heavy users make unprofitable, which is the classic AI trap. Ignoring the cash effect of monthly versus annual billing. Outcome-based pricing without a clear, agreed definition of the outcome, which leads to disputes. And designing a model around a payment processor that isn't available to you. Test your model like any assumption: offer monthly and annual options, pilot usage pricing with a small group, and measure uptake, retention and cash.

8:15 Recap and try this now

Let's recap. Your revenue model should charge in the rhythm and unit in which customers experience value. For AI products, remember that usage drives cost, so stress-test light, typical and heavy users, and use allowances, tiers or smarter model routing to protect margin. Check your payment rails before you commit, and test your model with real customers. Your try this now: build the three-column stress test for your idea with your own numbers, and write two alternative revenue models with the main risk of each. Next module: unit economics and pricing.