What a Startup Business Model Actually Is
Founders often conflate three different things: the idea, the product, and the business model. The idea is the insight ("people struggle to find a good barber"). The product is the thing you build (an app that books barbers). The business model is how money moves: who pays, how much, how often, and what you keep after costs.
Getting this distinction right matters because a great idea wrapped in the wrong model dies quietly. Plenty of useful products never found a way to charge that customers tolerated and that left a margin. When you describe your startup, you should be able to finish the sentence "We make money when ___" in one clause. If you cannot, the model is not yet real, only the idea is.
The rest of this post walks through the dominant model archetypes, what each one is good at, and where each one tends to break.
The Subscription and SaaS Revenue Model
In a subscription model, customers pay a recurring fee, monthly or annually, for ongoing access. Software-as-a-Service is the best-known flavor: Notion, Figma, and Spotify all charge you to keep using the thing rather than to buy it once.
The appeal is predictable, compounding revenue. Each retained customer pays again next month, so growth stacks on top of an existing base instead of resetting to zero. That predictability is also what investors price most generously.
The hard part is retention. Subscriptions live or die on churn: if customers cancel faster than you add them, the compounding works against you. This forces a discipline that one-time-sale businesses can dodge, you have to keep delivering value every single billing cycle, not just at the moment of purchase. Subscription only makes sense when the underlying need is recurring; nobody wants a monthly fee for something they use twice a year.
Transactional and Usage-Based Models
Transactional models charge per action. An e-commerce store earns on each sale; Stripe earns a percentage of each payment it processes; a courier app earns a fee per delivery. Revenue scales directly with activity rather than with a subscriber count.
Usage-based pricing is a close cousin that has become popular with infrastructure and AI products: you pay for what you consume, like AWS charging per gigabyte or an API charging per call. It aligns cost to value beautifully, customers who use more pay more, and small users are not scared off by a big upfront commitment.
The tradeoff is volatility. Revenue can swing with seasonality, a single large customer's behavior, or a slow month. You also lose some of the forecasting comfort of subscriptions. Many companies blend the two, a base subscription plus usage overages, to get predictability and upside at once.
The Marketplace Business Model
Marketplaces connect two sides, buyers and sellers, and take a cut of the transactions between them. Airbnb links hosts and guests; Uber links drivers and riders; Etsy links makers and shoppers. The platform usually owns no inventory, which is what makes the model so capital-efficient when it works.
The defining challenge is liquidity, the chicken-and-egg problem. Buyers will not show up without sellers, and sellers will not show up without buyers. Most marketplace founders spend their first year hand-solving one side of this, often by subsidizing or manually recruiting supply in a single city before expanding.
When liquidity tips, the model becomes a near-unbeatable flywheel: more supply attracts more demand, which attracts more supply. But before that tipping point, a marketplace can look like a ghost town for a painfully long time, and many die in that valley.
Freemium, Advertising, and Indirect Monetization
Freemium gives the core product away and charges for premium tiers. Dropbox and LinkedIn built large bases this way: the free tier is the marketing channel, and a small fraction of users upgrade. It works when the free product is cheap to serve and the paid features solve a real pain that intensifies with use, more storage, more seats, advanced analytics.
The risk is the conversion gap. If too few free users ever pay, you are funding a large non-paying base out of a thin paying one. Freemium is genuinely hard precisely because the free tier has to be generous enough to attract users but stingy enough to create upgrade pressure, and that line is narrow.
Advertising flips the relationship: the user is the audience, and advertisers are the customer. This needs scale, large audiences, to generate meaningful revenue, which is why it suits content and social products more than niche tools. A roast of a half-formed model is exactly what PitchRoast was built to deliver before you commit a year to the wrong one. Whatever you choose, name the paying customer explicitly, because in ad and freemium models the user and the buyer are not the same person.
How to Choose a Model and Test Your Unit Economics
Start from customer behavior, not from the model you admire. Ask how the customer already buys this category. People expect to subscribe to software but to pay per ride for transport; fighting that expectation is expensive. Match the payment rhythm to the need's rhythm.
Then sanity-check the unit economics on the back of an envelope. Roughly, what does it cost to acquire one customer, how much gross margin does that customer produce over their lifetime, and how long until you recoup the acquisition cost? You do not need precise figures early on, you need to know the relationship is not obviously upside-down. A model where it costs more to win a customer than they will ever pay is not a business, no matter how elegant.
Finally, remember you can combine archetypes deliberately. A marketplace can add a subscription for power sellers; a SaaS tool can add usage overages. Layering is fine. Copying a competitor's full stack without understanding why each layer exists is how founders end up with a model they cannot explain or defend.