Where the money actually comes from
Lesson 10 of 24, level 01Index
The founder wants a premium tier. You have no idea whether that is a good idea, because you have never looked at how the company actually makes money.
Five models cover almost everything
Trace the money
- Customer: Who pays?
- Unit: Seat, usage or sale?
- Revenue: What do they pay?
- Cost: What serves them?
Subscription, transaction or marketplace, advertising, licensing or enterprise, and freemium. Most real companies run two of them at once.
The number that reshapes your roadmap is marginal cost: what does one more active user cost to serve? If it is free you optimise for growth. If it is forty rupees a month you optimise for the users who will pay four hundred.
And in enterprise, the buyer is not the user. Which is why enterprise software is often bad at being used and excellent at being bought. Know which one your feature serves.
A product decision can create user value without creating revenue immediately. Understanding the business model helps you explain who pays, why they pay, and which behavior connects the two.
Use your chosen course project throughout. The additional examples below are fictional practice cases; transfer the method to your own evidence.
Trace the transaction
Draw the payer, user, unit sold, price, and delivery cost. The user and payer may be different people. A team member can love a tool that procurement will not approve. Identify the buying decision and the usage decision separately before interpreting free activity as purchase intent.
Name the revenue mechanism
A subscription, usage fee, transaction commission, and service contract each reward different behavior. Describe the actual charging rule for your chosen product using public pricing or a clearly labelled assumption. Revenue is not profit, and a prepaid annual contract is not the same thing as monthly cash arriving forever.
Connect a feature to a mechanism
Spell out the chain from the feature to an observable behavior and then to a business outcome. Each arrow is an assumption to test. Better onboarding might improve activation, but revenue also depends on eligible users converting, price, and retention. Avoid multiplying several optimistic assumptions and presenting the result as a forecast.
Follow one transaction
- Payer: Who approves?
- Unit: What is charged?
- Revenue: Price times units
- Cost: What serves them?
How Netflix and Spotify handled it
Netflix and Spotify
Both are subscriptions, both stream media, and their product decisions look nothing like each other.
Netflix owns most of what it shows, so an extra hour watched costs it roughly nothing and its product goal is engagement. Spotify pays rights holders per stream, so an extra hour costs real money and its goal is retention at the lowest cost to serve. That is why it pushed so hard into podcasts, where it owns the content.
Two identical looking companies will make opposite calls, and both will be right, because their marginal costs are different.
Doing it with AI, and where it breaks
Pick a product you use daily. Ask a model to lay out its revenue lines, then to name the product decisions that only make sense given those lines. Verify two claims against a primary source.
Models confidently invent revenue splits. They will give you a precise sounding “roughly 60 percent of revenue” with no source. Treat every number as a hypothesis until you find it in a filing, an earnings call, or a first party post.
Your AI workbench
Start with your own notes or clearly labelled practice data. Remove private details before sharing. Replace the placeholders, run the prompt in your chosen AI tool, and keep the output beside its source.
Map the business model from these supplied pricing facts [paste]. Separate user, payer, billing unit, revenue mechanism, and major variable costs. Mark unknowns. For my feature [describe], write the shortest plausible chain to revenue and a way each link could fail. Do not invent customer economics.
Before you use the output
- Pricing facts have a dated source
- Revenue and contribution are distinct
- Every projected effect is labelled an assumption
Stuck? Try this next
If AI cannot access the pricing page, paste the relevant public terms yourself. Ask it to calculate one customer example before discussing the entire market.
Keep a brief AI log: input used, useful output, what you checked, and what you rejected. The decision remains yours.
Build it
A one page business model teardown with two claims verified against primary sources.
Checkpoint
If you did the build, these take two minutes. If you cannot answer one of them, that is the part to go back to.
Moving on marks this lesson complete. Finish the build first, it is the part that counts.