Nine boxes. That's the whole thing. If you have ever sketched a business model on a single sheet of paper and felt, ten minutes later, that you understood your own company more clearly than after three days of writing a business plan, you know what the exercise does.
But most explanations of the Business Model Canvas stop at the labels. They list the nine blocks, paste a template, and call it a day. Which is useless if you actually want to fill one in for a real company and not produce a wall of vague nouns.
Here is what I want to do differently: walk you through the canvas as a working instrument, with two real examples filled in block by block, the mistakes I made when I started using it, and the reasons it sometimes fails you.
Key takeaways
- The canvas is a thinking tool, not a deliverable — its value comes from the arguments it forces, not the pretty boxes.
- Fill it in a specific order: customers first, revenue last. Starting with Key Partners is the classic beginner error.
- The nine blocks describe one customer segment. If you have two distinct segments, you need two canvases.
- Netflix and Amazon are not the same type of business even though both fit the canvas — the canvas hides the difference if you let it.
- Use a Lean Canvas instead if you are pre-market; the classic canvas is a better fit once you have paying customers.
Business model canvas explained: what the nine boxes actually ask you
Alexander Osterwalder and Yves Pigneur published the canvas in Business Model Generation. The core idea is simple enough: describe how an organisation creates, delivers and captures value, on one page, so that a team can argue about it out loud instead of reading a 60-page document nobody finishes.
The nine blocks sit in three zones. The left side is how you produce value. The right side is how you sell it. The middle is what you sell.
- Customer Segments — who exactly, described narrowly. "SMEs" is not a segment.
- Value Proposition — the specific problem you solve, phrased as a benefit the customer recognises.
- Channels — how the customer hears about you, buys, and gets support.
- Customer Relationships — self-service, dedicated account manager, community, automated? Each costs differently.
- Revenue Streams — subscription, one-off sale, commission, licensing, advertising. And who pays.
- Key Resources — what you absolutely must own or hire: a warehouse, a licence, a dataset, four great engineers.
- Key Activities — the two or three things you do better than anyone, or the model collapses.
- Key Partners — who you depend on, and what you hand over to them.
- Cost Structure — fixed vs variable, and what drives cost up as you grow.
Nine blocks is a lot to hold in your head. In practice, the canvas works because the blocks constrain each other. Change Revenue Streams from one-off sales to a subscription, and Customer Relationships, Key Activities and Cost Structure all have to move. That interlock is the real content. The boxes are just where the consequences show up.
Why nine blocks instead of a business plan?
Because a business plan freezes assumptions into prose, and prose is very hard to challenge in a meeting. A canvas puts each assumption in its own cell where a colleague can point at it and say "that's wrong."
I have watched this play out in workshops. Give a team a 40-page plan and the discussion drifts to formatting. Give them a canvas and within twenty minutes someone is arguing about whether Key Partners really includes their cloud provider or whether that's just a cost. That argument is worth more than the document.
Netflix, block by block: a subscription canvas that actually works
Take Netflix. It is the cleanest example I know for teaching the canvas, because every block follows logically from a single decision: the company sells access, not copies.
The right side: who pays and why
Customer Segments — households with broadband, split further in practice by region and language. Value Proposition — unlimited on-demand viewing without scheduling, at a flat monthly fee, with enough original content that cancelling feels like a loss. Channels — the app on the device the viewer already owns, plus the recommendation surface itself, which does most of the selling. Customer Relationships — fully automated; the only human interaction is a support ticket. Revenue Streams — monthly subscription, tiered by resolution and simultaneous screens.
Notice what the value proposition is not: it isn't "great films." It's convenience plus breadth. The moment you write "great content" in that box, the canvas stops telling you anything, because every competitor would write the same sentence.
The left side: what it costs to keep that promise
Key Resources are the content library and the encoding and recommendation infrastructure. Key Activities are licensing, commissioning originals, and running the delivery network. Key Partners are studios, production companies, and the cloud and CDN providers carrying the traffic.
Cost Structure is where the canvas earns its keep. Netflix's costs are overwhelmingly fixed and front-loaded: content is paid for before a single viewer presses play. Every new subscriber after that point is close to pure margin, which is precisely why the model scales — and precisely why a sudden drop in subscribers hurts so badly. One block explains the other.
Amazon: the same canvas, a completely different machine
Run the same exercise on Amazon and the blocks look superficially similar. They are not.
| Block | Netflix | Amazon (retail + marketplace) |
|---|---|---|
| Value Proposition | Unlimited viewing, flat fee | Vast selection, fast delivery, low prices |
| Revenue Streams | Subscription only | Product margin, third-party commissions, advertising, cloud services |
| Key Activities | Licensing and commissioning content | Logistics, marketplace operations, demand generation |
| Cost Structure | Fixed, front-loaded content spend | Mixed: warehouses and delivery are variable with volume |
| Customer Relationships | Fully automated | Automated plus a large returns and support operation |
Both models fit the canvas. That's the point and the trap. The canvas will not tell you that one is a content business with software economics and the other is a logistics business with a marketplace bolted on. It will only show you where to look.
The multi-sided trap most people fall into
Amazon's marketplace is a second customer segment — the third-party seller — with an entirely different value proposition ("reach millions of buyers, use our fulfilment") and a different revenue stream (commission plus fees). If you cram buyers and sellers into one canvas, both Value Proposition and Revenue Streams become incoherent mush.
The rule: one canvas per segment. Osterwalder calls these multi-sided models, and they are the single most common reason a canvas "doesn't work" for someone. It isn't the tool. It's that they tried to describe two businesses in one sheet.
How to actually fill one in (and the order that saves you time)
I got this wrong for months. My first canvases started on the left, because the left side felt like the "real business" — the operations, the partners, the costs. The result was a beautiful description of a machine with no customer attached to it.
Fill it in this order instead:
- Customer Segments — one narrow group, named specifically.
- Value Proposition — the problem, in the customer's words, not yours.
- Channels and Customer Relationships — how they find you and how you keep them.
- Revenue Streams — who pays, how often, how much.
- Only now: Key Resources, Key Activities, Key Partners, Cost Structure.
The logic is that the left side exists to serve the right side. Build the demand side first and the supply side has a job to do. Build it the other way round and you get a supply chain in search of a market.
Four mistakes I made, in order of embarrassment
Writing adjectives instead of mechanisms. "Strong brand" in Key Resources says nothing. "Exclusive 10-year distribution agreement for the region" says something you can test.
Treating it as a one-off. A canvas is a snapshot of assumptions. My first one was accurate for about five months before a channel shift made three blocks obsolete. Revisit it when something structural changes, not on a calendar.
Ignoring the numbers. A canvas with no cost drivers and no revenue per customer is a diagram. Add two figures per block where you can — even rough ones from your own operations.
Filling it alone. The tool's whole design assumes argument. Done solo, you just confirm what you already believed.
Where the canvas fails you
Honestly, the canvas has real blind spots, and pretending otherwise does you no favours.
It is static. Nothing in the nine blocks captures competition, timing, or how a market shifts under you. It is weak on platforms and ecosystems, where value is created by parties you don't employ or pay. And it says nothing about your unfair advantage — the thing that stops a well-funded competitor from copying the sheet next quarter. That gap is exactly why Ash Maurya built the Lean Canvas, which swaps Key Partners and Key Activities for Problem and Unfair Advantage, and which is the better instrument before you have paying customers.
Use the classic canvas once you have revenue and need to see the whole machine. Use Lean Canvas while you are still hunting for a problem worth solving.
The box that should scare you
If you take one habit away from this, make it this: after filling in all nine blocks, cover the sheet and try to say out loud, in one sentence, why a customer picks you over the alternative they already use.
If the sentence comes easily and matches what's written in Value Proposition, the canvas is doing its job. If it doesn't — and the first three times I tried this, it very much didn't — you have found the real work, and it was never in the boxes.
