The Business Model Canvas breaks a product into nine blocks: who you serve, what value you deliver, how you reach them, how you keep them, how you make money, what you need to build it, what you actually do, who helps you, and what it costs. I use it less as a one-time go/no-go and more as a workflow I come back to at each stage — idea, MVP, first paying customers, growth. Each stage sharpens the answers in a different block.
Read this once linearly. Then come back to one block at a time when that block is what’s blocking your next step.
Customer Segments
Who exactly are you building for?
Example. Airbnb runs two segments — travellers and hosts. Each side got its own value proposition (cheap authentic stays for travellers; spare-room income for hosts) before either flywheel could turn.
What to figure out
- Which specific group has the problem you solve?
- Are there sub-segments worth naming, or one tight beachhead to start with?
- Where do these people gather online?
What to do next
- Describe your ideal customer in one sentence — role, context, the pain.
- Find three places they cluster (subreddit, Slack group, hashtag, conference, app-store category).
- Read 50 reviews or posts where they describe the problem in their own words. Steal their vocabulary.
- Run a 5-question survey in one of those communities (use an open-source form tool if you don’t want Google holding the data) — 20 responses are enough to validate the pain.
- Sketch one persona per segment, grounded in those quotes — not from imagination.
Watch out for
- “Everyone” is not a segment.
- For two-sided products, neglecting one side kills both.
- “I am the customer” — your perspective is one data point, not a sample.
You’re solid when
- You can name the segment in one sentence.
- You can quote the pain in the customer’s own words.
- You know where to find more of them without paying for ads.
Value Proposition
Why would they pick you over what they already use, including doing nothing?
Example. Dropbox’s value prop — “your files, on every device, just work” — was clear enough that a single referral program drove 3,900% growth in 15 months. Users sold it to their friends in one sentence.
What to figure out
- What problem do you solve, and how severe is it?
- What does the customer use today, and what specifically is wrong with that?
- What single sentence captures why you’re different?
What to do next
- Write the value prop as one sentence: “for [segment] who [pain], we are [product] that [benefit], unlike [alternative].”
- List the closest alternatives — spreadsheets, manual workarounds, doing nothing, and (now mandatory) what the customer could rig up themselves in an afternoon with ChatGPT, Claude, or a no-code AI workflow.
- Put the sentence on a landing page with a waitlist signup. Send some traffic. Measure conversion.
- Talk to ten people in your segment. Ask: would you pay for this? If yes, what would you pay? If no, why not?
- Quantify the pain with one number — minutes lost, dollars wasted, errors per month. Use it everywhere.
Watch out for
- “Easy to use” and “saves time” describe almost every product. Get specific.
- Competing on price alone is a bad position unless you have a real cost advantage.
- Overpromising kills retention — claim only what you can deliver from day one.
- A clever user with a frontier LLM may already match a first draft of your product in a weekend. Name the part that can’t be replicated by a prompt — proprietary data, integrations, distribution, trust — and lean on that.
You’re solid when
- The sentence makes one specific promise.
- Strangers in your segment respond with “I need this” rather than “that’s nice.”
- Cold-traffic waitlist or trial conversion is above 10%.
Channels
How do customers discover you, try you, and keep coming back?
Example. Slack grew on a viral team-invite loop — one user invites four colleagues, the team becomes the unit of adoption — plus a self-service web signup. No salesperson touched the small accounts.
What to figure out
- Where does your segment already go for solutions like yours?
- Which one channel could carry you to your first 100 users?
- Is the channel cheap enough that lifetime value beats acquisition cost?
What to do next
- List 5 candidate channels (SEO, ads, communities, partners, referral, content, app store, sales). Score each on reach × cost × your skill.
- Pick the top one. Run a 2-week test with a fixed budget or time box.
- Measure cost per signup and cost per activated user — not just clicks.
- Diversify only after the first channel pays off; one working channel beats five half-running ones.
Watch out for
- Single-channel dependency — an algorithm change can wipe you out overnight.
- Picking channels you like (Twitter is fun) over channels your segment uses (LinkedIn pays).
- Paid ads as the only plan — CAC creep eats the model.
You’re solid when
- One channel reliably brings users at known cost.
- That cost is comfortably below what each user is worth over time.
- You have a second channel in early experiments, ready to scale if the first wobbles.
Customer Relationships
How do you keep customers around once they sign up?
Example. A 5% bump in retention can lift profits 25–95%. Most IT products underspend on retention because acquisition feels more measurable — but the math rewards keeping users far more than adding them.
What to figure out
- Will customers self-serve, need light support, or need a dedicated human?
- What brings them back next week, next month?
- Where in the journey do they currently drop off?
What to do next
- Map the user journey from first visit through month three. Mark each step with a measurable drop-off.
- Pick the biggest drop and fix it before adding anything new.
- Set up one onboarding email or in-app nudge tied to a behavior, not a date.
- Add a help channel that matches your scale: docs and chatbot for self-serve; live chat for SMB; account manager for enterprise.
Watch out for
- Promising white-glove support you can’t staff as you grow.
- Over-personalized emails (“we noticed you haven’t done X”) that feel surveillant.
- Silent churn — users who drift away without ever complaining.
You’re solid when
- You know your retention curve (week 1, week 4, week 12).
- The support model fits the price you charge — high-touch for high-paying, automated for free tier.
- Users find help fast enough that bad reviews don’t mention it.
Revenue Streams
Who pays you, for what, and how often?
Example. Apple’s App Store takes a 30% cut on most in-app purchases (15% for small developers and post-year-one subscriptions). If you sell through the App Store, you keep 70¢ on every dollar — that 30% has to be built into pricing from day one.
What to figure out
- Which revenue model fits the value you deliver — subscription, transaction, license, ads, freemium?
- What will the segment actually pay, based on what they pay today for alternatives?
- Does the math work — does lifetime value clearly exceed acquisition cost?
What to do next
- Pick the primary model. Look at three competitors’ pricing pages — copy what works, change one variable.
- Set three price tiers if subscription, with a clear “most pick this” anchor in the middle.
- Test willingness to pay early — in interviews ask “what would you pay” and watch faces, not just words.
- Build pricing into the model net of fees (Stripe ~2.9%, App Store 15–30%, partner cuts).
Watch out for
- “We’ll figure out monetization later” — ad models need millions of users to matter.
- Pricing too low to seem competitive, then never being able to raise it.
- One whale customer carrying revenue — losing them tanks you.
You’re solid when
- A real person paid the real price without negotiating.
- Lifetime value comes out to at least 3× acquisition cost at a reasonable scale.
- The pricing page is something you’d send a stranger without flinching.
Key Resources
What assets do you absolutely need to deliver?
Example. Google Maps’ moat is geospatial data — Street View cars, satellite imagery, decades of edits. The app is replaceable; the data isn’t. A newcomer can’t ship a credible competitor without that asset.
What to figure out
- What are the three-to-five resources you can’t operate without (tech, data, team, capital)?
- Which do you already have? Which do you need to acquire or build?
- Which one gives you a defensible position over time?
What to do next
- List your key resources by type: technology, data, people, brand, capital.
- For each, mark: already have / can build / must buy or partner for.
- Identify the one resource that compounds — the one that gets better the more you use the product (data, network, content library).
- Protect it: contracts, IP, retention plans for key people, redundancy for critical infra.
Watch out for
- Building on someone else’s API as your core — when they change terms, you break.
- Underestimating data — slow to acquire, hard to copy once you have it.
- Treating code as a moat. AI shrinks “clone this feature” from months to days. The moat is in the data, integrations, workflow, and trust around the code, not the code itself.
- Hiring senior specialists for problems a mid-level can solve.
You’re solid when
- You can name the resource that makes you hard to copy.
- You control or have stable access to every resource on the list.
- Adding the next customer doesn’t require adding a proportional amount of resource.
Key Activities
What do you and your team actually need to do well, every week?
Example. Slack’s three weekly activities for the first years: ship product improvements, keep the service up (downtime in a chat tool is fatal), and onboard enterprise teams into power use. Everything else was secondary.
What to figure out
- Which two or three activities directly create the value customers pay for?
- Which can you automate, which can you outsource, which must stay in-house?
- Where will execution risk bite you first?
What to do next
- List every activity your team does in a week. Sort by impact on the value prop.
- Star the top three — those are your key activities. Everything else is noise or support.
- For each, decide: do it manually (early), automate it (when patterned), outsource it (when not core).
- Set one metric per key activity (deploy frequency, signup rate, ticket resolution time).
Watch out for
- Treating everything as priority — three is the limit.
- Adding features that don’t tie to a key activity for your value prop.
- Founder-as-bottleneck — if only you can do a key activity, you can’t grow.
You’re solid when
- You can name the three activities and what they look like done well.
- Each has a metric and someone responsible.
- The team’s calendar reflects those three — not the urgent thing of the day.
Key Partnerships
Who outside the company makes the model work?
Example. Uber’s drivers aren’t employees, they’re partners — the entire supply side of the marketplace. Driver subsidies were marketing spend, paid back through ride volume. Without the partner side, the product is a button that does nothing.
What to figure out
- Which capabilities, distribution, or supply must come from outside?
- Which partner relationships are critical (you fail without them) versus convenient?
- What’s in it for the partner?
What to do next
- List partners by role: tech (APIs, infra), foundation models (OpenAI, Anthropic, open-weights you self-host), distribution (channels, resellers), supply (the other side of a marketplace), credibility (logos, certifications).
- For each critical partner, write a one-line “why they’d partner with us.”
- Avoid exclusivity unless it’s symmetric — they can’t drop you, you can’t be dropped.
- Have a Plan B for any single-source dependency (cloud provider, payment processor, data feed).
Watch out for
- Building a feature your platform partner will absorb once you prove it works.
- Single-cloud, single-API, or single-model lock-in without a migration path. Foundation-model providers change pricing, rate limits, and behavior on their own schedule — keep one fallback model wired up.
- “Once Apple bundles us…” — assuming a partnership that hasn’t been signed.
You’re solid when
- You know which partners are critical and which are convenient.
- Critical partners have signed terms, not handshakes.
- Each critical dependency has a documented alternative.
Cost Structure
Where does the money actually go?
Example. Uber spent — and for years still spent — heavily on driver and rider subsidies. That cost lived in marketing, not in COGS, but it was the price of growing both sides of the marketplace fast enough to stay ahead of competitors.
What to figure out
- What are the three-to-five biggest costs (salaries, infra, marketing, payments, support)?
- Which are fixed (don’t change with scale) versus variable (grow with usage)?
- At what scale does the model break even?
What to do next
- List costs in two columns: fixed monthly and variable per-customer or per-transaction.
- Estimate burn rate at 0, 100, 1,000, and 10,000 active users.
- Identify any cost that scales worse than linearly (support, moderation) — plan automation early.
- Apply for credits where they exist (AWS Activate, Google for Startups, accelerators) — real money for paperwork.
Watch out for
- Forgetting payment fees (~3%), platform cuts (15–30%), and compliance overhead.
- LLM inference costs scale per request, not per user. Model what a power user actually costs in tokens — a free-tier user running an AI loop can cost more than they’ll ever pay.
- Hiring ahead of revenue and burning runway you’ll need at the next milestone.
- Cost assumptions that only work in the optimistic scenario.
You’re solid when
- The cost stack maps cleanly to key activities and resources.
- Variable cost per customer stays below revenue per customer.
- A pessimistic case (half the revenue, 50% more cost) still gives you 12+ months runway.
Which blocks matter when
The whole canvas matters eventually, but not at once. A rough sequence:
- Pre-MVP: Customer Segments, Value Proposition, and a rough Cost Structure. If these three don’t make sense together, build nothing.
- MVP → first users: Channels, Customer Relationships, Key Activities. You need a way in, a way to keep people, and a clear sense of what you’re doing all day.
- First paying customers: Revenue Streams and Key Resources. Now the math has to work, and the assets that compound start to matter.
- Growth: Key Partnerships, plus a full revisit of every block — the answers from earlier stages get sharper or get replaced.
Revisit the canvas every time something blocks you. The block that’s stuck is usually one block that’s gone stale.