Overview
Based on "The Lean Startup" by Eric Ries and the Build-Measure-Learn framework. Ries argues that a startup is not a small version of a big company - it is an organization searching for a repeatable, scalable business model. The business model canvas is the artifact of that search. Each block is a hypothesis, not a fact. The job is to identify which blocks carry the most uncertainty and run the cheapest possible experiment to resolve that uncertainty before committing resources.
Workflow
Step 1: Fill each canvas block as a hypothesis
Work through all 9 blocks. Write each as a testable belief, not a statement of fact.
Customer Segments
Who specifically suffers from this problem? Not "SMBs" or "enterprise" - get to a job title, company size, industry, and behavior.
Example: "Series A SaaS founders with 5-15 person teams who do not yet have a dedicated finance hire."
Value Proposition
What specific outcome do you deliver, for which segment, better than any alternative?
Format: "We help [segment] achieve [outcome] by [mechanism], unlike [alternative] which [limitation]."
Channels
How do customers find you, evaluate you, and buy? Map the full path: awareness - consideration - purchase - onboarding - retention.
Mark each channel as owned (you control it), earned (press, word of mouth), or paid (ads, partnerships with fees).
Customer Relationships
How do you acquire, retain, and grow customers? Self-serve, high-touch sales, community, automated lifecycle?
State the expected CAC and LTV ratio. If you do not know them, write your assumption.
Revenue Streams
How do you charge? Subscription, usage-based, transactional, licensing, marketplace take rate?
Write the specific price point or range and the expected ACV for each segment. If multiple streams, rank by expected contribution.
Key Resources
What assets are essential to deliver the value proposition? Proprietary data, algorithms, brand, regulatory approval, specific talent.
Flag which resources you currently have vs. which you must build or acquire.
Key Activities
What does [your startup] actually do to create and deliver the value proposition? Product development, content, sales, operations, data curation?
Identify the 2-3 activities that would hurt most if you stopped doing them.
Key Partnerships
Who do you depend on that you cannot or should not own? Suppliers, distribution partners, technology providers, channel partners.
For each: what do they provide, what do you provide them, and what happens if they leave?
Cost Structure
What are the major cost drivers? Fixed (headcount, infrastructure) vs. variable (COGS, commissions, hosting per unit).
Write your monthly burn breakdown by category. Identify the largest single cost and whether it scales with revenue.
Step 2: Rank blocks by uncertainty
After filling all 9 blocks, score each on two dimensions:
- How certain are you this is true? (1 = pure assumption, 5 = validated with data)
- How much does [your startup] break if this block is wrong? (1 = minor, 5 = fatal)
The blocks with low certainty and high impact are your riskiest assumptions. These are what you test next.
Step 3: Identify the single riskiest hypothesis
Pick the one block that, if wrong, would invalidate the entire model.
For most early-stage startups this is one of:
- Customer Segments (do these people actually have the problem at the severity we believe?)
- Value Proposition (does our mechanism actually deliver the outcome better than alternatives?)
- Revenue Streams (will they pay what we need them to pay?)
State it as a falsifiable hypothesis: "We believe [specific claim]. We will know this is true when [observable evidence]. We will know it is false when [counter-evidence]."
Step 4: Design the minimum test
From Ries: the goal is not to build - it is to learn. Design the cheapest test that could falsify the riskiest hypothesis.
Test types by resource cost (low to high):
- Customer interview (the Mom Test approach) - tests whether the problem exists and how acute it is
- Landing page with email capture - tests whether people seek the solution unprompted
- Concierge MVP (manual delivery of the outcome) - tests whether people value the outcome, not the product
- Prototype with 5 users - tests whether the mechanism works
- Limited release to a waitlist - tests retention and word of mouth at small scale
For each test, define in advance: what result would confirm the hypothesis? What result would deny it? Run the test. Do not skip to building.
Step 5: Update the canvas after each learning cycle
The canvas is a living document. After each test cycle:
- Mark blocks as Validated (evidence supports), Invalidated (evidence contradicts), or In Progress
- Update the block content to reflect what you actually learned
- Re-rank uncertainty scores
- Identify the next riskiest hypothesis
A canvas that has not changed in 60 days either means [your startup] has no open questions (unlikely at early stage) or the team has stopped learning.
Step 6: Identify the model's unit economics
Before scaling, calculate:
- CAC: fully loaded cost to acquire one customer (include sales, marketing, onboarding)
- LTV: expected revenue from one customer over their lifetime (ACV x gross margin x average retention in years)
- LTV:CAC ratio: target 3:1 or higher for a viable model
- Payback period: months to recover CAC from gross margin (target under 18 months for most SaaS)
If unit economics do not work at current scale, identify which cost structure or revenue stream assumption needs to change and whether the business model can support it.
Anti-Patterns
1. Treating the canvas as a finished document
Bad: Filling the canvas once at founding and filing it away.
Good: Reviewing and updating after every major customer conversation or experiment.
2. Writing aspirations instead of hypotheses
Bad: Value Proposition block reads "the best solution in the market."
Good: "Reduces [specific workflow] from 4 hours to 20 minutes for [specific segment], validated with 3 customers."
3. Skipping cost structure and unit economics
Bad: A canvas with 8 well-developed blocks and "TBD" in Cost Structure.
Good: A rough burn breakdown by category even before you have real data. Assumptions are better than blanks.
4. Too many customer segments
Bad: Listing 5 segments to hedge uncertainty.
Good: Picking the one segment where the problem is most acute and the willingness to pay is highest. Expand after proving the model works there.
Quality Checklist