Design the business: the model on one page

• Right half is value and revenue: segments, the value promised to them, the channels and relationships that deliver it, the revenue that returns. This is the theater - what the customer sees.
• Left half is cost and capability: the partners, activities, and resources that make the promise deliverable, and what they cost. This is the backstage.
• Coherence between blocks is the test. A canvas is not graded block by block; it is graded on whether the blocks explain each other. High-touch relationships + a $33k median price + a self-serve dream do not cohere - something has to give.
• Fill it fast, argue it slow. Ten minutes to draft, an hour to interrogate. The draft is cheap; the argument is the work.

Put the whole machine on one page and two blocks visibly grind:

• Channels fight the value proposition. The promise is "the full digital layer for your consumers" - but a third of delivery runs through partner resellers with no activation playbook. Session 3 proved the cost: partner-onboarded clients stay single-module and churn. The channel block is quietly un-delivering the value block.
• Revenue has no cheap entry point. Subscription + take rate at a $33k median works for committed clients - and hands every price-sensitive prospect to the rivals at 40% less. The model has a front door and no side door.

A Lean Canvas is filled in a deliberate order - problem first, because everything else is downstream of being wrong about the problem. Problem → segment → unique value proposition → solution → channels → revenue/cost → key metrics and unfair advantage as the honesty checks near the end.

Himalaya's starter-tier canvas, drafted:

• Problem: SMB-end prospects are priced out at $33k and scared off by weeks-long onboarding - they default to the 40%-cheaper rivals.
• Segment: smaller consumer businesses Himalaya currently loses before the first call.
• Solution: a self-serve, usage-priced starter tier - one core module, activation in days, no CSM required.
• Key metric: activation to a second module within 30 days (the number Session 3 proved predicts retention).
• Unfair advantage: the existing 8M consumer touchpoints - data-powered defaults no AI-native rival can ship on day one.

The canvas exists to expose the one assumption that kills the bet if false. Not the ten assumptions - the one. For the starter tier it is this: "self-serve clients will activate without human onboarding." If that is false, the tier just industrializes the exact failure mode Session 3 diagnosed - clients who never activate, never form a habit, and churn.

• Rank assumptions by kill-power × uncertainty. "SMBs want cheaper software" is near-certain - do not test it. "They will self-activate" is uncertain AND fatal - test it first.
• Design the cheapest honest test. Canon for Himalaya: a 50-client self-serve pilot, with activation-to-second-module inside 30 days as the kill metric. Weeks and one squad - not a replatform.
• Pre-commit the kill line. Decide before the data arrives what number means stop. A pilot without a kill line is a pivot in disguise.

• Start on the circle, always. Jobs first ("run the consumer experience without an IT team"), then the pains that job produces, then the gains the client is really buying. If the circle is guesswork, the square is fiction.
• Claim fit left-to-right. Each pain reliever points at a named pain: self-serve setup → weeks-of-onboarding pain; usage pricing → paying-for-unused-modules pain. An arrow with no target pain is a feature looking for a reason.
• Verify with evidence. Every relief claim owes you a churn-log entry or an interview quote. "Clients hate slow onboarding" is a hypothesis; the Session 3 churn logs showing partner-onboarded clients never activating is evidence. No quote, no fit - just hope with a diagram.

Christensen's famous study: a fast-food chain optimized milkshakes by asking segments what flavors they wanted - nothing moved. Observing instead, they found morning buyers "hired" the shake for a boring commute: one-handed, slow to finish, tides you to lunch. The competition was not other shakes; it was bagels and bananas. Customers hire products for jobs; demographics describe who they are, not why they buy. Himalaya's clients do not want "a platform" - they hire it to run a consumer experience without building an IT team. Interview for the job ("walk me through the day you decided you needed this") and the VPC circle fills itself with facts instead of personas.

Porter's other canvas (1985): lay out the chain of activities from raw input to delivered value and ask where margin is actually earned. For Himalaya: build platform → sell → onboard → activate → support → expand. Two reads jump out:

• Himalaya's margin lives late in the chain - activation and expansion (multi-module NRR 112%) - yet its investment is concentrated early, in building and selling. The onboarding link, where Session 3 found the leak, is the cheapest link and the one that breaks the rest.
• The AI-native rivals rebuilt the chain, not the product. Their advantage is not a better module - it is deleting the onboarding link entirely (self-serve, days). When a competitor is cheaper, ask which link they removed before asking which feature they added.

Session 4 nominated the consumer-data asset as a moat. Jay Barney's VRIO is the cold shower every claimed moat deserves - four questions, in order, and the moat must pass all four:

Verdict: VRI yes, O no - which converts "we have a moat" into an org-design action, not a comfort blanket. Most claimed moats die on R or I; the honest ones often die on O.

Every canvas block eventually has to survive arithmetic. The unit is one client; the question is whether one client is worth acquiring.

• Contribution margin per client: what a median $33k/yr client leaves after the cost to serve them (cloud, support, CSM time). This funds everything else.
• The squeeze Himalaya is in: 14% churn cuts expected client lifetime to roughly seven years at best - and the churners leave early, before payback. Meanwhile a lengthening sales cycle and a falling win rate push CAC up. Margin down, CAC up: payback stretches from both ends.
• Why the starter tier must be usage-priced AND self-serve: a cheaper tier with the same sales-led CAC and CSM cost is arithmetic suicide. The tier only works if acquisition and service costs fall further than price does - which is exactly what the 50-client pilot must measure alongside activation.

Session 7 builds this into full LTV/CAC and cohort math; for now, the rule: no canvas is validated until the unit economics of one customer close.