business-model-canvas

name

business-model-canvas

description

Activate when: user says "business model canvas," "BMC," "Osterwalder canvas," "what's our business model," "map this business," or asks to evaluate/design/pivot/compare a business model before raising capital. Do NOT activate when: no customer segment exists yet (redirect to first-principles first); or the question is purely operational (which CRM, which feature to ship). More: deciqai.com/s/business-model-canvas

Business Model Canvas

Overview

A business creates, delivers, and captures value through nine interlocking blocks (Customer Segments, Value Propositions, Channels, Customer Relationships, Revenue Streams, Key Resources, Key Activities, Key Partners, Cost Structure). The canvas maps each block and exposes which claims are specific enough to be tested vs. still wishful thinking. Introduced by Osterwalder (HEC Lausanne, 2004); popularized in Business Model Generation (Wiley, 2010).

Compose: first-principles to clarify genuine dependencies · occams-razor to compare models · inversion to find the block whose failure kills the model · second-order-thinking to trace block cascades.

When to Use

Apply when: designing a new business · evaluating an existing business for block-level risk · considering a pivot · comparing against incumbents · before raising capital · sizing an AI-native model's economics under AI capex / usage-based margins / AI-native competition (does the foundation-model provider's pricing or its own product roadmap sit in your Cost Structure and Key Partners?).

When NOT to use: no customer named yet (do first-principles first) · canvas done and nothing validated since · purely operational decision · you will fill it in but never test it.

Coaching Novices (Adaptive Front Door)

• Engine mode: user has a concrete case → run The Process directly.
• Coach mode: user is unfamiliar or has no concrete case → guide step by step.

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.

1. One-line what-it-is: nine blocks map how a business creates, delivers, and captures value — so you can see which claims are testable and which are wishful thinking.
2. Check fit against When to Use / When NOT to use. If no customer named, redirect to first-principles.
3. Elicit their real business — one specific segment and one specific value prop before opening the other seven blocks.

[WAIT — do not advance until user responds]

4. Walk the nine blocks one at a time in dependency order (Segments → Value Prop → Channels → Relationships → Revenue → Resources → Activities → Partners → Cost). Force specifics.

[WAIT — do not advance until user responds]

5. Close by naming the single block with the most unvalidated assumption and the next-week experiment to test it.

[WAIT — do not advance until user responds]

The Process

Run the Business Model Canvas Audit — map all nine blocks, tag each claim, surface the riskiest assumptions, name next-week experiments.

Step	Block	Key question
1	Customer Segments	Specific who (not "consumers") — multi-sided businesses list each side
2	Value Propositions	What the customer gets in their own units (time/money/anxiety/status) — not features
3	Channels	Five phases: awareness, evaluation, purchase, delivery, after-sales
4	Customer Relationships	What relationship does each segment expect? (self-service / personal / community)
5	Revenue Streams	Model + price + why they pay — willingness-to-pay must be testable
6	Key Resources	Only the small distinctive set that enables the value prop
7	Key Activities	Only what must be done distinctively well to deliver the value prop
8	Key Partners	Named, with role and replacement risk
9	Cost Structure	Named drivers, fixed vs. variable; value-driven or cost-driven?

After mapping: tag every claim VALIDATED/ASSUMED/DEPENDENT · name the 1–2 ASSUMED claims that would kill the business · design the smallest next-week experiment for each.

Output template

# Business Model Canvas Audit: <business name>
| Block | Specifics | Status |
|---|---|---|
| Customer Segments | … | VALIDATED/ASSUMED/DEPENDENT |
| Value Propositions | … | … |
| Channels | awareness/eval/purchase/delivery/after-sales | … |
| Customer Relationships | … | … |
| Revenue Streams | model + price + why they pay | … |
| Key Resources | … | … |
| Key Activities | … | … |
| Key Partners | named + role + replacement risk | … |
| Cost Structure | fixed/variable drivers | … |

Multi-sided check: [list each side's segment + value prop + revenue]
Load-bearing ASSUMED claims: 1. … 2. …
Next-week experiments: To test <claim>: <smallest experiment by <date>>
Comparison to incumbents: [block-by-block diff vs. closest competitor]

→ Method in Action: Southwest Airlines (1971 → ongoing) · Nespresso's Razor-and-Blades Pivot → 2026 lens: Mapping an AI-Native Startup (2024–2026) — usage/outcome revenue, compute as dominant variable cost, the foundation-model provider as both key partner and key risk

Industry Canvas Packs

Domain	Load-bearing block	Typical ASSUMED claim	Common failure
B2B SaaS	Value Propositions	time-to-value short enough to survive the pilot	scaling sales before validating segment fit
Two-sided marketplaces	Customer Segments (×2)	cold-start liquidity on the harder side	mapping one side's canvas, skipping the other's Segments/Value Props/Revenue Streams
DTC consumer	Channels	CAC below contribution margin at scale	pricing set before any willingness-to-pay evidence; Channels covering purchase but not after-sales
Professional services	Key Resources	the specific people who deliver the value prop are retainable and replicable	selling capacity that walks out the door; Cost Structure left as "salaries" without utilization drivers

Contribute a pack for your domain — see the template at the repo root.

Applying It Well

• Generic words are theater. Replace "consumers," "online," "low cost" with specifics before a block counts as filled.
• The blocks must compose. A great Value Prop + Channels that can't reach the segment is not a business model.
• Multi-sided is the rule; the canvas points to experiments. Platforms have multiple Segments/Props/Revenue Streams — skipping a side breaks the canvas. Surface the 1–2 ASSUMED claims that would kill the business and test them next week.
• Compare block-by-block to incumbents. Identical canvas to a competitor = worse-positioned version of theirs.

→ Primary sources: references/sources.md

Common Rationalizations

[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.

Fake move	Reality
[D] Canvas as decoration — filled in after strategy is set	Canvas is for finding the strategy. Retrofitted = wall poster.
[D] Generic segments — "consumers," "businesses," "everyone"	Generic segments → generic canvas → untestable claims.
[D] Value Prop = feature list	Features = what the product does; value prop = what the customer gets in their own units.
[D] Single-sided thinking in a multi-sided business	Platforms have multiple Segments/Value Props/Revenue Streams — skipping a side breaks the canvas.
[D] Cost Structure left vague ("low" / "high" without driver names)	"Low cost" without named drivers is a wish, not an answer.
[D] Revenue Streams without willingness-to-pay evidence	Pricing without evidence is ASSUMED, not validated.
[D] Key blocks as generic lists · No experiment plan	Only distinctive-set items belong; without experiments, the canvas is a hypothesis.
→ Add [O] entries here after each real use — paste the actual failure pattern	What went wrong and why

Red Flags / Verification

• Canvas filled after strategy was set (decoration) · Segments contain "consumers"/"everyone" · Value Props are a feature list
• Revenue Streams give price without willingness-to-pay evidence · Key blocks list generic inventory · No multi-sided check on a platform
• No ASSUMED claims labeled · No next-week experiment named for load-bearing assumptions
• All nine blocks specific (not generic) · Channels covers all five phases · Each claim tagged VALIDATED/ASSUMED/DEPENDENT
• Canvas compared block-by-block to at least one incumbent's

---

Part of deciqAI Knowledge Skills — 228 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. See it run → https://www.deciqai.com/s/business-model-canvas · Built by deciqAI · github.com/deciqAI · Contributions welcome.

Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/business-model-canvas.json