Unit Model / Economies of Scale / Business Model (Cross-Author Thematic Synthesis)

When to open: When you need to break down a business model / calculate profit per SKU / select the minimum unit / want to understand that "scale effect is the gravity of the business world" / want to replicate 50 stores but fear the unit does not hold up. This wiki integrates the three perspectives of yitang Truman, Wang Huiwen, and Wang Kai.

One-sentence core:Truman: "The unit model is the most important ROI model within a company";Wang Huiwen: "Scale effect is basically the gravity of the business world";Wang Kai: "Do not challenge giants on their main battlefield; find niche points that hardware + systems + platforms + big companies are all unwilling to touch" —— The three schools point to the same essence from different dimensions:Scaling big or staying small depends on whether the unit model holds up + whether scale effect can amplify it.

0. The three schools' "unit model / economies of scale"

Perspective	Source	Core
yitang Truman system school	Yitang / Product kernel + complete unit model system (steps 2+3 of the five-step method)	6 levels of unit models + 15 classic ways to mess up + top 10 unit models
Wang Huiwen Tsinghua Product Class school	Yitang / Complete 66-page notes of Wang Huiwen's Tsinghua Product Class	ABC scale effects + network effect + high-frequency beating low-frequency + STP+4P
Wang Kai practical school	Wang Kai's AI-era entrepreneurship methodology and business judgment framework / Wang Kai's AI industry judgment and big-company competitive landscape observations	Hundred-billion-yuan monopoly + multi-technology cost accounting + four big mountains

1. Cross-author consensus (5 points)

1.1 Unit model = the minimum operating unit in a business model

Yitang / Product kernel + complete unit model system (steps 2+3 of the five-step method) Part B:

Definition:A certain minimum operating unit in a business model that reflects the cost-revenue relationship.

Examples	Unit
Cosmetics	Per-SKU model
Food delivery	Per-order model
Nail salon	Per-service model
Gashapon machine	Per-device model
Business district	Per-business-district model
Advertising	Per-customer-acquisition model

Truman quote:"The unit model is the most important ROI model within a company".

1.2 Scale effect = gravity of the business world

Yitang / Complete 66-page notes of Wang Huiwen's Tsinghua Product Class Part 2 Wang Huiwen:

"Scale effect is basically the gravity of the business world. After sales/user volume/customer volume becomes large enough (compared to smaller competitors), it produces (i)user experience advantages and (ii)cost advantages, depending specifically on the business model."

2 levels:

Different businesses inherently have different scale effects —— Some businesses are naturally unable to grow large
Competition within the same business —— Whoever first captures the elements of scale effect wins

1.3 If the unit model does not hold up = losing money

Yitang / Product kernel + complete unit model system (steps 2+3 of the five-step method) §B:Research data - over 50% of money-losing cases are due to incomplete unit models.

3 typical situations(all lead to losing money):

Situation	Manifestation
Too many blind spots	Used to looking at aggregate financial statements, lacking details per SKU / per salesperson / per fulfillment / per business district
Hindsight	Charging ahead aggressively, only discovering the unit fundamentally does not hold up after heavy losses
Static optimism	Rapid expansion after good data from 1-2 stores; opening in 15 cities after breaking even in 1 tested school

Wang Kai corroboration(Wang Kai's AI-era entrepreneurship methodology and business judgment framework §3.3):Single-technology cost accounting —— Full-stack engineer monthly salary 30,000 yuan → a product that can support 1 engineer = monthly revenue of 30,000 yuan / minimum monthly cost for a small team is 80,000 yuan = requires 13,334 monthly paying users.

1.4 "Replicate whichever you count, keep a close eye on whichever is most uncertain" (Truman exclusive)

Yitang / Product kernel + complete unit model system (steps 2+3 of the five-step method) §B Truman core mindset:

"Replicate whichever you count, keep a close eye on whichever is most uncertain!"

3 examples (L3):

Lighting business — find the core SKU unit
Southeast Asia container unicorn miss —— Per container → per SKU → per region → per city, replicating too early and stepping into pitfalls
Car-sharing rental —— 50,000 cars / 36 million users / 4 billion yuan in revenue,per order → per store → per SKU → per city → self-service transformation → per user → per outlet → per business district → per fulfillment 9-stage iteration

1.5 Do not challenge giants on their main battlefield (Wang Kai consensus)

Wang Kai's AI-era entrepreneurship methodology and business judgment framework §5 Four big mountains:

Hardware layer (Apple / OPPO / Tesla)
System layer (iOS / Android / Windows)
Platform / ecosystem layer (Chrome / WeChat / Slack / Notion)
Large model layer (OpenAI / Anthropic / Google)
Big-company application layer (Salesforce / Adobe / Meitu)

Practical advice:Find niche points with special demand that "hardware + systems + platforms + big companies" are all unwilling to touch.

2. Explicit disagreements (2 points)

2.1 Disagreement 1: Which unit to choose for replication granularity?

Truman's top 10 unit models + Wang Huiwen's scale effect classification:

By sales type	By headcount type	By space type
Per order	Per user	Per cabinet
Per SKU	Per customer	Per store
	Per salesperson	Per business district
	Per fulfillment	Per city

Truman's 3 types of misconceptions:

Only calculating big-picture accounts(aggregate financial statements lack details)
Blind spots exist(missing key units)
Choosing the wrong model(granularity too large or too small)

Real disagreement(which granularity to choose):

Lighting wholesale → per order → per customer → per SKU (three-stage iteration)
Southeast Asia containers → per cabinet → per SKU → per region → per city (replicating too early and stepping into pitfalls)
Car-sharing → 9 stages from per order to per fulfillment

2.2 Disagreement 2: Stay small and profitable vs scale big for IPO?

Perspective	Position
Wang Huiwen	"At the beginning we only wanted to build a 1-billion-USD market cap, 60-person company, we didn't expect it to grow big later"
Truman's 7-level spectrum	L1 side hustle → L7 IPO, choose a range not a point / secure A while striving for B
Wang Kai	"Doing business is more important than raising capital" / only place big bets for hundred-billion-yuan monopolies / running a small business with annual profit of several million yuan lets you travel and play Texas Hold'em

Real disagreement:Is there a premium for scaling big?

Wang Kai opposes "scaling big counts as success": small and medium businesses have good cash flow
Truman looks at ranges: L3 standalone hundreds of thousands of yuan vs L7 IPO billion-level
Wang Huiwen in hindsight: scaling big is an outcome, not a goal

3. Detailed explanation of Truman's 6 levels of unit models (exclusive complete version)

Yitang / Product kernel + complete unit model system (steps 2+3 of the five-step method) §B:

Level	Course	Capability
L1	Cognition module	Know the concept
L2	Cognition module	Distinguish 15 classic ways to mess up
L3 Select core model	Practice 1	Find the core unit + replication granularity
L4 Build complete model	Practice 2	Find all key items, no omissions
L5 Pin down baseline values	Practice 3	Plug in specific data
L6 Dynamic prediction model	Practice 4	Dynamic tug-of-war between economies of scale / diseconomies of scale

3.1 L6 Economies of scale (2021 Yitang's most complex diagram)

Truman's assessment:"As the conclusion of the 'Unit Model' course series, this lesson is the brightest pearl on the crown of Yitang's five-step business analysis, worth 100,000 yuan".

12 types of economies of scale (see course for details):

Network effect
Platform effect
Learning curve
Synergy effect
Economies of scope
Data flywheel
...

3.2 Classic case of L3 selecting core model

Complete 9 stages of car-sharing rental(from 0 to unicorn, 50,000 cars / 36 million users / 4 billion yuan in revenue):

单订单 → 单门店 → 单 SKU → 单城市
       ↓
   自助转型(模式升级)
       ↓
单用户 → 单网点 → 单商圈 → 单履约

Why 9 stages: the most critical minimum unit differs at each stage → "Replicate whichever you count, keep a close eye on whichever is most uncertain".

3.3 Student Chen's three-stage iteration in the lighting business

阶段 1: 单订单(早期不知道卖什么)
   ↓
阶段 2: 单客户(批发业务,大客户驱动)
   ↓
阶段 3: 单 SKU(找到核心爆款,组合定价)

4. Detailed explanation of Wang Huiwen's scale effects (exclusive)

Yitang / Complete 66-page notes of Wang Huiwen's Tsinghua Product Class Part 2:

4.1 Three shapes of scale effect curves

Type	Shape	Example
Type A	Monotonically increasing, no ceiling	Platform / network effect
Type B	Monotonically increasing, with ceiling	General SaaS / tools
Type C	Increases first then decreases	Traditional services / labor-intensive

4.2 High-frequency beating low-frequency (classic mindset)

When to use: When pursuing platform strategy / cross-category expansion, first build user base with high-frequency categories, then use high-frequency to drive low-frequency.

Meituan path:Group buying (high-frequency) → food delivery (higher frequency) → hotel & travel (low-frequency but high ticket size) → mobility (medium-frequency).

4.3 STP + 4P + PMF (classic framework)

STP:Segmentation → Targeting → Positioning
4P:Product / Price / Place / Promotion
PMF:Product-Market Fit

4.4 Four major traits of demand + supply-demand relationship

Yitang / Complete 66-page notes of Wang Huiwen's Tsinghua Product Class Part 2 high-level distillation:

Demand intensity(is it rigid)
Demand frequency(is it high)
Demand scale(is it large)
Demand willingness to pay(are users willing to pay)

4.5 Entry timing + Porter's five forces

Entry timing: Early stage high user education cost / mid-stage red ocean / late stage no opportunity
Porter's 5 forces: substitutes / new entrants / buyer bargaining power / supplier bargaining power / intra-industry rivalry

5. Wang Kai practice (exclusive)

Wang Kai's AI-era entrepreneurship methodology and business judgment framework + Wang Kai's AI industry judgment and big-company competitive landscape observations:

5.1 Three things for accounting

Traffic acquisition cost
Monetization efficiency
Long-term operation and maintenance cost

5.2 ARPU vs instant payment model

Model	Applicable to	Focus
ARPU(calculate total value over an extended period)	Long-retention products; even 500 yuan in total value over 3 years is worth doing	Retention / renewal
Instant payment	high-intent demand (writing papers / AI PPT)	Traffic arbitrage opportunity / single-transaction payback

5.3 "Three-step approach" for AI application layer

Step 1: 用 AI 红利切入(低门槛 + 快验证)
  ↓
Step 2: 获取流量 / 规模优势(趁红利期攒护城河)
  ↓
Step 3: 延伸业务壁垒(让 AI 占比降到 30% 以下,构建非 AI 护城河)

5.4 Evolution of "commercial monopoly" in the AI era

Internet era	AI era
Platform monopoly (Taobao/WeChat/Douyin)	AI Agent platform monopoly(ChatGPT/Claude)
Page Rank determines information distribution	Agent Rank determines Agent invocation frequency
SaaS intermediary-type monopoly (Salesforce)	End of intermediary-type monopoly, AI directly connects supply and demand
Content sites rely on search traffic	Content site weight transfers to AI training data

6. How to choose — unit model paths by identity

Who you are	Primary method	Starting point
Launching a business	Truman L3 select core model	Top 10 unit models + replicate whichever you count
Finding reasons for business losses	Truman L4 + 15 classic ways to mess up	Find all key items + see which item is missing
Whether to replicate and expand	Truman L5 pin down baseline values	A single store making money ≠ 50 stores making money / testing 1 school ≠ 15 cities
Want to pursue platform strategy	Wang Huiwen scale effect + high-frequency beating low-frequency	A/B/C three shapes + Meituan path
Small-team AI entrepreneurship	Wang Kai three accounting things + four big mountains	Do not challenge giants on their main battlefield + three-step approach for AI application layer
Want to scale big	Truman 7-level spectrum + Wang Huiwen scale effect	Secure A while striving for B + scale effect is the gravity of business

Universal disciplines:

If the unit model does not hold up = losing money(50% of money-losing cases are due to this)
Replicate whichever you count, keep a close eye on whichever is most uncertain(Truman)
Do not challenge giants on their main battlefield(Wang Kai)
Scale effect is the gravity of business(Wang Huiwen)
High-frequency beating low-frequency(Meituan gene)
Account before acting(Wang Kai's three things)

7. Reference list

Main original author wiki:

Yitang / Product kernel + complete unit model system (steps 2+3 of the five-step method) —— 6 Levels of Unit Models / 15 Classic Ways to Mess Up / Top 10 Unit Models / L6 Economies of Scale (Most Complex Diagram from a Yitang Class in 2021) / Excel Template
Yitang / Wang Huiwen's Tsinghua Product Class: Complete 66-Page Notes —— Wang Huiwen 66 Pages / ABCs of Scale Effects / Network Effects / High-Frequency Beating Low-Frequency / STP+4P+PMF
Yitang / Complete Five-Step Method System (Key Assumptions + Business Milestones + 5 Advanced Stages of the Five-Step Method + Implementation) —— Steps 2+3 of the Five-Step Method / Key Assumptions + Business Milestones
Wang Kai: Startup Methodology and Business Judgment Framework in the AI Era —— 6 Core Mindsets / Three Things to Calculate When Doing the Math / Four Great Mountains / Three Steps for the AI Application Layer
Wang Kai: Observations on AI Industry Judgment and the Competitive Landscape of Large Companies —— Hundred-Billion-Yuan Monopolies / Three Layers of Investment Framework / The End of Intermediary-Type Monopolies

History

2026-05-18: First release of the Topic-dimension wiki. Synthesized and rewritten from 5 author wikis. Explicitly notes 2 points of disagreement: replication granularity / making small profits vs scaling up for an IPO

2026-09 Supplement: What the free subdomain case can prove, and what it cannot prove

The case comes from DigitalPlat FreeDomain analysis notes saved on 2026-05-31. This section is used to practice breaking down unit models, and does not constitute verification of the service's current scale, operating status, or sustainability.

1. First identify what exactly is being delivered

The model described in the notes is: the operator holds the domain name and allocates subdomains under it to users for free. The usable address a user receives is not the same level of control as independently holding a registered domain name. For low-threshold needs such as practice projects and demo sites, this may be attractive; for businesses with high requirements for long-term address stability, the control rights and conditions for continued operation need to be examined separately.

"Holding the resource first, then splitting usage rights" can amortize part of the shared costs, but it cannot be deduced that "total costs are incurred only once" or that "adding more users costs almost nothing". The original notes do not provide a complete cost table, so this should be treated as a model to be verified, not an already established profitable case.

2. Free strategies require breaking down at least four sets of accounts

Accounts to calculate	Questions this case can prompt	Evidentiary boundaries of the current materials
Shared resource costs	Which costs can be spread across more users?	Explains the subdomain allocation idea, but does not provide complete cost details
Incremental service costs	How much additional service and maintenance burden does one more user add?	No actual measurement has been done; it cannot be written as zero
Assets obtained	What does free bring in exchange: users, spread, or business opportunities?	The notes mention claims about user counts and star counts, but provide no evidence of revenue or conversion
Continued use and exit	Why do users stay; what is lost if the service is terminated?	Control rights and dependency risks are worth checking; one cannot assert solely on the basis of it being free that users will definitely not stay

Model check formula:Average cost per user = Allocable shared costs ÷ Number of active users + Total variable costs ÷ Number of active users. A larger denominator only reduces the first term; omitting the second term is a blind spot in the unit model. In actual calculations, the two terms must use the same period and a consistent user definition.

3. When it is worth learning from

When you already have reusable resources and there is real demand blocked by a paywall, you can try using a free version to validate demand and spread. But you must set separate validation questions: whether anyone uses it continuously, how many resources are needed for each additional active user, and whether free users can form a clear revenue path.

Figures in the original notes such as "500,000+ users, 170,000+ GitHub stars, zero revenue" are retained in the original file only as unverified claims from the notes at the time; this section does not upgrade them to current operating facts. Scale, popularity, and profit are three separate questions, and evidence should be sought for each separately.