Startup Planning: A Practical Guide - From Problem Definition to the Business Model Canvas

The Starting Point of Business Planning

• What business planning is: The capability to develop research and analysis skills, and to classify and analyze collected data according to the planning purpose, then document it in numbers and charts
• References: 『The Art of Service Planning』(co-author), 『AI Business Theory』·『101 Business Model Stories』(co-author)
• Final goal: Complete 1 AI-based business plan

The core is these three:

• What business planning is: It's not about "what to make" but about "whose what problem to solve"
• The difference between technology and business: Technology asks "can it be made?", business asks "is there someone who will pay?"
• Slogan: "Good services don't start with ideas - they start with problems"

Finding Problems Instead of Ideas

Why you shouldn't start with ideas

Order	Flow	Result
Idea first	Idea → Build → Find customers	"Who uses this?" - mostly fails
Problem first	Problem → Confirm customers → Build	"When can I get this?" - people are waiting

The conclusion: "Ideas are cheap, problems are valuable." Problems where people are frustrated will find you if you go looking.

Good starting points (3), dangerous starting points (3)

• Good starting points: Something I've experienced (most reliable) · Others' complaints (a 1-star review = a signal of a problem) · Something I'm already struggling with (Excel, manual work = an opportunity)
• Dangerous starting points: "Because AI is trendy" (customers disappear) · "It doesn't exist yet" (there's a reason it doesn't exist) · "Everyone can use it" (no one actually uses it)

Frustration shows up in behavior, not just words

1. Just tolerating it - familiar frustration is the biggest opportunity
2. Working around it in weird ways (Excel, KakaoTalk "send to self") - workarounds are evidence of a problem
3. Spending more money or time - a signal of willingness to pay
4. Giving up entirely - the biggest market sits where people gave up

Three discovery questions: "Is this supposed to take this long?" · "Why are you still doing this by hand?" · "When was the last time you got angry about this?" TIP - Instead of "What do you need?", ask "When was the last time you were frustrated?"

Three Lenses

Time ("Why does this take so long?") · Effort ("Why isn't someone doing this?") · Emotion ("Where are you feeling anxious and annoyed?") - Services are born where time is being wasted, repetitive labor is AI's domain, and eliminating anxiety makes people pay.

One-sentence problem definition formula

Who + When/In what situation + Because of what + What harm they suffer

Example: "A 3-year solo apartment dweller doesn't know what's left in the fridge on the way home from work, ends up ordering delivery, and throws away ingredients they saved over the weekend" - all four elements are there, so the solution becomes visible.

4 questions to judge if it's a "solvable problem"

Real existence (3+ concrete people) · Happens frequently (weekly or more recommended) · Has a workaround · Can the effect be described in one line as time/money/mind relief. 3 types to avoid: Too big ("solve youth unemployment") · Only I experience it · Already well-solved ("it's hard to catch a taxi")

Customer Understanding and Personas

The customer might not be one person

Role	Definition	Key point
User	The person actually using it with their hands	Experiences the frustration directly
Buyer	The person opening their wallet	Most sensitive to price
Decider	The person approving "let's use this"	Most important in B2B

Example: A children's learning app - parents are the buyer; a company collaboration tool - team leads/executives are the decider; a college group project app - all three are the same person.

4 criteria for segmenting customers - stronger as you go down

WEAK Demographics ("20s female" is weak) · NORMAL Job/role ("1-cafe owner") · STRONG Situation/context ("3 days before closing" = the "when" from earlier) · STRONGEST Behavior habit ("manages with Excel" = evidence they're tolerating it)

4-step funnel to narrow the target

Total market (50 million) → Role (living alone, ~7.5 million) → Situation (after-work solo apartment dweller at 10 PM, hundreds of thousands - the moment they turn it on) → One person (2-year solo apartment dweller Kim Min-su (22) = persona).

"Isn't 1 person too small?" - There are hundreds of thousands of people like Kim Min-su. Satisfy one person well and similar people will follow; aim for everyone and you'll catch nothing. The narrowed targets share a common time constraint ("3 months into fitness," "night before a presentation") - people in a hurry become your first customers.

7-cell persona template

Definition: A fictional character representing one key customer, written as if they're a real person with a name, situation, daily routine, and frustrations. "20-30s male/female smartphone user" type summaries are statistics, not personas.

Cell	Content
01 Name/Age	Must use a fake name
02 Job/Role	Daily routine should be imaginable
03 Daily routine	The activation moment comes from here
04 Frustrations experienced	In this person's own words, from the problem defined earlier
05 How they're tolerating it now	If skipped, the problem is weak
06 Desired outcome	Not features - results
07 One-liner quote	The strongest line for presentations

4 common mistakes: Listing statistics · Creating an ideal type · No visuals · Pre-solving - only write the problem, save the solution for value proposition. Test: "I know someone like this" = success, "Who are you talking about?" = insufficient. 3 principles: Real person · No solutions yet · Only 1 number.

Market and Competitive Service Analysis

SESSION 1 - How to look at the market

Stage	Question	Nature
TAM (Total market)	How big is this entire field?	Credibility
SAM (Target market)	Which range within that does our customer belong to?	Direction - plug in persona conditions
SOM (Serviceable obtainable market)	What can we realistically capture in 1 year?	Planning - most important in planning

TAM is for credibility, SAM is for direction, SOM is for planning - you need all three to be persuasive. Forbidden phrases: "The market is huge" · "Even if we capture just 1%" · "There are no competitors" - all sound like "no research done."

Formula: Number of customers × Amount per person × Number of times. Numbers need sources; if none, label as "estimate." 4 research methods: Public statistics (Statistics Korea etc.) · App reviews (1-2 star reviews are the best data) · Community searches · Try using it directly.

SESSION 2 - 3 types of competitors and 6-item comparison table

Starting point is the persona's "how they're tolerating it now" - if they search YouTube then open a delivery app, those two are competitors.

Type	Definition · Key point
Direct competitor	Same customers, same method, same problem - easy to find, e.g., recipe recommendation app
Indirect competitor	Different method solving the same problem - easy to miss but actually the strongest, e.g., delivery app, meal kit
Just tolerating it	Doing nothing and putting up with it - hardest to beat, where many new entrants come from

Competitors take the customer's time and money - everything. The comparison table needs at least 2 types. 4-step finding: Pull from "how they're tolerating it" → Search using the customer's own words → Deduce 1 direct, 1 indirect, 1 tolerating → Try using them directly.

6 comparison items: Name/type · 1 core feature · Price · What they're good at · What's lacking · Where we win. Item 05 "What's lacking" is the core - our service is born here. Item 04 must also be filled for trust, and the judgment standard is your persona's taste, not your own preference.

SESSION 3 - 3 conditions for differentiation, positioning map, one-sentence formula

"Better" is not a differentiation - everyone says the same thing, there's no standard, and there's no reason to switch.

Condition	Question	Example
Specificity	Can you say it with numbers/situation?	"Convenient" ✕ → "Done in 3 minutes" ○
Customer value	Is it something the persona would care about?	"Latest AI technology" ✕ → "No need to go to the store" ○
Competitive advantage	Are competitors not doing this right now?	Must pass all three to be a differentiation

Positioning map: 2 axes → plot competitors → find empty space. Good axes: price, speed, expertise, preparation burden, usage frequency. Bad axes: good↔bad types. Note: an empty space isn't automatically an opportunity - it might be empty because "nobody wants it," so ask "why is it empty?"

Formula: ① For whom ② In what situation ③ Differently from competitors ④ What we do - numbers make it stronger. 4 common mistakes: No competitors (tolerating is also a competitor) · Market too big (focus on SOM) · Boasting features (translate into benefit) · Putting others down (acknowledge strengths first) - the common thread is not talking about the customer.

Making the Value Proposition

Features vs. value - distinguished by the subject

If the subject is "service" it's a feature; if the subject is "customer" it's value. Customers don't buy features - they buy a different day.

• Feature: "AI recognizes photos and recommends recipes" → Value: "No need to go to the store, dinner is solved and less money is wasted"
• Feature: "Automatically predicts order quantities with data" → Value: "Ordering time disappears, so you can leave work early"

3 shapes of value

Value is the flip side of the lenses we saw earlier. Time (gives it back) · Effort (gives some back) · Emotion (makes it easier) - don't say all three, pick the biggest pain point and emphasize it large.

Value proposition canvas 6 cells

• Our side: Product/service · Pain killer · Gain maker
• Customer side: Job to be done · Pain · Gain

Order: Customer side first (moved from the 7-cell persona) → pair pain with pain killer. 3 check questions: Is every feature paired? · Does it directly address the biggest pain? · Is this pairing better than competitors?

Core one-liner message - 3 formula types

Formula	Form	Example
A Pain removal type	"No longer [pain]"	"Fridge ingredients - don't throw them away anymore."
B Result promise type	"Just [action], then [result]"	"One photo, and 10 minutes later dinner is done."
C Target designation type	"For [who], [value]"	"Fridge savior for people who can't cook - for solo apartment dwellers."

Rules: Around 20 characters · No jargon · Customer as subject. Words to delete: AI-based, smart, platform, innovation, convenient, best. 30-second pitch: Who → what pain → our solution → one-liner finish.

Understanding the Business Model Canvas

Business model = 5 questions

For whom → Customer segments · What → Value proposition · How delivered → Channels/customer relationships · What it costs → Resources/activities/partners/costs · What it earns → Revenue streams. Definition: A way of drawing how a business runs on one page, a tool for assuming, testing, and revising based on evidence - Osterwalder & Pigneur, 『Business Model Generation』(2010).

9 cells and the order to fill them

No.	Cell	Question asked
①	Customer segments	For whom?
②	Value proposition	What do we give?
③	Channels	How do we reach them?
④	Customer relationships	How do we keep them?
⑤	Revenue streams	Who pays?
⑥	Key resources	What must we absolutely have?
⑦	Key activities	What must we absolutely be good at?
⑧	Key partners	With whom?
⑨	Cost structure	Where does the money go?

Fill the right side (customer side) first, left side later. ① comes from the earlier persona, ② from the value proposition one-liner, and ⑤⑨ show "is this a profitable business?"

7 revenue stream types

Asset sales (Coupang) · Usage fees (-) · Subscription (Netflix, Wow) · Rental/leasing (-) · Licensing (-) · Intermediary fees (Baemin, Airbnb, Toss) · Advertising (KakaoTalk, Baemin). Mixing types is common.

6 case examples - user ≠ payer

Service	Who pays	Revenue type
Baedal Minjok	Restaurant owners + member customers	Intermediary fees · Advertising
KakaoTalk (neighborhood)	Residents free, local stores/advertisers	Advertising
Toss	Stores · financial companies · financial customers	Fees · interest
Netflix	Viewer pays directly monthly	Subscription
Airbnb	Host + guest both	Intermediary fees
Coupang	Buying customers + marketplace sellers	Sales · subscription · fees

Key point: Baemin takes money from stores (9.8% fee), Coupang's biggest cost is logistics, and Airbnb's key asset isn't buildings - it's "trust (reviews)."

Concretizing the Business Model

The final step is rewriting one-word cells like "advertising" / "?" / "server costs."

4 questions for revenue streams

"Making money from ads" isn't a revenue structure. Who (the payer) · For what (what value is priced) · How much (price) · How (method) - if any is missing, it's under-specified. Example: "A neighborhood side-dish store pays monthly for the 'ingredient recommendation' spot next to recipes."

Real numbers: Baemin delivery fee 9.8% (2024.7~) · Wow monthly 7,890 won · Airbnb host-only 15.5%. Finding: 2+ revenue streams are standard, numbers change yearly so the canvas is a living document.

Rough 1-year revenue calculation

Payer count × amount per payment × number of payments per year = 1-year revenue. Example (assumed): 20 stores × 50,000 won/month × 12 = 12 million won. If insufficient, add more revenue streams or grow customers.

The left 4 cells are a value chain

② Value proposition → ⑦ Key activities → ⑥ Key resources → ⑧ Key partners → ⑨ Cost structure. When value changes, the left 4 change together - Coupang's logistics becoming the biggest cost because of "fast delivery" value is the example.

Make it or partner for it?

Make directly = key activities (the heart of value, e.g., Coupang operating delivery itself) · Partner = key partners (what others do better, e.g., Airbnb accommodations are hosts, Baemin food is restaurants). Judgment question: "If someone else does this, would customers still choose us?" - If yes, partner.

Cost structure - fixed vs. variable

Basket	Definition	Example
Fixed costs	Money going out every month even with 0 customers	Personnel, office, basic server
Variable costs	Money that grows as customers grow	AI usage fees, payment processing fees

Examples: Coupang logistics centers (fixed) + delivery fees (variable), Baemin system (fixed) + rider/payment fees (variable). Trap: If variable cost per transaction exceeds revenue per transaction, more customers means more loss - check first.

Weak cells - 5 signals

• Revenue stream is one word ("ads") → unpack with the 4 questions
• Partner cell is "?" → "What we can't do" is the partner
• 10+ activities → keep only the heart 3
• Costs are just "server" → minimum 4 fixed + variable
• Value and costs are separate → connect them as a chain

Direct quote from the material - "Finding weak cells isn't failure. Finding them now means you can fix them most cheaply."

The full flow, in one line

Problem discovery → Customer understanding → Market/competition → Value proposition → Canvas → Concretization - this is the sequence that produces a complete business plan.

This is	Its material
Problem definition · "how they're tolerating it"	"Frustration" behavior signals · competitor list
Persona · differentiation	① Segments · value proposition
One-liner message · canvas draft	② Value proposition · concretization

Core formulas and matching at a glance

Starting with the formula clusters:

• Problem definition: Who · When/in what situation · Because of what · What harm
• Market size: Number of customers × amount per person × times
• 1-year revenue: Payer count × amount per payment × times per year
• Differentiation one-liner: For whom · In what situation · Differently from competitors · What we do
• 7 revenue types: Asset sales · Usage fees · Subscription · Rental/leasing · Licensing · Intermediary fees · Advertising
• 4 revenue questions: Who · For what · How much · How
• BMC 9 cells: Customer segments · Value proposition · Channels · Customer relationships · Revenue streams · Key resources · Key activities · Key partners · Cost structure (right side first)

One-line matching summary:

• TAM is credibility, SAM is direction, SOM is planning
• Competitors are direct · indirect · just tolerating - indirect is actually strongest
• "Better" isn't differentiation - specificity · customer value · competitive advantage
• If the subject is service it's a feature; if customer it's value
• User ≠ payer (KakaoTalk: residents free, stores pay for ads)
• Heart is made directly, the rest goes to partners
• If variable cost per transaction exceeds revenue per transaction, more customers means loss

Concept matching self-quiz

Answer in 10 seconds to pass.

Question	Answer
If technology is "can it be made?", then business is?	"Is there someone who will pay?"
What type does "just boiling ramen and eating it" fall under?	Just tolerating it - the type where many new entrants come from
The market stage we can actually capture within 1 year?	SOM
Among the 4 customer criteria, the strongest is?	Behavior habit (how they're tolerating it now)
"One photo, and 10 minutes later dinner is done" is?	B Result promise type
From a recipe app's perspective, a delivery app is?	Indirect competitor
The most important cell in the comparison table?	⑤ What's lacking
Money going out every month even with 0 customers?	Fixed costs

Final checklist

Going down from the top, you can check the entire flow at once.

• Starting point - business planning definition, technology vs. business difference
• Problem finding - 3+3 starting points, 4 behavior signals, 3 lenses, problem definition formula, 4 judgment questions
• Customer/persona - 3 customer faces, 4 criteria, 4-step funnel, 7-cell persona, 4 mistakes
• Market/competition - TAM/SAM/SOM, 4 research methods, 3 competitor types, 6 comparison items, 3 differentiation conditions, map, one-sentence formula
• Value proposition - features vs. value (by subject), 3 value shapes, 6-cell canvas · pairing, 3 one-liner formulas
• Canvas - 5 questions, 9 cells, 7 revenue types, 6 case examples matched
• Concretization - 4 revenue questions, revenue formula, chain, activities vs. partners, fixed/variable, 5 weak-cell signals

In the end, one tip per line with one example each will be enough.