Do things that don’t scale — to understand customers and create "wow"
🎯 Goal: Understand why, early on, you should hand-hold every customer and do things manually — even if it "doesn’t scale" — to understand customers deeply, create "wow" moments, and win a few who truly love you.
Paul Graham’s famous advice to young startups: "Do Things That Don’t Scale". Early on, instead of dreaming of automating everything, serve each customer by hand, do things manually, meet people in person. That lets you understand customers in ways automation never teaches, and create "wow" moments customers remember and retell. At this stage you don’t need many customers — you need a few who truly love you, because they are the base of all later growth.
Lesson content
🎯
The paradox: to build something that scales later, you first do things that DON’T scale. The deep reason: doing it by hand puts you right next to real pain and you learn incredibly fast — something reports and aggregate numbers can’t give. Paul Graham (in "Do Things That Don’t Scale") argues a startup is like a cold engine you must hand-crank with manual work before it runs on its own. This isn’t wasted time — it’s how you learn and how you build initial momentum.
🧭
4 "unscalable" moves for the early stage: 1) Recruit customers yourself — don’t wait; proactively meet/message the first ones. 2) Delight (over-serve) — "wow" customers with care that big companies can’t match. 3) Onboard by hand — walk each user through the first step, see where they stumble, fix the product. 4) Run the "backstage" manually (by hand, not code) — test ideas fast before building systems. All four share one goal: learn fast + create your first loyal customers.
🧰
TOOL — A list of "unscalable" things worth doing early (and what each teaches/creates):
Do it by hand
What it teaches / creates
Meet/call your first 10 customers yourself
Learn their real language and pain
Onboard each new user by hand
See where they stumble, fix the product
Text a thank-you / check-in after purchase
Create "wow" and word of mouth
Handle returns / complaints personally
Turn angry customers into loyal ones
Run the "backstage" by hand, not code
Test fast before building systems
Pick 3–5 and do them with real people — each time is a free lesson.
🌏
Case study — Airbnb & Vietnamese shops (step by step): 1) Early Airbnb: listing photos were so bad that few people booked. 2) The founders personally went to photograph homes for hosts in New York — impossible for millions of listings. 3) Result: better photos → bookings rose; and they saw firsthand why hosts hesitated to post. 4) Vietnamese shops grow by texting each customer by hand: asking sizes, advising outfits, thanking after purchase, handling returns via personal messages. 5) Early Coolmate was also obsessed with experience: up to 60-day returns, fast replies — "manual" details that make customers feel cared for and build loyalty.
⚠️
4 traps around "not scaling": • Automating too early: building systems at just 20 customers — optimizing a model that may be wrong. • Chasing an indifferent crowd instead of a few who truly love the product. • Doing it by hand without extracting lessons: serving customers but not recording what you learned wastes the effort. • Never letting go: manual work is to LEARN and start the engine; once the model is proven you must gradually standardize — don’t stay stuck doing everything by hand.
✅
Early "unscalable" checklist: ① Did I proactively meet/message each first customer? ② Did I do one thing that "wows" customers that a big company can’t? ③ Did I onboard by hand and watch where they stumble? ④ Did I record one surprising thing learned from each customer? ⑤ Do I have a few customers willing to retell/refer?
Practice exercise
🔬 APPLIED EXERCISE: Pick your first 5 (real) customers/users. For each, do one "unscalable" thing to create a "wow" (a check-in call, hand-deliver, a personal thank-you note, hands-on onboarding...). Write down one surprising thing you learned from each person that you didn’t know before, and one product change you’ll make because of it.
Worked template: FILLED EXAMPLE — selling cakes online: hand-deliver the first 5 orders. Surprise: customers buy them as GIFTS, not to eat → add gift-wrap + a handwritten card, matching the real need. A "manual" detail you only learn by delivering yourself.
Quick quiz
1. What does Paul Graham’s "Do Things That Don’t Scale" mean?
→ Early on, accept doing things manually, serving each customer to understand them and create "wow"
"Do Things That Don’t Scale" (Paul Graham) = early on, accept manual work, serving each customer to understand them and create "wow".
2. Why was Airbnb photographing each home valuable even though it "doesn’t scale"?
→ Better photos raised bookings, and they directly learned why hosts hesitated to post
Airbnb photographing homes raised bookings and gave direct insight into why hosts hesitated — the fastest way to learn.
3. At the first-100-customers stage, what matters most?
→ Having a few who TRULY love the product and will retell it
At the first 100 customers, what matters is having a few who truly love the product and retell it, over many who are indifferent.
4. What’s the advantage of running the "backstage" by hand instead of coding right away?
→ Test quickly before spending effort building systems
Running the "backstage" by hand lets you test fast before spending effort building systems.
5. Why does "doing things that don’t scale" NOT contradict the ambition to build a big company?
→ Because it’s how you learn fast and create your first loyal customers — the base to build something scalable later
Manual work doesn’t contradict big ambition: it’s how you learn fast and build a loyal base — the foundation to scale later.
6. After doing manual work with customers, what is the key thing to take away?
→ Record the surprising thing learned and turn it into a product change
After hand-caring, you must record what you learned and turn it into a product change, else the effort is wasted.
Advanced
A deeper framework
At an advanced level, "doing things that don’t scale" is a learning strategy, not a desperate act. Each time you serve a customer by hand you gather qualitative data (what they say, where they stumble, what makes them "wow") — things a spreadsheet never gives. That data is precisely what tells you what system to build later.
Timing principle: do things by hand to find and prove the model; once the model is clear and starts to overload, gradually standardize and automate the right parts. Automating too early is "optimizing something not yet proven" — fast, but it multiplies mistakes.
What hand-caring for 20 users teaches you
First 20 users, message each by hand
~6 hours/week of manual work
Finding: 14/20 stumble at the same onboarding step
Fix one step → activation clearly rises
5/20 use the product in a way you didn’t expect
Opens up one new feature matching real need
3/20 proactively refer friends
Your first loyal customer base
6 "unscalable" hours/week buys 3 costly lessons — exactly what guides what to automate later.
Common trap: Automating too early: spending 2 months building a care system when you only have 20 customers. You optimize an unproven model and lose the chance to learn from direct contact — a cost far higher than the time you "saved".
Advanced questions
1. Your team has its first 20 users and is debating: (A) spend 2 months building an automated care system, or (B) personally text and meet each of those 20. Which and why?
→ (B), because at small scale, hand-care builds deep customer understanding and "wow" — things automation can’t teach
At 20 users, automating is premature optimization. Hand-caring for each teaches real pain and creates customers who love the product — that qualitative data tells you what system to build later.
2. Why does "doing things that don’t scale" NOT contradict the ambition to build a big company?
→ Because it’s how you learn fast and create your first loyal customers — the base to build something scalable later
Manual early work is precisely how you LEARN and get initial momentum; that insight and loyal base is the foundation to later build the right scalable systems.
3. When should you START standardizing/automating instead of doing everything by hand?
→ Once the model is clear, customers stay, and manual work starts to overload
Do things by hand to find and prove the model; once it’s clear and overloading, standardize the right parts. Too early optimizes the unproven; too late gets you stuck and unable to grow.
🎯 Real-life mission
REAL-LIFE MISSION: For your idea or small project, pick 3 "unscalable" actions and actually do them with at least 3 real people (a check-in call, hand-deliver, a personal thank-you, hands-on onboarding...). Record: one surprising thing each person taught you, and one product change you’ll make because of it.