Why "Do Things That Don't Scale" Isn't About Inefficiency
Paul Graham's 2013 essay changed how YC thinks about early-stage validation. The core idea: the fastest path to product-market fit is doing things manually that will eventually be automated. Not as a stepping stone. As a learning tool.
Most founders misread this. They think: "I'll do manual work temporarily, then hire people to scale it." Wrong. The manual work is the validation. Once you automate without understanding customer needs, you've built the wrong machine.
Airbnb's founders photographed 10,000+ apartments themselves. Not to save money. To understand what made listings convert. They saw patterns in lighting, framing, cleanliness. This became their design philosophy. Competitors hired photographers later and missed the psychology entirely.
The Three-Stage Framework
Stage 1: Concierge MVP (First 10-50 users)
Build nothing. Deliver value manually. Use Google Sheets, email, WhatsApp, direct calls. For a B2B SaaS founder in India, this means: pick 10 potential customers, solve their problem yourself without software. Do this every week for a month.
Example: a founder building logistics software should spend 20 hours a week literally managing deliveries, calling customers, tracking orders manually. Watch where your process breaks. Those breaks are your first features.
This stage kills bad ideas in 2 weeks. It saves 6 months of engineering time.
Stage 2: Manual Onboarding (Users 50-500)
Once you have a basic product, onboard each user personally. Call them. Walk them through setup. Watch them use it. Stay on the call while they struggle.
Rippling's Dennis Gartman onboarded the first 200 HR customers himself. He heard the same objection 47 times. That objection became a feature in v2. The next 100 customers onboarded in half the time.
Most Indian SaaS founders hire a support team at 20 users. They miss the feedback loop. They never learn why people churn.
Stage 3: Instrumented Automation (Users 500+)
Only now do you automate. But you automate based on patterns you've discovered manually. Your onboarding flow, your email sequences, your product UI—all informed by real data.
This is when you hire ops, sales, support. But you know exactly what to ask them to do because you've done it yourself.
Why Indian Founders Get This Wrong
There's cultural pressure to "scale fast." Investors ask: "How many users will you have in Year 1?" They don't ask: "How well do you understand each user?" This flips the priority backwards.
Another trap: hiring too early. A founder raises ₹50 lakhs, immediately hires 3 BDRs, 2 engineers. Now they're managing people instead of talking to customers. Their 30 users aren't validated; they're just noise. The company runs out of runway defending a broken hypothesis.
Scott Belsky calls this "the messy middle"—the phase where you're not validated but also not efficient. Indian founders hate the messy middle. They want to jump from idea to "scale." That's where the mistakes live.
A Concrete Indian Example
A food delivery aggregator founder in Bangalore decided to validate demand manually. Instead of building an app, he:
1. Partnered with 5 restaurants directly (no platform).
2. Took orders on WhatsApp himself.
3. Coordinated deliveries via spreadsheet.
4. Delivered 10 orders personally the first week.
Two insights emerged:
- Restaurants didn't want another platform; they wanted delivery coordination. The MVP should optimize for restaurant workflow, not consumer UX.
- Delivery boys weren't the bottleneck; restaurant order preparation was. The real feature should be kitchen dashboards.
He built the right product because he understood the system manually first. Competitors built slick consumer apps without talking to restaurants. They failed.
The Non-Obvious Advantage
Here's what nobody discusses: unscalable work is a competitive moat.
When you spend 50 hours a week talking to customers, you're building a dataset competitors can't access. You're learning local market dynamics. You're building relationships. You're developing taste.
A well-funded competitor can outspend you on marketing and engineering. They can't outspend your customer intelligence. By the time they decide to do manual work, you're already integrated with 200 customers.
This is especially true in India, where market fragmentation is extreme. What works in Bangalore doesn't work in Pune. What resonates with D2C e-commerce doesn't work with SMBs. You must build local knowledge. That requires manual work.
The Actionable Framework
This week:
1. Pick one specific customer segment (25 people max).
2. Solve their problem manually for 30 days.
3. Document every friction point—every email sent, every call, every workaround.
4. Count the hours. Track the LTV. Measure the churn.
5. Only then build software around what you've learned.
If you can't sustain manual operation with 10 users, your business model is broken. Software won't fix a broken model. It just makes it expensive to break.
The founders who become household names do unscalable things at the right time. They embrace the friction. They learn from it. They build the machine after they've learned to dance.