Do Things That Don’t Scale

A foundational principle of early-stage startups: founders must manually do unscalable work — recruiting users one by one, providing extraordinary service, operating in narrow markets — before any scalable growth can take hold.

Origin

Coined by Paul Graham in his July 2013 essay, this became one of Y Combinator’s most commonly given pieces of advice. The insight is that most founders instinctively want to “launch” and have growth happen automatically, but this almost never works.

The Five Categories

1. Manual User Recruitment

Nearly all startups must recruit users by hand at the start. The “Collison installation” (Stripe’s founders literally taking laptops to set people up) is the canonical example. The math: 10% weekly growth = 14,000 users in year one.

2. Extraordinary Delight

Going beyond product quality to create memorable experiences. Wufoo sent handwritten thank-you notes to every user. This approach compounds — it becomes culture.

3. Narrow Focus

Start with a deliberately small market to achieve critical mass. Facebook launched at Harvard only. Pinterest’s founder recruited users at design blogger conferences.

4. Manual Manufacturing

Build hardware by hand before scaling production. Pebble assembled their first watches manually before their $10M Kickstarter.

5. Consulting Mode

Do the work manually before automating. Viaweb built online stores for merchants who couldn’t use the software themselves.

Why It Works

Compound growth from small beginnings is vastly underestimated
Big launches almost never drive sustained growth
Partnerships with large companies rarely work for early traction
Extraordinary early experiences create organic word-of-mouth
Manual work teaches you what to automate

When to Stop (The Transition)

The transition from unscalable to scalable is one of the hardest moments in a startup’s life — covered in depth in scaling. Knowing when to stop doing things manually and start building systems is a judgment call with no clean answer.

Signs it’s time to systematize:

You’re personally the bottleneck — customers are waiting on you, and only you, to deliver
The same manual process is running for the 100th time with no variation
Quality degrades because you simply can’t do it all yourself anymore

Keith Rabois’ barrels framework is useful here: you need to find “barrels” — people who can take your manual process and turn it into a repeatable system. Not everyone can do this. Most people are ammunition; barrels are the ones who can own an entire process end-to-end and ship it without supervision.

The dual risk:

Transitioning too early means losing the magic of personal touch. You systematize before you truly understand what makes the experience special, and you automate away the very thing that was working.
Transitioning too late leads to founder burnout and quality collapse. You can’t personally onboard user #5,000 the way you onboarded user #5.

Airbnb waited 4 months after fundraising before hiring anyone — PG calls this a key strategic choice. They used that time to keep doing things manually, learning deeply about what their users needed before encoding that knowledge into systems and people.

From the Trenches: Real Stories

Jessica Livingston’s account of what goes wrong at startups reveals just how far founders go to make things work before product-market fit kicks in.

Airbnb nearly exhausted their credit cards doing things manually before being accepted into Y Combinator. They were literally going door-to-door in New York, taking professional photos of listings themselves. There was no growth hack — just three founders doing everything by hand until it worked.

Pebble was rejected by every investor they pitched. Rather than giving up, they built their own crowdfunding platform in a single week when Kickstarter wasn’t an option for their timeline. They went on to raise over $10M — the most successful crowdfunding campaign at the time.

Stripe had young founders who knew their age could be a liability in meetings with banks and financial partners. Their workaround: they called banks on the phone before in-person meetings to establish the relationship and prevent age bias from derailing deals before they started.

Lockitron was rejected by Kickstarter entirely. Instead of waiting for approval, they built their own crowdfunding platform in days and raised $2M directly from customers.

These are all examples of founders doing whatever it takes — the most unscalable thing of all. Not just manual user acquisition or handholding, but bending the entire process around whatever obstacle appeared, often with no playbook and no safety net.

Entrepreneurship KB

Explorer

Do Things That Don't Scale