Do Things That Don't Scale
In one line
Early on, stop waiting for customers to find you and win them one at a time, by hand.
The scene
Two hundred dollars a week: that was all Airbnb earned in the winter of 2009, and the number would not move. Brian Chesky, Joe Gebbia and Nathan Blecharczyk were in Y Combinator's batch, short on cash and carrying credit card debt. They had a working site, and people did use it. They just didn't use it much.
Paul Graham, who co-founded Y Combinator, looked at the numbers the way a programmer looks at a bug. Most of their listings were in New York. So why were the three founders sitting in California?
He and his partners told them to go to New York and meet the hosts face to face. The founders went. They visited hosts at home and looked at their listings with their own eyes. The photos hosts had uploaded were dark and blurry, so the founders took better ones themselves with a rented camera. It was slow, and it was embarrassing work for the founders of a technology company. Nobody could have called it a strategy.
Graham wrote later that the revenue figure doubled, from $200 a week to $400. It didn't stop there. When he wrote the essay summarized here, in 2013, that trip was the example he reached for first.
The big idea
Most people picture a startup as a great product that spreads on its own. Graham says it almost never goes that way. A startup takes off because its founders push it, one customer at a time, long before anything spreads on its own.
The trap is that big companies look automatic. They have ads, sales teams and software that handles thousands of customers at once. A new founder copies that shape and waits for the machine to turn on. But the machine needs fuel, and in the beginning the fuel is you.
Doing things by hand works for a simple reason: it teaches you faster than anything else can. When you personally sign up a customer, you hear the hesitation in their voice. When you personally fix their problem, you see what the product is really for. Automation hides all of that. Do the work yourself while the numbers are small. Once you understand it, you will know exactly what is worth building a machine for.
The things that look too small to matter are often what get a company started.
Read it if / Skip it if
- Read it if you are launching a product, a service, a newsletter or a side business and the first customers aren't showing up on their own.
- Skip it if you already have steady, growing demand and a team, and your problem is running the machine rather than starting it.
How it works
1. Go and get your first users yourself. The shift: from "launch and wait" to "go and find them." Graham says nearly every startup has to recruit its first users by hand. Nobody hears about you until you tell them, one conversation at a time. He tells the story of Stripe's founders, who would not let a new user go away to try the product later. They offered to set it up on the person's own computer right then. Graham named this the "Collison installation," after Patrick Collison, one of the co-founders.
2. Treat your first few customers as if they were your only ones. The shift: from seeing support as a cost to seeing it as the thing people remember. Graham argues that a small company can beat a big one by being more attentive than a big company can afford to be. He mentions Wufoo, a web-form startup, whose team sent handwritten notes to its customers. A big company would never do this, which is why the notes got noticed.
3. Do the job manually before you build the software. The shift: from building first to learning first. If you can deliver the result by hand, even clumsily, you find out whether people want it before you spend months automating it. The manual version also shows you which steps are painful, so you automate the right ones.
4. Read weak numbers as information, not as a verdict. The shift: from "this isn't working, so I'm failing" to "this is telling me something." Airbnb's flat $200 a week could have been the end of the story. Instead, Graham and the founders asked a plain question about it: where are the users, and what do they see? The answer pointed to New York and to better photos. The setback changed the plan, not the people's opinion of themselves. That is the whole habit.
Where it breaks
Graham's examples are mostly winners that he knows well from Y Combinator. We don't see the many companies that also did things by hand and still failed, so the essay shows a pattern in successful startups, not proof that the method causes success.
It also fits some businesses better than others. It suits software and marketplaces, where early users have to be recruited one by one before the product becomes useful to anyone else. It fits less well where the first sale already takes months of formal selling, or where heavy investment is needed before the first customer.
Graham admits the risk himself: this work can become a habit you never drop. A founder who spends years answering every email personally has an exhausting business, not a startup. The essay is better at telling you how to begin than when to stop, and that judgment is yours to make. Finally, the idea is empirically testable, which helps: you can try it for a month and count the results.
Do this Monday
Pick one person who has tried your work, product or idea, and send them a personal message today (under ten minutes): thank them, and ask one specific question about what nearly stopped them from using it. "Done" means the message is sent and their name is on a list you keep.
Repeatable practice: once a week, do one thing for one customer that you could never do for a thousand.
The Wall Note
- Go to your users. Don't wait for them.
- Be too attentive on purpose.
- Do it by hand first.
- Learn before you automate.
- Flat numbers are information, not verdicts.
- Start with a few people who love it.
- Scale later, not now.
Send this to someone
Send this to the friend who has been polishing a side project for months and is waiting for people to discover it. Because the essay's main point is that waiting is the wrong move.
"Read Paul Graham's 'Do Things That Don't Scale.' Then pick three people and personally get them to try it this week. I'll check in Friday."
Sources
- Paul Graham, "Do Things That Don't Scale" — the original essay (July 2013), including the Airbnb, Stripe and Wufoo examples.
- Paul Graham, "Startup = Growth" — a second essay by the same author on why early growth rate is the number that matters.
- Paul Graham (programmer) — Wikipedia — an independent overview of his career, Viaweb and Y Combinator.
Get the full book
The essay is free at paulgraham.com, and for more of the author's thinking, pick up Paul Graham's Hackers & Painters (O'Reilly Media) — amazon.com, at bookshops, or your local library.
This is an original editorial commentary created for personal inspiration. All ideas, frameworks, proprietary concept names, and registered trademarks belong to their respective authors and publishers — this site is not affiliated with, sponsored by, or endorsed by the author or publisher. No sentences or passages from the original book are reproduced verbatim. This summary is not a substitute for the original work. We strongly encourage you to read the full book.
Rate 1-5 when you like.