Paul Graham
Do things that do not scale
The early work is meant to be handmade
Do Things That Don't Scale, 2013
Founders think about scale constantly. He argues for the opposite direction — and is specific about what the handmade phase actually buys.
What actually happened?
The essay opens against the instinct: startups do not take off by themselves; the founders make them take off. The common mistake is chasing something scalable on day one, when the early period calls for recruiting users by hand. Airbnb had stalled, so the founders flew to New York week after week, knocked on hosts' doors and photographed their apartments themselves. Completely unscalable — and it found the switch that moved conversion. The output of that phase is not the handful of users; it is an unusually concentrated understanding of them, plus people so over-served they will tell others.
Doing it by hand is the highest-bandwidth research there is
Surveys and dashboards show the shadow a behaviour casts; being in the room shows the behaviour. In people's living rooms it turned out photograph quality was the bottleneck — an insight the analytics could only ever have shown as low conversion. The expensive thing early on is not slow growth, it is understanding wrongly.
Have the founders deliver the first hundred customers personally. Every one is a piece of research arriving with its full context. Time saved by delegating it comes back doubled on the wrong roadmap.
Over-serving is how you manufacture the first advocates
If early users get service indistinguishable from a large company's, they have no reason to speak about you. The handmade phase gives what no scaled operation could — a founder replying in minutes, the product bent around one person. That unreasonable care is the raw material of word of mouth.
Answering the first cohort at any cost brings customers later that are cheaper than any advertising, because the story they tell cannot be bought.
Unscalable is a stage, not a creed
The essay gets quoted in halves. The correct loop is: do it by hand, extract the one move that was actually doing the work, then build that move into the product or the process. Airbnb turned professional photography into a platform feature. Handwork without extraction is diligence covering for an absence of leverage.
After three months of founders closing deals personally, the deliverable is a list of the three actions that decide a sale — then those get standardised. Not a fourth month of closing deals.
How do I use it today?
Where you are: the product just launched and you are torn between building growth machinery and serving users by hand.
Ask first: is our understanding of why people buy or leave at dashboard level or living-room level?
Where it goes wrong: keeping it handmade past the point of scale, so founders are still carrying the business on their backs; or doing it by hand for a long time and never extracting the repeatable move.
Lines to keep
Startups do not take off by themselves. The founders make them take off.
In the early days you have to recruit users manually.
They went door to door photographing the hosts' apartments themselves.
Same situation, other people are asking
It's live and nothing is happening. Do I market it or go talk to people?We hit the numbers and I can't say who we serve.There's no market for this yet. How long can I fund it?Everyone still using it says they like it. Who am I not asking?I'm building something nobody asked for · all 4 questions →If this one named what you are going through,
send it to someone who needs it, or keep it somewhere you will find it again.