Do Things that Don't Scale — Paul Graham

The common myth is that startups build something and if it's good, users magically appear. In reality, startups take off because the founders make them take off. There is usually a separate, laborious early process — like the crank on an old car engine — to get growth started.

Recruit

The most common unscalable thing is manual user recruitment. You have to go out and get users, especially at the beginning.

Stripe (one of YC's most successful) is famous inside YC for aggressive early acquisition. When someone agreed to try the beta, the Collison brothers would say "Right then, give me your laptop" and set them up on the spot ("Collison installation").

Founders resist for two reasons:

Shyness + laziness (prefer coding to talking to strangers who will mostly reject you).
The numbers look tiny at first. They underestimate compound growth. 100 users + 10% weekly growth = 14,000 after a year, 2 million after two.

You can start manual and later switch to less manual methods once the market exists.

Airbnb is the classic example: founders went door-to-door in New York with rolly bags, recruiting users and improving listings. For a period it was so fragile that ~30 days of in-person engagement made the difference between success and failure.

Early startups are fragile. Inexperienced observers judge them by the standards of mature companies. Even Bill Gates returned to Harvard briefly because he didn't yet see how big Microsoft would become. The right question is not "is this taking over the world now?" but "how big could it get if the founders do the right things?"

Fragile and Initial Measures

Almost all startups are fragile at the start. The heroic early measures (manual recruitment, personal photos, etc.) often look inconsequential or even embarrassing in retrospect, but they were the optimal path.

How to find the first users: solve your own problem (then your peers are the users), or do an untargeted launch and notice which users are most enthusiastic, then seek more like them (Ben Silbermann at Pinterest going to design bloggers).

Delight

You must also take extraordinary measures to make early users happy — not just acquire them.

Wufoo sent every new user a hand-written thank-you note for as long as they could.

Founders (often ex-engineers) resist because:

Customer service isn't part of engineering training.
They assume it won't scale (but in the larval stage you have nothing to lose, and delighting users tends to become part of culture).
Their standards were set by big companies (Tim Cook doesn't send handwritten notes; you can).

Once you realize conventions are not an upper bound, it is pleasant to think how far you can go to delight users. I have never seen a startup lured into a blind alley by trying too hard to make initial users happy.

Steve Jobs' "insanely great" applies in a larval startup not (yet) to the product being robust and elegant, but to the experience of being your user. You can deliver an insanely great experience with an early, buggy product if you make up the difference with attentiveness.

See its strong emphasis in the generalist reading list (Strategy section) for the importance of doing things that give maximal information early: a-reading-list-for-generalists-dylan-bowman.

The practice is captured in do-things-that-dont-scale.