Do Things that Don't Scale

Do Things that Don't Scale. An essay by Paul Graham, July 2013.

Startups need a push

One of Y Combinator's most common tips: do things that don't scale. Why does that need saying?

Many founders think a good product draws users by itself. And if nobody comes, there was no market.

Graham disagrees: “startups take off because the founders make them take off.” Very few grow on their own. Most need a push.

Like cars before electric starters: someone had to crank the engine by hand. Hard work, but once it caught, it ran on its own.

Recruit users by hand

Users won’t come on their own. So the most common unscalable job is recruiting them by hand.

Even Stripe, with an urgent problem to solve, went hunting for users. A timid founder asks if you’d try the beta, then promises to send a link.

The Collison brothers asked for your laptop and set you up right there. At Y Combinator, that’s called a Collison installation.

Why founders hold back

Why do founders resist? Partly shyness and laziness: coding at home beats hearing no from strangers. But someone, usually the CEO, has to sell.

And the numbers look tiny. But growth compounds: 100 users growing 10% a week become 14,000 in a year, and 2 million in two.

You won’t do it by hand forever: as you grow, you shift to less manual ways. But the start is by hand.

Airbnb, door to door

Airbnb is the classic case. Marketplaces are so hard to start that heroic effort is normal, so its founders went knocking on doors in New York.

They signed up hosts and improved their listings, even taking the photos. Graham pictures them always towing rolling suitcases.

Airbnb looks unstoppable now. But early on, about a month of meeting users in person decided whether it lived or died.

Fragile at first

Nearly every startup starts out fragile. Yet people judge them by grown-up standards, like writing off a newborn because it can’t do much yet.

The real danger: dismissing it yourself. After starting Microsoft, even Bill Gates went back to Harvard. He’d never have gone had he guessed how big it would get.

Ask instead how big it could grow if the founders did what’s needed. That often looks tedious and trivial: two guys in Albuquerque writing Basic for hobbyists.

Finding the first users

Who should you recruit? If you built it to fix your own problem, just find people like you. That’s usually easy.

If not, launch loosely, see who’s keenest, and find more like them. Pinterest’s early users loved design, so Ben Silbermann went to a design bloggers’ conference.

Delight your users

Don’t just get users; delight them. Wufoo sent every new user a handwritten thank-you note, and kept doing it surprisingly long.

Why does this feel unnatural? Founders are often engineers, trained to build things, not pamper users. But a founder is the captain now: Kirk, not Scotty.

Worried it won’t scale? Early on, there’s nothing to lose. Having too many users to pamper would be a lovely problem.

And most founders have never been treated this way. Big companies set the bar. Tim Cook can’t send you a note when you buy a laptop. You can.

An insanely great experience

How far should you go? Steve Jobs had the phrase: insanely great. At an established Apple, that meant the Mac itself, down to its packaging.

In a new startup, what must be insanely great is what it’s like to be your user. Attention can make up for a buggy, unfinished product.

So ship once it’s barely useful and watch what users do. Your first version is never right, and this early feedback is the best you’ll get.

A contained fire

Sometimes the trick is a market narrowed on purpose. Like a fire: get a small one burning really hot, then add the logs.

Facebook began only at Harvard: a few thousand possible users. But it felt made for them, so a critical mass signed up.

Then it spread one college at a time. Zuckerberg said building each school’s course lists was hard work, but made students feel at home.

Pulling a Meraki

Hardware has its own version, called pulling a Meraki. Meraki’s founders got started by building their routers themselves.

Why? Factories want huge minimum orders, often hundreds of thousands of dollars. No product, no growth; no growth, no money to make the product.

Pebble hand-built its first few hundred watches, which probably made its $10 million Kickstarter possible. And it taught them things, like the value of good screws.

Build for one user

For business software, try picking a single user and acting as their consultants. Fit that user perfectly, and others usually want it too.

Consulting is classic unscalable work. It’s safe as long as they aren’t paying for it. Paid by the hour, they’d expect everything.

At Viaweb, Graham’s team built merchants’ online stores for them. It felt lame, but it showed them how the software felt to use.

Be the software

A step further: be the software. With few users, do by hand what you’ll automate later. You launch sooner, and learn what to build.

It can feel like a practical joke. Stripe’s instant merchant accounts? Behind the scenes, the founders were signing users up by hand for ordinary ones.

Some startups could start entirely by hand. Scary? Less scary than the usual case: an automatic product that solves nobody’s problem.

No big launch

What usually fails? The big launch. Some founders treat startups as projectiles, not powered aircraft, as if launch speed decided everything.

How many famous startups’ launches can you recall? A launch only needs to bring a core of early users. How happy they are matters more than how many.

Big partnerships rarely work either. Any plan that skips the hard work at the start, a big launch or a big partner, deserves suspicion.

Ideas as vectors

Graham suggests treating a startup idea as a vector, not a single number: what you’ll build, plus the unscalable thing you’ll do to get going.

Usually that second part is the same: recruit users by hand, and make their experience overwhelmingly good. You have to work in two dimensions.

At best, that early effort becomes the company’s DNA. Work hard to delight your first few users, and the habit stays as you grow.

In short

Startups take off because founders push them.

Recruit users by hand, one at a time, and delight them.

Make being your user insanely great, even while the product is buggy.

Every idea has two parts: what you build, and the unscalable work that gets it going.

The original is full of examples, and better than this. Read it: paulgraham.com/ds.html