Do Things that Don't Scale

Startups don't take off by themselves. Founders have to make them take off with an early, laborious, unscalable push — the crank before the engine runs on its own.

Recruit

Manual user acquisition is the most common form. Stripe's "Collison installation": when anyone agreed to try Stripe, the founders said "Right then, give me your laptop" and set them up immediately — rather than sending a link and waiting.

The power of compound growth means small early numbers are misleading. 100 users at 10% weekly growth becomes 14,000 in a year and 2 million in two. The absolute number is not the thing to optimize; the growth rate is. And you can start manual, then switch.

Fragile

Almost all startups are initially fragile. The biggest mistake founders, investors, reporters, and critics make is to judge larval startups by the standards of established ones — like looking at a newborn baby and concluding it couldn't possibly accomplish anything.

Airbnb looked like nothing. About 30 days of going door-to-door in New York made the difference between survival and failure. Even Bill Gates returned to Harvard for the fall semester after starting Microsoft — he didn't realize Microsoft would become what it became.

The right question about an early startup is not "is this taking over the world?" but "how big could this get if the founders did the right things?" The right things often seem both laborious and inconsequential at the time. That's the point.

Delight

Go to extraordinary lengths to make early users happy. Wufoo sent hand-written thank-you notes to every new user for as long as they could. Not as a marketing trick — as a commitment.

Being small gives you an ability no big company has: you can provide a level of service that is physically impossible at scale. Tim Cook cannot send you a hand-written note after you buy a laptop. You can. "Insanely great" in a larval startup means the experience of being your user, not yet the polish of the product. The product can be incomplete and buggy; the experience cannot.

The feedback from over-engaging with early users is also the best you will ever get. Once you're using focus groups, you'll wish you could go to users' homes and watch them use your product — which you can still do now.

Fire (contained)

A narrowly targeted early market works like a contained fire — you get it really hot in one place before adding more logs.

Facebook launched at Harvard only, then specific colleges. The narrow scope meant students felt it was genuinely for them, which drove a critical mass to sign up. Starting broad often means starting nowhere.

The question to ask: is there a subset of the market where we can get a critical mass of users quickly? For marketplaces especially, you almost always have to start in a subset.

Vector framing

Startup ideas should be thought of as vectors, not scalars: (1) what you're going to build, plus (2) the unscalable things you'll do to get the company going. Most founders only think about component 1. The second component is equally important — and often equally differentiating.

An idea where the second component is empty (no unscalable way to get started, no users you can recruit manually) is probably a bad idea, at least for those founders.

The information angle

The reading list specifically highlights this essay for "activities that provide maximal information." The early manual work is not just a necessary evil — it generates feedback that is qualitatively better than anything you can get later. You learn things from building hardware yourself, consulting for one user, or manually doing what your software will eventually do that you could never learn from aggregated metrics. The unscalable phase is an information-gathering phase.