The Save Is Already In The Building

The Core Insight

Charles Geschke raised 2.5 million dollars to build printers, typesetters, and software. Two customers told him to throw the plan away. Gordon Bell of DEC asked why Adobe did not sell him the software alone, and Steve Jobs said the same thing two months later.

About half a dozen companies were funded in the same twelve-to-eighteen-month window to do roughly the same thing. The other five followed the plan they had raised on, and all five disappeared. Adobe incorporated in December 1982, signed Apple in December 1983, and turned profitable inside its first twelve months.

Founders at Work collects thirty-two interviews about the early days of startups that worked. Jessica Livingston published it in 2007. One at a time they read as war stories. Together they read as a survey, and the survey is where the value sits.

Most founders read startup history as one person with a good idea making a sequence of good decisions. These interviews carry a different shape. The idea arrived as a private annoyance, and the first version took days. The plan got thrown out once a customer spoke, and the save was already running somewhere inside the company.

Livingston names determination as the single most important quality in a founder. What the interviews add is where it gets spent.

The Framework

The book carries no framework of its own. The repetition across thirty-two interviews is the framework, and it holds in six parts.

• The idea arrives as a private annoyance and gets recognized as a market later.
• The first version ships in days and does one narrow thing.
• Customers replace the plan that raised the money.
• The near-death moment resolves with something already inside the company.
• The first users get found by hand, on somebody else's surface.
• Cofounder alignment decides more outcomes than the technology does.

The exceptions are few and instructive. Four people left Sun's Java team to start Marimba with no idea at all, betting that a proven team beats a proven idea. Each put in about 25,000 dollars, and the whole office cost about 1,400 dollars to furnish. Arthur van Hoff calls the first idea a catalyst rather than a plan.

Adobe, Alliant, and Lycos began from technology looking for a market. All three needed real capital, which is the price of starting from that end.

Key Ideas

The Idea Arrives As A Private Annoyance

Sabeer Bhatia and Jack Smith built Hotmail because the firewall at their day job blocked their personal email. Joshua Schachter built del.icio.us out of a text file of 20,000 links he had no way to search. Paul Buchheit built Gmail against a college irritation about walking back to his dorm to read mail. Dan Bricklin built VisiCalc after running business school numbers by hand, and Steve Wozniak designed a computer he had no money to buy.

The leap in each case was noticing that other people carried the same problem. Craig Newmark started craigslist as a mailing list of ten to twelve people on Pine. Mark Fletcher states the rule flat: solve a problem that you have, first.

The First Version Is Days Of Work And Narrow On Purpose

Buchheit built the first Gmail in one day on Google Groups code, and it searched only his own mailbox. That was already useful to colleagues, because they shared so much of the same mail. He shipped it with 1 gigabyte of storage against a standard of 2 to 4 megabytes. The launch date was April 1, so people took it for a joke.

Steve Perlman built the WebTV prototype in three days and two nights on about 3,000 dollars of parts from Fry's. Bricklin wrote the VisiCalc prototype in Basic over one weekend in the fall of 1978. Caterina Fake had the first Flickr running eight weeks after the idea. Basecamp launched at 4,000 lines of code, written at ten hours a week by the only programmer, with Rails at 1,000 lines underneath it.

James Hong coded HOT or NOT in spare time by the Wednesday of the week the idea came up. He emailed 40 friends on the Monday and took about 40,000 hits that day.

Customers Replace The Plan You Raised On

Max Levchin changed the business plan six times before PayPal. The chain ran from crypto libraries to enterprise security, then a consumer wallet, then Palm IOUs, then web payments. eBay bought the last one in 2002 for 1.5 billion dollars.

TiVo raised on a home server network. The founders went back to the investors, said they had changed their minds, and narrowed to the one application they knew how to explain. Flickr was a side feature inside a game called Game Neverending, which had 20,000 people signed up for the prototype. The game was shelved in July 2004 with six people left.

Fog Creek started with three consulting clients and had zero by February. Joel Spolsky then sold the bug tracker the two founders already used internally. Geschke's board chair put the rule in one line. Customers tell you what your business is, and the plan existed only to get the money.

The trigger is usually a demand signal the founders resisted. Levchin told the eBay sellers asking for PayPal logos to go away. Evan Williams and Meg Hourihan kept agreeing that Blogger was too small and too trivial to be the company.

The Save Is Already In The Building

Stephen Kaufer ran the wrong TripAdvisor for eighteen months. Licensing the database to portals produced one deal, with Lycos, whose quarterly check did not cover the weekly free lunch. After September 2001 the company cut from eleven people to eight.

The save was a demo site nobody had paid attention to, running about 5,000 visitors a day. A copied Expedia banner drew 100 clicks from 3,000 visitors. Deep links to the exact hotel booking page, priced per lead, ran at 10 percent click-through. The industry rate was a quarter to half a percent.

They gave Expedia a free month. Expedia called before it ended and offered 10,000 dollars for December against about 20,000 leads, then 20,000 dollars for January, then hundreds of thousands. Break-even was about 75,000 dollars a month, and they got there in about four months from no revenue.

The same shape repeats across the book. Spolsky's bug tracker was sitting on his own machine. James Currier's personality tests launched and drew nobody, while a throwaway fifteen-question dog-breed test drew a million people eight days later. The paid Meet system at HOT or NOT started as a way to block porn. At 6 dollars a month it made about half a million dollars a year.

One company in the book did not get the save. Software Arts died of a lawsuit with its own publisher, filed in September 1983. The filing landed one or two days before a cash-and-stock acquisition worth about 50 million dollars was to close. Lotus later bought the assets for a few million to avert bankruptcy.

The First Users Get Found By Hand

Paul Graham gave the first five or six Viaweb customers the software free and built their sites himself. TiVo put all sixty employees into Fry's stores for a weekend. Kaufer hired people to read every travel article on the net, about half an hour each, and classify them by hand. People told him the work had no end.

The cheap distribution rode on somebody else's surface. Hotmail put a signature line on every outbound message and reached 100,000 subscribers in its first three months. PayPal told each new user that 10 dollars was waiting. Levchin rates that line above every other viral driver he saw.

Firefox had no budget at all. Spread Firefox signed up over 250,000 people in 2004, matched by region and skill. Ten thousand of them donated 10 to 30 dollars each, which bought two facing full-page New York Times ads. Blake Ross states the conclusion: marketing is making the product good enough that people spread it, and giving them a way to do that.

Small numbers carried real signal. Flickr had 60,000 users when a bomb hit the Australian embassy in Jakarta, and three people uploaded tagged photos within 24 hours.

Cofounders Decide More Outcomes Than Technology

Levchin and Peter Thiel ran a puzzle-solving showdown as a compatibility test. Levchin notes that neither of them was in a funk at the same time. Wozniak and Jobs never argued except over the number of expansion slots on the Apple II. Bricklin and Bob Frankston argued constantly for twenty-five years without animosity.

The fractures cost more than any technical failure in the book. Williams and Hourihan ended in layoffs, a lawyer, and a campaign among mutual friends. Ann Winblad, from the investor's chair, says most companies fail by self-inflicted wounds from the leadership team rather than by a competitor.

Two structural moves are worth copying. Joe Kraus and his five Excite cofounders moved equity away from six equal sixths at Vinod Khosla's prompting. Kraus says the change survived only because they were friends first. Tim Brady agreed the terms of his own exit with Jerry Yang before joining Yahoo, so the question never occupied him afterward.

Spolsky sets the floor at one other person. Do not start a company unless you can convince somebody to quit and come with you.

Practical Applications

Write down the annoyance you already solved for yourself with a script, a spreadsheet, or a text file. That artifact is the most reliable starting point in the whole book.

Cap the first build in days, and narrow the scope until it fits. A version that serves only you is the version that ships this week.

List everything already running that you have written off as a side project, an internal tool, or a demo. Kaufer's demo site was carrying 5,000 visitors a day while the company sold licences nobody wanted. Read what that traffic already does before you build anything new.

Get the first ten users by hand and do the work that does not scale. Build their sites, staff the store, classify the articles yourself. Then put one mechanism inside the product that carries it outward to the next user.

Settle the cofounder terms before the money arrives. Split, decision rights, and the conditions under which each of you walks. Kraus got that renegotiation for free because the six of them were friends first, and most teams do not have that credit to spend.

Who This Is For

Founders before their first ten users get the most from this. The book lives in the gap between having an idea and having a user. Every interview covers that stretch in detail.

Operators scaling a working machine can skip it. The interviews stop at the end of the first act, and the second-order history after that is thinner than the early chapters.

Every founder in the book had already succeeded when Livingston interviewed them. The sample is selected on the outcome. The book cannot separate the moves that caused the win from the habits the winners happened to share. Teams that did the same things and died are not in the room. The figures are self-reported, and some conflict inside the book itself. Kraus puts the Excite bid for the Netscape button at 3 million dollars, and Brady recalls it as five.

Livingston ran the interviews in the mid two thousands. Seed rounds ran in the hundreds of thousands, and a button on Netscape was a distribution channel. Word of mouth inside one technical community reached a million people. The mechanics of raising money and reaching users have changed. The patterns about ideas, first versions, and cofounders survive the change.

The Decision

Run the test on your own users this week. Write down every person using what you built, by name.

Next to each name, put how they arrived and what they use it for. The founders here got their first users from a signature line, a free month, or an email to 40 friends. Some built the customer's site by hand. An empty list means the user count is zero, whatever the dashboard reports.

Then read the second column. Anyone using the product for something you did not design is telling you what the business is. That is the signal every founder in this book resisted first and followed second. Follow it earlier.