ISSUE 008
The Co-Founder Question
Somebody asked me last month whether they should find a partner before starting, because they had read that solo founders fail. I went looking for the study behind that. There isn't one. There is an essay from 2006, written by a man who invests in startups, which says it in the first paragraph and cites nothing. Meanwhile the Census Bureau counts owners, and the count says the one-person business is not the exception. Neither of those settles the question, which is the honest finding here.
An idea worth thinking about
The most repeated rule in startup advice is an essay from 2006 with no data in it
Paul Graham's essay "The 18 Mistakes That Kill Startups" is dated October 2006. Mistake number one, ahead of every other failure mode he lists, is Single Founder. His reasoning, in his words: a single founder is "a vote of no confidence", it "probably means the founder couldn't talk any of his friends into starting the company with him", and "starting a startup is too hard for one person". He adds that the low points are so low that few could bear them alone.
The essay cites no study, no dataset and no survey, and it does not pretend to. It is an experienced investor writing down what he has observed. That is a legitimate thing to publish. What is not legitimate is the twenty years of advice that has treated it as a finding. Graham co-founded Y Combinator, which invests in startups and has long preferred teams, so he is also not a neutral party to the question — worth saying out loud, not as an accusation but because nobody repeating the advice ever mentions it.
The opening line is the part I would flag, and this is my reading rather than his claim. "Have you ever noticed how few successful startups were founded by just one person?" That sentence looks at successful startups and counts founders. It is the same shape as noticing that few lottery winners bought one ticket. To learn anything you would have to start with all the businesses, solo and team, and follow them forward — which is what entries 001 and 004 in this pillar kept finding nobody had done.
The closest thing to real comparative evidence I found is a working paper by Jason Greenberg at NYU and Ethan Mollick at Wharton called "Sole Survivors", which followed Kickstarter projects that became real companies and reported that solo-founded ventures survived longer than team-founded ones, with the pattern reversing for nonprofits. I am not resting anything on it and I am not linking it. It is unpublished, it is about crowdfunded projects, and a Kickstarter campaign is not an HVAC company in Jackson. I mention it only because it points the opposite way from the famous advice, which should tell you how settled the question is. The absence of evidence for "never found alone" is not evidence for "always found alone". It is just an absence.
A tool I disappeared into
The Annual Business Survey, as the raw file, for free, with no interface in the way
Everything in slot five came out of one directory on a Census file server. It holds the Annual Business Survey for 2023 as plain pipe-delimited text, zipped, free, no account and no query builder. The file I used is AB2300CSCB01.zip. Its companion AB2300CSCB01_FIELDS.txt lists all fifty-two columns and is the only documentation you get.
Two warnings before you click. That zip is about 45 megabytes and expands to roughly 472 megabytes, which will surprise a laptop. And the file is not sorted for humans — every row is a combination of geography, industry code, sex, ethnicity, race, veteran status, question code and answer code, so you are filtering, not reading. If you want the national all-sectors total you want the rows where the geography type is national, the industry is 00, and sex, ethnicity, race and veteran status are all set to their total codes. Miss one of those and you will double-count your way to a number that is wrong by a factor of four.
What makes it worth the afternoon is that the questions are the ones you would actually ask. Number of owners. Whether the business is family-owned. Whether it is jointly owned by spouses. What kinds of customer it sells to. What kinds of worker it employs. Whether it has ceased operating. And the industry codes go down to a level where you can find something close to your own trade rather than "services".
The honest limit, which applies to everything in this entry: these are employer businesses only. The roughly twenty-eight million American businesses with no employees are not in this file. If you are one of those, this dataset does not describe you, and the pillar's entry 005 is the one that does.
Learn something
The partnership agreement Missouri writes for you if you never write one
If the question is whether to bring in a partner, the thing worth reading is not a book about co-founders. It is Chapter 347 of the Missouri Revised Statutes, free and searchable, which is the limited liability company law. Section 347.081 is the one about operating agreements. The dozens of sections around it are the defaults — what happens when your operating agreement is silent, which for a great many two-person Missouri businesses means what happens, full stop.
Take one example. Section 347.111 governs how profits and losses get allocated. If your operating agreement says nothing, losses are allocated according to the contributions each member made or promised, and profits follow the members' shares of distributions after contributions are repaid. That is a rule with real money in it, and its effective date is 1 December 1993. It was written long before your business existed and it will apply to your business by default.
Then read 347.121 and 347.123, on withdrawal of a member and the events that end membership, and 347.115, on whether an interest in the company can be assigned to somebody else. Those three answer the question people actually mean when they ask about partners, which is never "will we work well together". It is "what happens when one of us wants out, or dies, or stops showing up".
This is not legal advice and I am not a lawyer; it is the text of the statute and you should take a real one before you sign anything. Chapter 347 also covers LLCs only — corporations are Chapter 351, and a general partnership is a different chapter again. But reading two or three sections of it will tell you more about what a partner means in practice than any amount of writing about founding teams, and it costs an hour.
What we’re seeing
Sixty-one percent of American employer businesses have exactly one owner
The 2023 Annual Business Survey asked employer businesses how many owners they have. Of the firms that answered, 61.0 percent said one person. Another 34.8 percent said two to four. Five to ten owners accounted for 1.3 percent, eleven or more for 0.4 percent, and 1.9 percent were owned by a parent company, estate or trust. So roughly ninety-six out of every hundred employer businesses in this country are run by one person or by a group small enough to fit in a pickup.
The single-owner firms are not a rounding error and they are not toys. They employed 14.7 million people. They also took in 12.1 percent of all receipts, against 34.8 percent of firms holding two to four owners and taking 14.7 percent — which tells you something true and unflattering about scale: ownership concentration is not where the revenue is. The firms owned by a parent company or trust are 1.9 percent of the count and 29.5 percent of the receipts.
Three things about this number before anybody builds an argument on it. First, the survey asks how many owners the business has now, not how many people founded it. A business started by three partners and now owned by one is counted as one, and the Census table cannot see the difference. Second, this is employer businesses only — the roughly twenty-eight million American businesses with no employees are not in the table, and they are overwhelmingly single-owner, so the real national picture is more solo than this, not less. Third, 2,916,044 firms answered the question out of 5,934,950 employer firms, so a bit under half the population is doing the talking.
I also checked the same question in the 2018 survey, because a tidy story would be that solo ownership is rising. In 2018 it was 59.1 percent. That looks like a two-point increase and it is not one: the standard error on the 2023 figure is 1.8 points and on the 2018 figure 1.3, so the two numbers are not statistically distinguishable. Five years apart, the honest answer is that the share has not visibly moved.
One more row, because it connects to what entry 005 found. The survey separately asks about spousal ownership, and 10.2 percent of reporting firms — 288,736 of them — are jointly owned and equally operated by spouses. The most common business partner in America is not a co-founder you met at a conference. It is the person you live with. Nobody writes essays about that one.
What I would take from all of it. The question "should I find a co-founder" has no evidence-based answer, and anyone who gives you one confidently is working from an essay. What entry 004 in this pillar did find, in administrative data on 2.7 million founders, is that prior experience in the industry predicts outcomes where age does not. That points somewhere practical. Stop asking whether you need a partner and start asking what specific thing you cannot do — the trade, the books, the selling, the licence — and whether a partner is the cheapest way to get it. Sometimes it is. Often the answer is a contractor, a part-time hire, or six months of learning it yourself, none of which require you to give away half of a business for the rest of your life.
One question
If you have a business partner, what is the one thing they do that you could not?
And if you do not have one, the same question about the gap you are living with. One line either way.
More in this pillar as it's written.
— Shana
Sources link to the original publisher, never to a summary.