September 2012A startup is a company designed to grow fast. Being newly foundeddoes not in itself make a company a startup. Nor is it necessaryfor a startup to work on technology, or take venture funding, orhave some sort of "exit." The only essential thing is growth.Everything else we associate with startups follows from growth.If you want to start one it's important to understand that. Startupsare so hard that you can't be pointed off to the side and hope tosucceed. You have to know that growth is what you're after. Thegood news is, if you get growth, everything else tends to fall intoplace. Which means you can use growth like a compass to make almostevery decision you face.RedwoodsLet's start with a distinction that should be obvious but is oftenoverlooked: not every newly founded company is a startup. Millionsof companies are started every year in the US. Only a tiny fractionare startups. Most are service businesses — restaurants, barbershops,plumbers, and so on. These are not startups, except in a few unusualcases. A barbershop isn't designed to grow fast. Whereas a searchengine, for example, is.When I say startups are designed to grow fast, I mean it in twosenses. Partly I mean designed in the sense of intended, becausemost startups fail. But I also mean startups are different bynature, in the same way a redwood seedling has a different destinyfrom a bean sprout.That difference is why there's a distinct word, "startup," forcompanies designed to grow fast. If all companies were essentiallysimilar, but some through luck or the efforts of their foundersended up growing very fast, we wouldn't need a separate word. Wecould just talk about super-successful companies and less successfulones. But in fact startups do have a different sort of DNA fromother businesses. Google is not just a barbershop whose founderswere unusually lucky and hard-working. Google was different fromthe beginning.To grow rapidly, you need to make something you can sell to a bigmarket. That's the difference between Google and a barbershop. Abarbershop doesn't scale.For a company to grow really big, it must (a) make something lotsof people want, and (b) reach and serve all those people. Barbershopsare doing fine in the (a) department. Almost everyone needs theirhair cut. The problem for a barbershop, as for any retailestablishment, is (b). A barbershop serves customers in person,and few will travel far for a haircut. And even if they did, thebarbershop couldn't accomodate them. [1]Writing software is a great way to solve (b), but you can still endup constrained in (a). If you write software to teach Tibetan toHungarian speakers, you'll be able to reach most of the people whowant it, but there won't be many of them. If you make softwareto teach English to Chinese speakers, however, you're in startupterritory.Most businesses are tightly constrained in (a) or (b). The distinctivefeature of successful startups is that they're not.IdeasIt might seem that it would always be better to start a startupthan an ordinary business. If you're going to start a company, whynot start the type with the most potential? The catch is that thisis a (fairly) efficient market. If you write software to teachTibetan to Hungarians, you won't have much competition. If youwrite software to teach English to Chinese speakers, you'll faceferocious competition, precisely because that's such a larger prize.[2]The constraints that limit ordinary companies also protect them.That's the tradeoff. If you start a barbershop, you only have tocompete with other local barbers. If you start a search engine youhave to compete with the whole world.The most important thing that the constraints on a normal businessprotect it from is not competition, however, but the difficulty ofcoming up with new ideas. If you open a bar in a particularneighborhood, as well as limiting your potential and protecting youfrom competitors, that geographic constraint also helps define yourcompany. Bar + neighborhood is a sufficient idea for a smallbusiness. Similarly for companies constrained in (a). Your nicheboth protects and defines you.Whereas if you want to start a startup, you're probably going tohave to think of something fairly novel. A startup has to makesomething it can deliver to a large market, and ideas of that typeare so valuable that all the obvious ones are already taken.That space of ideas has been so thoroughly picked over that a startupgenerally has to work on something everyone else has overlooked.I was going to write that one has to make a conscious effort tofind ideas everyone else has overlooked. But that's not how moststartups get started. Usually successful startups happen becausethe founders are sufficiently different from other people that ideasfew others can see seem obvious to them. Perhaps later they stepback and notice they've found an idea in everyone else's blind spot,and from that point make a deliberate effort to stay there. [3]But at the moment when successful startups get started, much of theinnovation is unconscious.What's different about successful founders is that they can seedifferent problems. It's a particularly good combination both tobe good at technology and to face problems that can be solved byit, because technology changes so rapidly that formerly bad ideasoften become good without anyone noticing. Steve Wozniak's problemwas that he wanted his own computer. That was an unusual problemto have in 1975. But technological change was about to make it amuch more common one. Because he not only wanted a computer butknew how to build them, Wozniak was able to make himself one. Andthe problem he solved for himself became one that Apple solved formillions of people in the coming years. But by the time it wasobvious to ordinary people that this was a big market, Apple wasalready established.Google has similar origins. Larry Page and Sergey Brin wanted tosearch the web. But unlike most people they had the technicalexpertise both to notice that existing search engines were not asgood as they could be, and to know how to improve them. Over thenext few years their problem became everyone's problem, as the webgrew to a size where you didn't have to be a picky search expertto notice the old algorithms weren't good enough. But as happenedwith Apple, by the time everyone else realized how important searchwas, Google was entrenched.That's one connection between startup ideas and technology. Rapidchange in one area uncovers big, soluble problems in other areas.Sometimes the changes are advances, and what they change is solubility.That was the kind of change that yielded Apple; advances in chiptechnology finally let Steve Wozniak design a computer he couldafford. But in Google's case the most important change was thegrowth of the web. What changed there was not solubility but bigness.The other connection between startups and technology is that startupscreate new ways of doing things, and new ways of doing things are,in the broader sense of the word, new technology. When a startup both begins with anidea exposed by technological change and makes a product consistingof technology in the narrower sense (what used to be called "hightechnology"), it's easy to conflate the two. But the two connectionsare distinct and in principle one could start a startup that wasneither driven by technological change, nor whose product consistedof technology except in the broader sense. [4]RateHow fast does a company have to grow to be considered a startup?There's no precise answer to that. "Startup" is a pole, not athreshold. Starting one is at first no more than a declaration ofone's ambitions. You're committing not just to starting a company,but to starting a fast growing one, and you're thus committing tosearch for one of the rare ideas of that type. But at first youhave no more than commitment. Starting a startup is like being anactor in that respect. "Actor" too is a pole rather than a threshold.At the beginning of his career, an actor is a waiter who goes toauditions. Getting work makes him a successful actor, but he doesn'tonly become an actor when he's successful.So the real question is not what growth rate makes a company astartup, but what growth rate successful startups tend to have.For founders that's more than a theoretical question, because it'sequivalent to asking if they're on the right path.The growth of a successful startup usually has three phases: There's an initial period of slow or no growth while the startup tries to figure out what it's doing. As the startup figures out how to make something lots of people want and how to reach those people, there's a period of rapid growth. Eventually a successful startup will grow into a big company. Growth will slow, partly due to internal limits and partly because the company is starting to bump up against the limits of the markets it serves. [5]Together these three phases produce an S-curve. The phase whosegrowth defines the startup is the second one, the ascent. Itslength and slope determine how big the company will be.The slope is the company's growth rate. If there's one number everyfounder should always know, it's the company's growth rate. That'sthe measure of a startup. If you don't know that number, you don'teven know if you're doing well or badly.When I first meet founders and ask what their growth rate is,sometimes they tell me "we get about a hundred new customers amonth." That's not a rate. What matters is not the absolute numberof new customers, but the ratio of new customers to existing ones.If you're really getting a constant number of new customers everymonth, you're in trouble, because that means your growth rate isdecreasing.During Y Combinator we measure growth rate per week, partly becausethere is so little time before Demo Day, and partly because startupsearly on need frequent feedback from their users to tweak whatthey're doing. [6]A good growth rate during YC is 5-7% a week. If you can hit 10% aweek you're doing exceptionally well. If you can only manage 1%,it's a sign you haven't yet figured out what you're doing.The best thing to measure the growth rate of is revenue. The nextbest, for startups that aren't charging initially, is active users.That's a reasonable proxy for revenue growth because whenever thestartup does start trying to make money, their revenues will probablybe a constant multiple of active users. [7]CompassWe usually advise startups to pick a growth rate they think theycan hit, and then just try to hit it every week. The key word hereis "just." If they decide to grow at 7% a week and they hit thatnumber, they're successful for that week. There's nothing morethey need to do. But if they don't hit it, they've failed in theonly thing that mattered, and should be correspondingly alarmed.Programmers will recognize what we're doing here. We're turningstarting a startup into an optimization problem. And anyone whohas tried optimizing code knows how wonderfully effective that sortof narrow focus can be. Optimizing code means taking an existingprogram and changing it to use less of something, usually time ormemory. You don't have to think about what the program should do,just make it faster. For most programmers this is very satisfyingwork. The narrow focus makes it a sort of puzzle, and you'regenerally surprised how fast you can solve it.Focusing on hitting a growth rate reduces the otherwise bewilderinglymultifarious problem of starting a startup to a single problem.You can use that target growth rate to make all your decisions foryou; anything that gets you the growth you need is ipso facto right.Should you spend two days at a conference? Should you hire anotherprogrammer? Should you focus more on marketing? Should you spendtime courting some big customer? Should you add x feature? Whatevergets you your target growth rate. [8]Judging yourself by weekly growth doesn't mean you can look no morethan a week ahead. Once you experience the pain of missing yourtarget one week (it was the only thing that mattered, and you failedat it), you become interested in anything that could spare you suchpain in the future. So you'll be willing for example to hire anotherprogrammer, who won't contribute to this week's growth but perhapsin a month will have implemented some new feature that will get youmore users. But only if (a) the distraction of hiring someonewon't make you miss your numbers in the short term, and (b) you'resufficiently worried about whether you can keep hitting your numberswithout hiring someone new.It's not that you don't think about the future, just that you thinkabout it no more than necessary.In theory this sort of hill-climbing could get a startup intotrouble. They could end up on a local maximum. But in practicethat never happens. Having to hit a growth number every week forcesfounders to act, and acting versus not acting is the high bit ofsucceeding. Nine times out of ten, sitting around strategizing isjust a form of procrastination. Whereas founders' intuitions aboutwhich hill to climb are usually better than they realize. Plus themaxima in the space of startup ideas are not spiky and isolated.Most fairly good ideas are adjacent to even better ones.The fascinating thing about optimizing for growth is that it canactually discover startup ideas. You can use the need for growthas a form of evolutionary pressure. If you start out with someinitial plan and modify it as necessary to keep hitting, say, 10%weekly growth, you may end up with a quite different company thanyou meant to start. But anything that grows consistently at 10% aweek is almost certainly a better idea than you started with.There's a parallel here to small businesses. Just as the constraintof being located in a particular neighborhood helps define a bar,the constraint of growing at a certain rate can help define astartup.You'll generally do best to follow that constraint wherever it leadsrather than being influenced by some initial vision, just as ascientist is better off following the truth wherever it leads ratherthan being influenced by what he wishes were the case. When RichardFeynman said that the imagination of nature was greater than theimagination of man, he meant that if you just keep following thetruth you'll discover cooler things than you could ever have madeup. For startups, growth is a constraint much like truth. Everysuccessful startup is at least partly a product of the imaginationof growth. [9]ValueIt's hard to find something that grows consistently at severalpercent a week, but if you do you may have found something surprisinglyvaluable. If we project forward we see why.weeklyyearly1%1.7x2%2.8x5%12.6x7%33.7x10%142.0xA company that grows at 1% a week will grow 1.7x a year, whereas acompany that grows at 5% a week will grow 12.6x. A company making$1000 a month (a typical number early in YC) and growing at 1% aweek will 4 years later be making $7900 a month, which is less thana good programmer makes in salary in Silicon Valley. A startupthat grows at 5% a week will in 4 years be making $25 million amonth. [10]Our ancestors must rarely have encountered cases of exponentialgrowth, because our intuitions are no guide here. What happensto fast growing startups tends to surprise even the founders.Small variations in growth rate produce qualitatively differentoutcomes. That's why there's a separate word for startups, and whystartups do things that ordinary companies don't, like raising moneyand getting acquired. And, strangely enough, it's also why theyfail so frequently.Considering how valuable a successful startup can become, anyonefamiliar with the concept of expected value would be surprised ifthe failure rate weren't high. If a successful startup could makea founder $100 million, then even if the chance of succeeding wereonly 1%, the expected value of starting one would be $1 million.And the probability of a group of sufficiently smart and determinedfounders succeeding on that scale might be significantly over 1%.For the right people — e.g. the young Bill Gates — the probabilitymight be 20% or even 50%. So it's not surprising that so many wantto take a shot at it. In an efficient market, the number of failedstartups should be proportionate to the size of the successes. Andsince the latter is huge the former should be too. [11]What this means is that at any given time, the great majority ofstartups will be working on something that's never going to goanywhere, and yet glorifying their doomed efforts with the grandiosetitle of "startup."This doesn't bother me. It's the same with other high-beta vocations,like being an actor or a novelist. I've long since gotten used toit. But it seems to bother a lot of people, particularly thosewho've started ordinary businesses. Many are annoyed that theseso-called startups get all the attention, when hardly any of themwill amount to anything.If they stepped back and looked at the whole picture they might beless indignant. The mistake they're making is that by basing theiropinions on anecdotal evidence they're implicitly judging by themedian rather than the average. If you judge by the median startup,the whole concept of a startup seems like a fraud. You have toinvent a bubble to explain why founders want to start them orinvestors want to fund them. But it's a mistake to use the medianin a domain with so much variation. If you look at the averageoutcome rather than the median, you can understand why investorslike them, and why, if they aren't median people, it's a rationalchoice for founders to start them.DealsWhy do investors like startups so much? Why are they so hot toinvest in photo-sharing apps, rather than solid money-makingbusinesses? Not only for the obvious reason.The test of any investment is the ratio of return to risk. Startupspass that test because although they're appallingly risky, thereturns when they do succeed are so high. But that's not the onlyreason investors like startups. An ordinary slower-growing businessmight have just as good a ratio of return to risk, if both werelower. So why are VCs interested only in high-growth companies?The reason is that they get paid by getting their capital back,ideally after the startup IPOs, or failing that when it's acquired.The other way to get returns from an investment is in the form ofdividends. Why isn't there a parallel VC industry that invests inordinary companies in return for a percentage of their profits?Because it's too easy for people who control a private company tofunnel its revenues to themselves (e.g. by buying overpricedcomponents from a supplier they control) while making it look likethe company is making little profit. Anyone who invested in privatecompanies in return for dividends would have to pay close attentionto their books.The reason VCs like to invest in startups is not simply the returns,but also because such investments are so easy to oversee. Thefounders can't enrich themselves without also enriching the investors.[12]Why do founders want to take the VCs' money? Growth, again. Theconstraint between good ideas and growth operates in both directions.It's not merely that you need a scalable idea to grow. If you havesuch an idea and don't grow fast enough, competitors will. Growingtoo slowly is particularly dangerous in a business with networkeffects, which the best startups usually have to some degree.Almost every company needs some amount of funding to get started.But startups often raise money even when they are or could beprofitable. It might seem foolish to sell stock in a profitablecompany for less than you think it will later be worth, but it'sno more foolish than buying insurance. Fundamentally that's howthe most successful startups view fundraising. They could grow thecompany on its own revenues, but the extra money and help suppliedby VCs will let them grow even faster. Raising money lets youchoose your growth rate.Money to grow faster is always at the command of the most successfulstartups, because the VCs need them more than they need the VCs.A profitable startup could if it wanted just grow on its own revenues.Growing slower might be slightly dangerous, but chances are itwouldn't kill them. Whereas VCs need to invest in startups, andin particular the most successful startups, or they'll be out ofbusiness. Which means that any sufficiently promising startup willbe offered money on terms they'd be crazy to refuse. And yet becauseof the scale of the successes in the startup business, VCs can stillmake money from such investments. You'd have to be crazy to believeyour company was going to become as valuable as a high growth ratecan make it, but some do.Pretty much every successful startup will get acquisition offerstoo. Why? What is it about startups that makes other companieswant to buy them? [13]Fundamentally the same thing that makes everyone else want the stockof successful startups: a rapidly growing company is valuable. It'sa good thing eBay bought Paypal, for example, because Paypal is nowresponsible for 43% of their sales and probably more of their growth.But acquirers have an additional reason to want startups. A rapidlygrowing company is not merely valuable, but dangerous. If it keepsexpanding, it might expand into the acquirer's own territory. Mostproduct acquisitions have some component of fear. Even if anacquirer isn't threatened by the startup itself, they might bealarmed at the thought of what a competitor could do with it. Andbecause startups are in this sense doubly valuable to acquirers,acquirers will often pay more than an ordinary investor would. [14]UnderstandThe combination of founders, investors, and acquirers forms a naturalecosystem. It works so well that those who don't understand it aredriven to invent conspiracy theories to explain how neatly thingssometimes turn out. Just as our ancestors did to explain theapparently too neat workings of the natural world. But there isno secret cabal making it all work.If you start from the mistaken assumption that Instagram wasworthless, you have to invent a secret boss to force Mark Zuckerbergto buy it. To anyone who knows Mark Zuckerberg, that is the reductioad absurdum of the initial assumption. The reason he bought Instagramwas that it was valuable and dangerous, and what made it so wasgrowth.If you want to understand startups, understand growth. Growthdrives everything in this world. Growth is why startups usuallywork on technology — because ideas for fast growing companies areso rare that the best way to find new ones is to discover thoserecently made viable by change, and technology is the best sourceof rapid change. Growth is why it's a rational choice economicallyfor so many founders to try starting a startup: growth makes thesuccessful companies so valuable that the expected value is higheven though the risk is too. Growth is why VCs want to invest instartups: not just because the returns are high but also becausegenerating returns from capital gains is easier to manage thangenerating returns from dividends. Growth explains why the mostsuccessful startups take VC money even if they don't need to: itlets them choose their growth rate. And growth explains whysuccessful startups almost invariably get acquisition offers. Toacquirers a fast-growing company is not merely valuable but dangeroustoo.It's not just that if you want to succeed in some domain, you haveto understand the forces driving it. Understanding growth is whatstarting a startup consists of. What you're really doing (andto the dismay of some observers, all you're really doing) when youstart a startup is committing to solve a harder type of problemthan ordinary businesses do. You're committing to search for oneof the rare ideas that generates rapid growth. Because these ideasare so valuable, finding one is hard. The startup is the embodimentof your discoveries so far. Starting a startup is thus very muchlike deciding to be a research scientist: you're not committing tosolve any specific problem; you don't know for sure which problemsare soluble; but you're committing to try to discover something noone knew before. A startup founder is in effect an economic researchscientist. Most don't discover anything that remarkable, but somediscover relativity.Notes[1]Strictly speaking it's not lots of customers you need but a bigmarket, meaning a high product of number of customers times howmuch they'll pay. But it's dangerous to have too few customerseven if they pay a lot, or the power that individual customers haveover you could turn you into a de facto consulting firm. So whatevermarket you're in, you'll usually do best to err on the side ofmaking the broadest type of product for it.[2]One year at Startup School David Heinemeier Hansson encouragedprogrammers who wanted to start businesses to use a restaurant asa model. What he meant, I believe, is that it's fine to startsoftware companies constrained in (a) in the same way a restaurantis constrained in (b). I agree. Most people should not try tostart startups.[3]That sort of stepping back is one of the things we focus on atY Combinator. It's common for founders to have discovered somethingintuitively without understanding all its implications. That'sprobably true of the biggest discoveries in any field.[4]I got it wrong in "How to Make Wealth" when I said that astartup was a small company that takes on a hard technicalproblem. That is the most common recipe but not the only one.[5]In principle companies aren't limited by the size of the marketsthey serve, because they could just expand into new markets. Butthere seem to be limits on the ability of big companies to do that.Which means the slowdown that comes from bumping up against thelimits of one's markets is ultimately just another way in whichinternal limits are expressed.It may be that some of these limits could be overcome by changingthe shape of the organization — specifically by sharding it.[6]This is, obviously, only for startups that have already launchedor can launch during YC. A startup building a new database willprobably not do that. On the other hand, launching something smalland then using growth rate as evolutionary pressure is such avaluable technique that any company that could start this wayprobably should.[7]If the startup is taking the Facebook/Twitter route and buildingsomething they hope will be very popular but from which they don'tyet have a definite plan to make money, the growth rate has to behigher, even though it's a proxy for revenue growth, because suchcompanies need huge numbers of users to succeed at all.Beware too of the edge case where something spreads rapidly but thechurn is high as well, so that you have good net growth till you runthrough all the potential users, at which point it suddenly stops.[8]Within YC when we say it's ipso facto right to do whatever getsyou growth, it's implicit that this excludes trickery like buyingusers for more than their lifetime value, counting users as activewhen they're really not, bleeding out invites at a regularlyincreasing rate to manufacture a perfect growth curve, etc. Evenif you were able to fool investors with such tricks, you'd ultimatelybe hurting yourself, because you're throwing off your own compass.[9]Which is why it's such a dangerous mistake to believe thatsuccessful startups are simply the embodiment of some brilliantinitial idea. What you're looking for initially is not so much agreat idea as an idea that could evolve into a great one. Thedanger is that promising ideas are not merely blurry versions ofgreat ones. They're often different in kind, because the earlyadopters you evolve the idea upon have different needs from therest of the market. For example, the idea that evolves into Facebookisn't merely a subset of Facebook; the idea that evolves intoFacebook is a site for Harvard undergrads.[10]What if a company grew at 1.7x a year for a really long time?Could it not grow just as big as any successful startup? In principleyes, of course. If our hypothetical company making $1000 a monthgrew at 1% a week for 19 years, it would grow as big as a companygrowing at 5% a week for 4 years. But while such trajectories maybe common in, say, real estate development, you don't see them muchin the technology business. In technology, companies that growslowly tend not to grow as big.[11]Any expected value calculation varies from person to persondepending on their utility function for money. I.e. the firstmillion is worth more to most people than subsequent millions. Howmuch more depends on the person. For founders who are younger ormore ambitious the utility function is flatter. Which is probablypart of the reason the founders of the most successful startups ofall tend to be on the young side.[12]More precisely, this is the case in the biggest winners, whichis where all the returns come from. A startup founder could pullthe same trick of enriching himself at the company's expense byselling them overpriced components. But it wouldn't be worth itfor the founders of Google to do that. Only founders of failingstartups would even be tempted, but those are writeoffs from theVCs' point of view anyway.[13]Acquisitions fall into two categories: those where the acquirerwants the business, and those where the acquirer just wants theemployees. The latter type is sometimes called an HR acquisition.Though nominally acquisitions and sometimes on a scale that has asignificant effect on the expected value calculation for potentialfounders, HR acquisitions are viewed by acquirers as more akin tohiring bonuses.[14]I once explained this to some founders who had recently arrivedfrom Russia. They found it novel that if you threatened a companythey'd pay a premium for you. "In Russia they just kill you," theysaid, and they were only partly joking. Economically, the factthat established companies can't simply eliminate new competitorsmay be one of the most valuable aspects of the rule of law. Andso to the extent we see incumbents suppressing competitors viaregulations or patent suits, we should worry, not because it's adeparture from the rule of law per se but from what the rule of lawis aiming at.Thanks to Sam Altman, Marc Andreessen, Paul Buchheit, PatrickCollison, Jessica Livingston, Geoff Ralston, and Harj Taggar forreading drafts of this.
