If you ladies leave my island, if you survive recruit training, you will be a weapon, you will be a minister of death, praying for war. But until that day you are pukes. You’re the lowest form of life on Earth. You are not even human fucking beings. You are nothing but unorganized grabastic pieces of amphibian shit!
—R. Lee Ermey as Gunnery Sergeant Hartman
Full Metal Jacket (1987)
JUNE 2010
This is how Y Combinator works.
For three months, the selected startup founders meet weekly for a dinner with some eminent startup personage. The dinners are not relaxed social affairs: they’re competitive demos in which founders try to one-up each other with increasingly developed products, upward-sloping user graphs, or funding news, within a context of technocamaraderie and shared suffering. The weekly cadence imposes order on the always-full-throttle startup chaos, and is a welcome respite from the grinding toil and stomach-churning stress.
I always tried to sit directly in the front row to take the speaker’s measure.
Marissa Mayer’s hands shook and she spoke at an anxious clip; she arrived escorted by a Google “handler,” the only speaker to do so.
Reid Hoffman, a large man, thundered like an emperor at the head of this troops, and spilled wonderful stories about the PayPal saga, the founding of LinkedIn, and going to war with Microsoft. He was by far one of the best speakers; you felt like kicking your way through a brick wall when he was done.
And on it went with the creators of Gmail and Yahoo Mail, partners at Sequoia and Google Ventures, the founders of Airbnb, Eventbrite, and Groupon and the like, regaling us with war stories from the tech trenches.
The dinner food was what Paul Graham politely called “goop on rice”: an occasionally foul, sporadically delicious stew of random ingredients tossed over a bowl of rice. PG or one of the other partners would dole it out themselves, the founders forming a meandering line, like a Depression-era soup kitchen. YC’s main space had a full kitchen, and supposedly PG actually cooked the “goop” himself—or at least he had in the early days. The creamy stuff with peas and some mystery fish was without a doubt the worst. Those were bad weeks. Microwaved spackling paste surely tasted better. There was general agreement the meatballs on pasta were the global maximum (which was saying something).
You obviously weren’t there for the food, or even the distinguished speaker’s company.
The real YC selling points were the following: access to the YC partners, access to the network of YC founders, a bankable patina of prestige, and Demo Day.
The value of the partners will be made clear as this story progresses.
The value of the network of YC companies is that it essentially constitutes a private microcosm of the greater tech world. This is often derided as the “YC mafia” by outsiders.
“Mafia” is a pejorative term bringing to mind tracksuit-wearing Russians bolting vodka, shouting into burner cellphones, and whispering to each other in raspy, Slavic tones. There is a feeling of collective defense around YC, and it does know how to circle the wagons if threatened.
The real network effect, however, was this: With the pool of YC companies expanding every year, you could probably re-create 80 percent of the consumer and infrastructure technology used in our Internet-enabled age exclusively with Y Combinator companies. Whatever need you had, whether system-monitoring tools, mobile development, or even marketing tools (“Have you heard of our product, AdGrok?”), you could find a YC company to fill it. Given that you were one of the family, you could expect excellent service and a steep discount. You could also be sure that whatever you were building would receive preferential adoption by others in the YC network, providing you with an instant set of savvy and patient users, as well as impressive logos to feature in a pitch deck. This tendency to “dogfood” YC products came straight from the top. The day YC funds an airline, I suspect PG will fly that one, and no other.
As for the value of the bankable prestige, we’d see just what that was soon enough.
So what were we building in Y Combinator?
To understand that, we’ll need one last lecture on online media, and then I’ll step down from the pedagogical soapbox. It’s pure money, sex, and death after this, I promise.
How does Google manage to generate yearly revenues of $70 billon, greater than the GDP of Luxembourg or Belarus? It invented this magical website called Google Search, where all of humanity goes and tells Google what it wants: “Nikon D300 camera.” “Online nursing degree.” “Divorce lawyer in Atlanta.” A world of of three- and four-word desires and needs, all craving to be satisfied, all backed by wallets waiting to be opened. Injecting themselves in that last moment of purchase, at the apex of desire, Google invites you to click on an ad. Its cash register rings every time you click.
It doesn’t even need to figure out what a search query is worth and price it accordingly. It simply holds an auction for every search query entered, at the moment the query occurs. The net result is that billions of times a day, Google runs an auction of keywords and accompanying bids. By looking at the bid, and estimating the likelihood of a click, Google takes the product of the two (which is how much it will make per query) and picks the highest. Then it displays the associated ad that the advertiser has created and uploaded to Google for that keyword.
Actually printing physical money would be harder.
So how many such search queries, or “keywords,” in Google-speak, are there?
The second edition of the Oxford English Dictionary, released in 1989 and since supplemented, contains 291,000 word entries. The Woordenboek der Nederlandsche Taal, a dictionary of the Dutch language and the largest monolingual dictionary in the world, runs to 50,000 pages and 431,000 entries. Both works are dwarfed by the size of the keyword lists maintained by those lexicographers turned word merchants, the search engine marketers. Like a stock portfolio manager, who keeps a set of assets with a theoretical and current price, the paid search manager maintains encyclopedic word lists along with dollar-sign values, and constantly adjusts bids to reflect realized performance. These lists literally number in the millions, and look something like:
KEYWORD |
COST PER CLICK |
REVENUE PER CLICK |
“divorce lawyer in reno” |
$1.45 |
$0.90 |
“nevada cheap divorce” |
$0.75 |
$1.10 |
“nevada divorce lawyer” |
$5.55 |
$2.75 |
Like the classic stock picker buying low and selling high, our search engine marketer curates and trims a keyword list like a bonsai tree, buying more of well-performing keywords, and fewer of bad. If the revenue generated by postclick sales outpaces cost, up go the bid and the budget, and the reverse in the opposite case. The ratio of revenue to cost is known as “return on advertising spend” (ROAS) and is the basic metric all marketers in every medium use. As an example, ROAS is $1.10 ÷ $0.75 – 1 = 47% for “nevada cheap divorce” above. This means for every dollar I put into Google for that keyword, I get $1.47 back, at least as projected based on historical data. I’m happy to do that all day. Time to up the budget.
Such is the essential busywork behind how a Google makes more than some European countries produce in a year. That’s it. You now know as much as the best search engine marketers in the world.
As a piece of clickbait-y news, want to know which is the most expensive word in the English language? Around 2011 or so, and probably still to this day, the priciest word in the global auction on words was “mesothelioma.” This tongue twister is a rare form of lung disease common among former asbestos-plant workers. Thanks to a series of class-action lawsuits against former factory owners, filed by plaintiffs’ attorneys who make fortunes on contingency fees, the value of this word was bid up as high as $90 per click. Want to screw a slimy lawyer? Google “mesothelioma” and start randomly clicking on the ads that appear. You’re costing a lawyer almost a whole benjamin every time you do that.
Lung-cancer words, despite their superlative cost, are still a pretty niche market. What are the costliest Google keywords among relatively high-volume keywords? The ranking changes, but the top ten is always composed of some combination of “insurance,” “loans,” “mortgage,” “classes,” “credit,” “lawyer,” and so on. These are Google’s moneymakers, which pay for the Android phones, the Chrome browser, the self-driving cars, the flying Wi-Fi balloons, and whatever weird, geeky, philanthropic shit the company is up to recently.
Think about this in the context of more traditional industries for a moment. Chain restaurants like McDonald’s have a best-performing outlet in a particularly busy high-rent district. Automakers have a particularly popular, bestselling model like the Ford Fusion or the Chevy Impala that makes their quarter. Google has the words “insurance” and “loans.” Those are its flagship moneymakers. It is a tech empire built on snippets of words and phrases flitting through people’s minds. You as a mere consumer don’t see it, but how Google ranks keywords and runs the auction determines the fate of billion-dollar companies and industries. As the gatekeeper on buying interest, Google is the bouncer at the door of almost any Internet-enabled business today, and if you as the business owner don’t pay your bouncer well, he’ll shut you down, as Google has done more than once.
So what was our angle on this?
Very simple.
This mountain of money Google earns doesn’t flow through the simple ads-buying tools it provides. Those are too rudimentary for the experienced marketer. Middlemen companies, like brokers on an exchange, offer sophisticated tools for those spending millions a year on Google keywords. Small businesses, though—the custom jewelry maker with an online store on Etsy, the local plumber—have no such tools, as Google’s crude tools are also too confusing for them, and the auction in keywords too dynamic. It’s as if the financial world had large investment banks like Goldman, but no Charles Schwab for the average investor to use. Since Google makes most of its money elsewhere, and in any case doesn’t have the patience or internal culture to create tools for small advertisers, the tools it does offer are dauntingly complex and ineffectual. The outstanding problem here is like the last-mile problem of an Internet service provider; some piece of technology is needed in order to travel from the fat-pipe fiber-optic cable to the home user wanting to stream Netflix. We would be that last link, allowing a mom-and-pop to finally spend money on Google, rather than pissing away some experimental budget and then getting burned out due to poorly performing keywords or ill-advised bids.
We had plenty of competition.
A company named Clickable (now dead) had an all-star cast of advertising investors, and had raised over $32 million to crack this same nut. Lexity (now dead) was founded by a former Yahoo exec, and raised $6 million. Trada (now dead) had pursued the interesting approach of crowdsourcing the problem, establishing a marketplace where advertisers could easily find campaign managers to do the workaday management of Google search campaigns. It raised $19 million, including money from Google Ventures, the venture arm of the very partner company it was connecting to advertisers.
You’ll notice lots of casualties here. We didn’t know it at the time, but nobody would succeed in closing the last-mile gap between Google and the universe of small business; to this day, it’s an unsolved problem. Despite all this techie fix-it boosterism, not every problem has an engineering solution. That doesn’t mean, of course, that you can’t sell one.
Following the YC playbook, we’d do boot camp close to YC headquarters, isolated from the beguiling distractions of San Francisco.
I found us a cheap one-bedroom apartment to serve as an office three blocks west of Castro Street, the main drag in Mountain View. Other than serving as Google’s hometown, Mountain View is just one more in the string of towns dotting the 101 and the Caltrain line from San Francisco to San Jose. More down-market and working-class than posh Palo Alto or Menlo Park, it housed a couple of startups, as well as the law firm Fenwick & West, an entity we would, sadly, come to know well. Smack in the middle of downtown was Red Rock Coffee, about the most hacker and startup-y café on the Peninsula, whose weaponized sugar-and-caffeine mochas would keep us going through the coming weeks.*
I had just moved out of my Mission bachelor pad in SF and in with British Trader and little Zoë (at this point our relationship situation was tenuous but hopeful), and had furniture to spare. Our corporate headquarters soon had a futon, a mattress we’d toss on the ground in fine Tijuana-whorehouse style, and three desks we made out of cheap interior doors from Home Depot and sawhorses. We used the YC nickels to buy some computer hardware—monitors, new machines for the boys—and we set to work coding our fucking asses off.
In the first week in the pad while we waited for Argyris to disentangle himself from Adchemy, MRM and I sketched out the vision for what the AdGrok product, called the GrokBar, would look like on two big 4×8 sheets of markerboard ($11.99 at Home Depot!).†
Our initial discussion of the GrokBar actually dated to an email from me to the boys on April 16, while we were still at Adchemy, and was a shameless rip on the recently discontinued DiggBar. This was a special window that lived inside the user’s browser, and allowed him or her to “Digg” ’ (an early version of a Facebook Like) a piece of content. Rather than requiring site owners to include special Digg code on their website, or requiring a user to cut and paste URLs into digg.com, a browsing user could simply comment and “Digg” stuff as part of his or her normal browsing behavior. It was as if someone surfing the Web had a heads-up display that showed the Internet’s opinion of each piece of content, everywhere that person went online.
Given its parentage, the early GrokBar closely resembled the DiggBar, in that it accompanied users as they navigated their own online shop. Stuck in the upper reaches of a browser pane, it was slim and relatively unobtrusive, listing stats for the Google Ads you were running. As you browsed your online store, the stats would slice the data by the product whose page you were looking at, providing an in-context view of what ads on Google were driving sales for that product, and at what price. In the same way that Digg showed your friends’ comments and likes of an article, the GrokBar showed you how Google was driving traffic to that product’s page, which keywords people were searching for to get there, and how much you were paying to show ads alongside the search results. Of course, the bar would be present only when you were on your online store or company page, and not otherwise.
But how to sell it?
Most small startups decide to go after the small-to-medium-sized businesses (SMB) market because they think it’s an easy mark. The enterprise sale is too hard, the sales cycle too long, and big companies too untrusting (perhaps reasonably so) of early-stage companies. So they sell to that venerable couple, that mythical bedrock of American values so loved by politicians: Mom and Pop. While it’s true that Mom and Pop are likely to try anything once, and that the quality of their typical software and service would make Italian phone companies look cutting edge by comparison, it is also true that Mom and Pop are phenomenally flaky, and are likely to cancel their subscription to even a useful service, making user turnover a problem. Also, given that you’re making fifty to one hundred dollars per month for every sale, those nickels and dimes mean any sort of high-touch sales process is unscalable even for underpaid startup founders. So you’ve got to scale the sale somehow, either by partnering with someone who already owns the SMB relationship (think Salesforce, or the advertising departments of newspapers), or by building on an existing platform with small-business clients, assuming that’s possible with your technology.
In our case, we barely had any success with partnering, and had modest success with the building. But that’s getting ahead of our story. Let’s go back to that dumpy one-bedroom on Oak Street in Mountain View where three scared guys were bailing water to save their lives.