DoorDash Wants to Own the Last Mile

From left to right, Tony Xu, Co-founder and CEO, Stanley Tang, Co-founder and CPO, Ding Zhou, Vice President of Engineering, and Andy Fang, Co-founder and CTO of DoorDash.It’s 11 am on a weekday, and the magic hour for lunch orders has begun. Tony Xu is ready to do some Dashing. He’s wearing his bright red DoorDash t-shirt. His water bottle is full. He’s got the Dasher app loaded on his iPhone. He checks the trunk of his 2001 Honda Accord to verify he’s got enough hot bags. Checks the dashboard to make sure he’s got enough gas. And then he checks the Dasher app to see if he’s got action. Yep. A pickup at Asian Box, at the Town and Country center, across El Camino Real from the Stanford campus. Delivery will be to downtown Palo Alto. Xu backs out of his spot at the former Palo Alto Animal Hospital, and rolls onto the boulevard. His phone provides navigation to the restaurant and an estimated time of arrival.

Xu belongs to a huge squad of red-shirted workers blanketing the Peninsula today. One in three Silicon Valley households uses DoorDash. Nationwide, there are “tens of thousands” of Dashers. But unlike almost all of these contract workers, Xu is an actual employee. In fact, he’s the CEO. While he’s making this particular delivery for my benefit, one day a month he puts on the bright red tee to drive, bike, or walk to local eateries and trundle comestibles to the customer’s door. It’s required of everyone in the company.

Before he arrives at Asian Box — a busy operation focusing on takeout, where Stanford students and startup techies line up to construct meals in the spirit of Bangkok street stalls — he’ll get lucky. His phone will buzz with a second order from the same restaurant, to the same address. Dashers love these batch orders, as they can make twice the money without much more effort.

That sort of leverage is indicative of DoorDash’s grander plans. It’s a classic example of the on-demand company: building a vast service empire with a full-time staff that can easily be accommodated at an Il Fornaio dinner seating, and an army of contract workers that would barely squeeze into Levi’s Stadium. Though typically fueled by a feel-good mission (in DoorDash’s case it’s a desire to see local merchants thrive; Uber cites lofty ideals of a more evolved transportation policy), these enterprises see themselves not in terms of the prosaic tasks they perform for customers but more like utilities — human/machine infrastructure. The business, they claim, is not delivery but “logistics.” DoorDash, typical of this genre, believes that its real advantage lies in its software.

While Uber has taught us to associate on-demand with transportation, getting stuff to the front door — notably, food — is turning into the on-demand economy’s most competitive arena. (Spare me the analogies to dotcom-bust poster child Kozmo.com — that was pre-iPhone.) Xu and his founders face not only direct competition from earlier startups like Postmates and GrubHub, but from internet giants like Amazon and Google. Other services like Seamless make it easy for customers to get food from a wide range of restaurants, but don’t hire their own delivery people. Besides competition, obstacles include a lack of parking, routing challenges as knotty as the Fermi Paradox, and a class action lawsuit charging that DoorDash and similar companies are misclassifying and underpaying its contract workers.

None of this has stopped DoorDash from growing, growing, growing. It has spread to 18 markets and will add three more this month. (Most recent: Phoenix, Seattle, and Atlanta.) Later this year it plans to go international. Tony Xu and his team are on a dash to own the last mile.

Early in 2014, that last mile led to the front door of venture capitalist John Doerr. He’d ordered Howie’s Artisan Pizza. Apparently he liked the experience, because he soon was in contact with Xu, who impressed him immediately. In fact, the 31-year old founder reminded him of another guy he once funded, Jeff Bezos. Both, says Doerr, are similarly focused on great customer experience first, then building a profitable business. Doerr is now a DoorDash funder and a board member. Competition to own the last-mile doesn’t faze him. “The prize,” he says, “is really big.”

Tony XuXu’s own story, like that of many founders, begins an ocean away. He came with his parents from Nanjing, China, in 1989, when he was five, with only $300 in their pockets, but unlimited opportunity. Xu’s father had a choice of attending grad school in aeronautical engineering at either Stanford or the University of Illinois. The latter offered more money, so Xu grew up in Champaign, Illinois. While he eventually followed in his father’s grad school footsteps — neatly completing the circle by going to Stanford after getting his undergrad degree at Berkeley — Xu was equally inspired by his mother’s career in small business. “She owned a restaurant and now owns an acupuncture clinic,” he says. Early in 2013, Xu and some Stanford CS undergrads — Andy Fang and Stanley Tang — were deciding on what kind of company to start. (When you are a Stanford business or engineering student, the startup question is not whether but what.) They all seemed to share Tony’s passion to help small, local retailers. As they interviewed various proprietors, they often would notice clipboards of unfulfilled orders. “There were many pages of orders, not just one or two,” he says. Finally, when spotting this phenomenon at a Palo Alto macaroon store one day, Xu asked why these orders weren’t being delivered. The answer was that the owner had opened his bakery to bake, not deliver stuff. He got similar answers from people in other businesses, ranging from toys to furniture, to restaurants.

Stanley Tang, Co-founder and CPO, and Ding Zhou, Vice President of Engineering.(As Xu tells the story, he has to be reminded that there is a fourth founder, Evan Moore, who no longer is associated with DoorDash. Xu describes the split as a corporate version of an amicable divorce. Moore basically confirms this vague account, and now speaks of DoorDash with respect, though not much enthusiasm. [Update: Moore says that his reticence was simply a consequence of not wanting to speak for the company, and that he’s enthusiastic and supportive of his former company’s mission.] He is associated with a startup of his own, which, awkwardly enough, is located in the same San Francisco building as DoorDash’s new luxe HQ, a big upgrade from DoorDash’s previous digs at the aforementioned animal hospital.)

The foursome decided to explore whether a third party could make the deliveries that merchants didn’t want to bother with. The big question they had was whether people wanted such a service. While food delivery was common in places like Chicago or New York City, it was limited to pizza and a few Chinese restaurants in less urban locales like Palo Alto. Did that indicate a lack of desire?

One Saturday that February, they decided to find out. They hacked out a prototype web site in an hour. It included a choice of eight local restaurants, whose menus they uploaded in PDF format, without the knowledge of the owners. The actual orders would not be placed online but to a Google Voice phone number shared by the founders. They called their business Palo Alto Delivery, mainly because the Internet domain was available. The fact that they could secure that URL was proof that no one else was doing last-mile delivery in Palo Alto. The center of the New Economy was acting like the old economy.

That would change.

The founders didn’t expect anyone to call soon, as it might take Google several days to index the site. But 45 minutes after going live, their phones buzzed. “It wasn’t a friend — we didn’t know this person,” says Xu. “The only way this person could have found us was by typing paloaltodelivery.com.” The caller wanted some Pad Thai and spring rolls from Bangkok Cuisine in downtown Palo Alto. They quickly called the restaurant and placed the order for takeout. Then they hopped into Xu’s Honda, picked up the food, and delivered it to the customers. They collected payment on Xu’s Square account.

Within days, the service began to take off. “We effectively became Stanford’s delivery service,” says Xu. “We delivered every single day for six months, seven days in a row, every single week.” To learn more, in the next few months, they also took temporary jobs for existing delivery services. “We drove for Domino’s Pizza, FedEx, and 1–800-Courier, which is the largest partner to Google Shipping Express,” says Xu.

The experience was illuminating and gratifying. Stanley Tang recalls a delivery to a hospital to a woman who had just given birth; the soup DoorDash brought to her was her first meal as a mother. “If it weren’t for DoorDash, she wouldn’t have gotten that food,” he says.

Now convinced that there was a demand, they sought to find out whether the restaurants would actually kick back part of their revenues to be included. They contacted the eight emporiums on the app, and then spread out to others in the area. Finally they decided that, considering the limited peak times of food delivery, it made the most sense to use contract workers rather than employees. Xu says it was not to emulate Uber or other on-demand companies. “It was just a question of how can we pay per job and not have to buy all this infrastructure, to create a model in which we could sustain ourselves.”

Later that year, they joined the Y Combinator incubator and changed the name to DoorDash. They also zeroed in on a business model. The Dashers would keep the entire delivery fee (it varies by locality — it’s $6 in the Bay Area and a low $2 in New York) as well as the tip. DoorDash gets paid by the restaurants, at a range of percentages that the company chooses not to divulge. (One restaurant I spoke to said the fee was 20 percent of the food order; DoorDash confirms this is within their range.)

Now they had to build and scale their service.

Unlike in the early days of DoorDash, when Google Voice and PDFs were the tools, the current system is a labyrinthine combination of multiple apps and dynamic algorithms. There’s one app for thecustomer, one for the dasher, and another for the merchant, in a DoorDash-supplied iPad.

One particularly thorny problem is routing the Dashers. “ The media makes a huge deal about the contractors, etcetera, etcetera — but no one talks about how hard deliveries are to do,” says Xu.

Andy Fang, Co-founder and CTO and Tony Xu, Co-founder and CEO.Xu ticks off all the vexing problems in rapid-fire: “How do you determine which driver should get which order? How do you time it so there’s no wasted time at the restaurant? If you get there too late, the food is already cold. How do you determine whether or not more than one order should go to a driver? What happens when you don’t have enough drivers out on the road? How do you estimate how long it takes to make something?”

The solution involves variations of the Traveling Salesman Problem, a search for the shortest path through a complicated itinerary of multiple cities. It touches on the Routing Problem, an even knottier optimization problem, which postulates how a fleet of vehicles might best deliver from a central depot. And because DoorDash was delivering food from multiple locations— with tight windows on freshness and a hungry customer waiting on the other end — there are problems that mathematicians have yet to name. “To make one delivery successful there’s fifteen small things you have to get correct, which is shocking,” says Xu.

Xu and his colleagues got a taste of how complicated this was even when they were making all the deliveries themselves. “The first version of the algorithm was my brain,” he says. “And I was like, Wow, this is really hard.” Of course, as the company moved to hundreds of deliveries from dozens of restaurants in a given market, hard became an understatement. One of their Stanford profs told them, “The problem you guys have has been theorized before but never solved.” No one had even collected the data to try and solve it.

But that’s exactly what DoorDash set out to do: gather a massive amount of data and then use machine learning to iterate to the best possible way to estimate delivery and execute it in a timely, mistake-free fashion. “We had to build a tech company,” he says. “Because if we’d tried to build an operations company, like FedEx or even Amazon, we would literally die trying.”

Hiring a team of engineers poached from Apple, Google and, yes, Amazon, DoorDash began building a system it called Deep Red, an homage to the IBM system that beat Garry Kasparov. “We’re trying to get it to do for local logistics what Deep Blue did in chess,” says co-founder and CTO Andy Fang, 23. It collects data on a secret number of “predictors” that contribute to an estimate of how long it takes to deliver the food. It works with partner apps to take into account conditions like traffic and weather. It crunches menus and ready times to determine which dishes take longer to prepare. It takes note of the efficiency of different Dashers. Then it takes note of its performance and what might have been right and wrong. With every order, Xu says, DoorDash gets smarter. And its next delivery might be better.

Recently, DoorDash hired Ding Zhou as VP of engineering, whose resume includes high posts at Klout and Facebook; most recently he was a director of engineering at Square. (At 33, he’s kind of the old man of the company.) “I’m very surprised at the quality of the system — beyond my expectations,” he says. “Obviously this is a very hard problem so there’s a lot of work to do. But this is how we will differentiate ourselves — to use all the data you can think of to measure every step of the delivery, to find the balance between efficiency and customer quality.”

So far, however, it’s sometimes hard to see how much that learning is actually reaching the Dashers. In the past couple of months of testing DoorDash, I’ve made multiple food orders to my home. Delivery people in my building must appear at the service entrance — if they go to the front door, they get bounced and must go through a process that can add ten extra minutes to the delivery time. This mistake was made several times — and since there are hundreds of people in my building, that misstep must have been repeated many more times. Yet the app never managed to alert Dashers to this twist. (DoorDash told me that the system doesn’t handle such details about customer location, though it does for the restaurants.)

Still, DoorDash seems to be pleasing customers — and impressing investors. In the Spring of 2014, DoorDash raised $17 million, largely from Sequoia, whose star VC Alfred Lin joined the board. A year later they got $40 million more, in a round led by Kleiner Perkins; that’s when Doerr became a director. (This combination may be known as the Google Double, reflecting the most famous parlay of those two venture giants.) At that point, the company was valued at $600 million.

Of course, DoorDash’s competitors are not as impressed as its funders. Its most direct rival — the Uber to its Lyft, or vice versa — is Postmates, a service that launched in 2011, two years before Tony Xu’s Accord delivered its first spring roll. Thus CEO and co-founder Bastien Lehmann refers to DoorDash as a “copycat.” Postmates originally began as a way to deliver bulky items like furniture. But it quickly realized that food delivery was the ideal category for a force of contract workers. Now, seventy percent of the sorties taken by Postmates (a term that seems oddly centerfoldish) involve prepared food. Some workers fill orders for both services, but Postmates also has deals with Apple, Starbucks and Chipotle to fulfill orders tapped into the apps of those companies. It wants to deliver everything.

So does DoorDash. Its first big non-food partner is 7-Eleven. More announcements are to come. Xu’s dream is to become the go-to service for all local merchants. And even some national ones. Asked if he’d work with Costco, he doesn’t hesitate.

“We’re open to it, right?”says Xu. “In order to truly build a delivery network in a city you have to be able to connect all the merchants to all the consumers.”

Co-founder Tang, 23, who holds the title of Chief Product Officer, says that the company’s broader ambitions are actually baked into the app’s visual design. Tang spent part of his childhood in Japan and commuted to school by bullet train, an experience that fascinated him. “It’s very modern, it’s super fast, it’s very efficient — and always on time. It’s also a utility, part of the fabric of Japanese society. People rely on it every single day.” That’s why he says the DoorDash app doesn’t scream food, but tries to convey a sense of speed and efficiency. “When people get DoorDash, we don’t want them to think about pizza,” he says. “We want them to think of all the qualities and values that are associated with the Japanese bullet train.”

The bullet train may be slowed by a yellow light. On September 23 of this year, a Boston attorney named Shannon Liss-Riordan initiated DoorDash into a not-so-exclusive fraternity: on-demand companies sued in class action suits for allegedly misclassifying its workers as contractors. Because Dashers wear “uniforms” (well, red t-shirts), and because DoorDash specifies work standards and relies on its contractors for a key component of its business, they should be treated as employees, she charges, and paid for their expenses like gas and phone data charges. “There’s this whole trend of these companies that seem to think they can classify these employees as contractors and not have to worry about wages and benefits,” she says. “I think they’re skirting the law.”

DoorDash isn’t commenting beyond a bland statement: “While we will vigorously defend these lawsuits and we feel strongly that our business model is lawful, we recognize that the issues being raised are important ones. We are proud to have created opportunities for Dashers that offer them flexibility, freedom and a meaningful source of income.” Postmates’ Lehmann has a spicier take. “Everybody in this space is getting sued,” he says. “Unfortunately, there seems to be an extortion of money happening one way or another. When we launched there was no idea whether 1099 was good or bad. We hired people on contract because it was difficult to find reliable people when people want to eat.”

Pretty much all the companies being sued — including Uber and Lyft — believe that its contract workers are better off by being able to work when and how much they choose. DoorDash reports that most of its contractors use their Dashing fees as supplementary income, not their primary support. Take this away, the thinking goes, and there go the options for people to make this extra income at the hours they choose. These companies insist they’re creating jobs!

Liss-Riordan finds this risible. “Why should we say we like this? Because you’re creating jobs that aren’t really jobs?” she says. “I say good riddance to those companies who can’t afford to pay people.”

Ultimately, all of this might not matter. The 1099 controversy may resolve itself when Dashers and Postmates are replaced by drones (Google and Amazon are building them) and rolling robots. Just last week news broke of a company called Starship Technologies, launched by two co-founders of Skype. It does not make rockets, but an “autonomous delivery vehicle,” creeping along suburban streets at walking pace, designed to make drop-offs within 30 minutes for a cost of one dollar.

Will human delivery squads indeed be replaced by machines? “We think so,” says Xu. In fact DoorDash recently hosted a drone hackathon. “We invest in our version of moon shots,” he says. He’s already thinking about problems like getting hot food and other items on and off the robots, and what an autonomous delivery infrastructure might be like.

Ofcourse, Xu notes, that future is “years off.” Until that happens, the streets and sidewalks of every metro market will be jammed by workers who set their own hours and pay their own expenses to deliver food and various sundries. As Tony Xu himself does on the day I accompanied his Dash.

Here’s CEO Xu’s dash. Photos by Steven LevyWhile waiting for the orders in Asian Box that morning, two other Dashers show up, not unusual in Palo Alto. When Xu’s orders are ready he lifts the bulging bags to a table and painstakingly checks each item to make sure none are omitted. In this case the app gives a specific warning to double check that the free spring rolls are in the bag, apparently an artifact of previous lapses. Once everything is verified he hoists the food into insulated hot bags, and drives for seven minutes or so to a venture capital firm in downtown Palo Alto, where a receptionist meets him to take both orders.

It is 11:32 am. The whole process has taken around 45 minutes. Xu lets me view his haul as a Dasher. The two orders have cost $77 and $43 dollars. This results in a tip of $24.04. With the six dollar delivery fee, Xu has made $30.04, almost five times minimum wage, more than compensating for expenses incurred by operating the 2001 Accord.

Pretty good. But not nearly as good as the job Xu returns to after he takes off his red shirt — a full-time CEO who commands a workforce of 140. If you don’t count tens of thousands of contract workers.

Photographs by Jason Henry for Backchannel okCorrection: An earlier version inaccurately reported that GrubHub does not have its own delivery staff.*

This is one of a series of deep dives that illuminate issues concerning the future of the workplace. On November 12–13 in San Francisco, Medium and O’Reilly Media will co-host a conference on these issues called Next: Economy. You can find out more about it here____