Defending A16z: Noam Scheiber mistakes a VC portfolio for his 401K

Labor and workforce reporter for the NY Times Noam Scheiber takes on Andreessen Horowitz in this recent piece:

Rather than profiting like Mr. Ovitz and his fellow agents, the venture capitalists may be more like the Hollywood studios — chronically overpaying for projects whose costs they can rarely recoup. Mr. Andreessen and his partners have invested so much in so many start-ups that it would take a remarkable string of successes to make the approach pay off. For all their skill — the firm bought into the likes of Airbnb, Instagram and Pinterest relatively early — their track record suggests it’s unlikely. Already, they’ve suffered a few impressive flameouts, including Fab, on which they are likely to lose tens of millions of dollars.

Even when they pick well, they often bid so much for stars that the return is relatively modest. It’s easier to triple or quadruple your money when you’ve invested $10 million in a $100 million company than when you’ve invested nearly $100 million in a $1 billion company, as they did with the daily deal site Zulily. There are only so many companies that are acquired for billions of dollars or reach that kind of price through an initial public offering. Fewer retain such valuations — Zulily’s stock price has fallen sharply since last year.

Noam doesn't make a data-based argument. He uses an anecdote. Because data would ruin a really good story in this case. He compares Silicon Valley juggernaut Andreessen Horowitz to the excesses of Hollywood and CAA.

There's a big difference here. Let's take Avatar, for instance. It's a film that was made in 2009 for $237 million. It grossed over $2.7 billion worldwide — a roughly 11X return on capital. That's the highest grossing film ever made, and a good proxy for how profitable Hollywood can be at best. 

Let's take another example from Silicon Valley— Facebook. Peter Thiel invested $500,000 in the fledgling company in its first seed round in 2006, and from public records held 22.4 million shares of the stock at IPO. Those shares, if he hadn't sold them, would be worth $1.9B today (at about $74 per share). That's a 3,800X return on capital. 1 

Multi-billion dollar companies happen when non-obvious ideas and huge market needs meet perfect execution. We've seen it before our eyes — Uber, Airbnb, Dropbox, Stripe, Instacart — and when you have the potential for 100X to 4000X returns, it's not about avoiding loss or minimizing downside. A proper venture portfolio is not like your 401K. The only way startup investors truly lose is if they miss the Uber. 

And that, in a nutshell, is why using anecdotes (e.g. Zulily in Sheiber's piece above) as evidence against Andreessen Horowitz makes no sense at all. An individual investment may fail but it's just one in a portfolio. The returns that are possible in early stage technology investing far outweigh anything Hollywood has ever seen or ever will see. Software is eating the world, and the numbers bear it out. 

1 David Hammer suggested a better comparison would be Accel, their Series A partner. Their 10% stake at IPO is now worth $14.8B, so their $12M investment yielded roughly 1100X return.

Spotify as a simple case study in making something people want

A passage in the New Yorker long form piece this week on Spotify really illuminates the key parts of what we spend a lot of time trying to help founders create: Something people want. 

Problem: Piracy was how people did it at the time. 

“It came back to me constantly that Napster was such an amazing consumer experience, and I wanted to see if it could be a viable business,” Ek went on. “We said, ‘The problem with the music industry is piracy. Great consumer product, not a great business model. But you can’t beat technology. Technology always wins. But what if you can make a better product than piracy?’ ” Ek continued, “Piracy was kind of hard. It took a few minutes to download a song, it was kind of cumbersome, you had to worry about viruses. It’s not like people want to be pirates. They just want a great experience. So we started sketching what that would look like.”

They started with a simple problem statement. All of the great era-defining startups seem to start that way. What was broken? They knew exactly what was broken. 

Solution: 200 milliseconds

Their “product vision,” in tech parlance, was that the service had to give the impression that the music was already on your hard drive. “What would it feellike?” Ek asked. “That was the emotion we were trying to invoke.” The key was to build something that worked instantly. Streaming, whether audio or video, tends to have built-in delays while you wait for the file, which is stored on a server in the cloud. But if the music starts in two hundred milliseconds or less—about half the time it takes, on average, to blink—people don’t seem to perceive a delay. That became Ek’s design standard. He told his lead engineer, Ludvig Strigeus, a brilliant programmer he had worked with before, “I don’t accept anything that isn’t below two hundred milliseconds.”

What would solve that very direct problem? Well, a great streaming service that was superior in a specific way. They knew how to measure success. Clearly if you could make a streaming service act as if it were on your local hard drive, you could win. 

Hard-to-build proprietary technology — the first time it was done, just enough of an advantage

Strigeus responded, “It can’t be done. The Internet isn’t built like that.”

“You have to figure it out,” Ek insisted.

The solution involved designing a streaming protocol that worked faster than the standard one, as well as building their own peer-to-peer network, a decentralized architecture in which all the computers on it can communicate with one another. In four months, they had a working prototype.

“And I knew when we had it that it was going to be very special,” Ek said.

This is a signature piece of why Spotify was special early on. There are two parts to this. The first is could it be done? Until these early prototypes, it hadn't. The second part is can it be cloned? As with almost anything software related — yes. Novel tech is not an infinite defense, but it was enough of one for Spotify to get off the ground. You don't need a technological advantage that lasts forever (though of course that'd be preferable) — just one that will last enough such that your competitors can only copy your innovation from 6 months ago. It can take that long to fast follow, which is enough to keep your unoriginal would-be competitors at bay while you blaze ahead. 

The Schlep = The Moat

Ek’s original idea was to launch Spotify in the U.S. at the same time that he launched the service in Europe. Ken Parks, Spotify’s chief content officer, said, “Daniel thought he could just go down to the corner store in Stockholm and pick up a global license.” He didn’t realize that he would have to negotiate directly with all the different copyright holders, a herculean task. Not surprisingly, the labels weren’t interested. Ek was an outsider—a techie, and a Swedish one at that. Parks, an attorney who’d worked at E.M.I., recalled, “We needed to overcome the music-is-free mentality that Spotify represented.” Of the labels’ attitude, he went on, “If you have something you’ve invested a ton of money in, and you’ve been selling it for a lot, and you feel raped by piracy—to say to that person, ‘The only way to beat this is to co-opt the people who are stealing from you,’ that was a challenge.” Ek said, “If anyone had told me going into this that it would be three years of crashing my head against the wall, I wouldn’t have done it.”

This is a common refrain. If founders knew how much work these things would be, often they wouldn't do it. But they do, and that's protectable long term enterprise value. 

This analysis is of course with 20/20 hindsight. In the moment, Ek and the Spotify cofounders and investors had no idea they'd be right. But they figured they might be, so they built it. 

Why Flow, a new low cost super high precision controller, is important for designers and creatives

Flow is launching this morning. They're YC alums who have created a low cost, high precision wireless controller in the form of a dial. It's highly programmable and most designers and creatives will find this to be super valuable because that's where precision really matters.

When I'm in Photoshop making pixel-perfect mockups, or when I'm in Lightroom editing photos, I'm constantly making micro-adjustments on specific settings, whether it be brush size, exposure, etc. I have to acquire the target, then move my mouse, and then click-drag to the point where I'm happy. We're exercising one fundamental law of UX over and over again - Fitts' law. 

Fitts' Law states that the difficulty of an action is determined by the movement time needed to complete that action, which is in turn defined by the size of the target to be acquired. Sliders are by nature long and thin. If I had to guess, a good chunk of the cognitive load of doing creative work is just moving a mouse pointer to a tiny slider bar.

Not only is it a tiny target to acquire, but there are finite number of steps in those sliders that can make a mountain of difference. For instance, photographers are always looking for that absolutely perfect exposure or temperature. With a slider, you're limited to the number of pixels that slider has on screen - 200px? That's only 200 gradations, and in my experience that perfect level is always in between two of those notches. 

Enter Flow. There are over 3600 distinct values in one full 360 degree turn of the device. And since you can link them directly to specific values e.g. exposure or brush size, you don't have to acquire the target over and over and over again. 

That's why I bought one, and that's why Flow is an important programmable hardware device that creative people should keep an eye on. They're accepting preorders now and are on Product Hunt, and if you get one early it's an extra good deal. 

Silicon Valley's Congressional Hope: Time to sweep away a congressman "who mostly votes the right way"

Silicon Valley hasn't had a real voice in Congress before. We've got a shot at one now though. In about 20 days, voters will go to the polls in a race for a congressional seat for the 17th district, and democrat Ro Khanna has a fighting shot at toppling incumbent Mike Honda. The polls are running in a dead heat, so this is one you should care about. 

There's nothing wrong with Honda — he's an old-line Democrat. Except that's actually the problem. He hasn't done much in the way of defending the things we really care about: Immigration reform, free Internet rights, supporting entrepreneurship, and reforming education. If Silicon Valley can elect Ro, then we as citizens are making a statement that business as usual for the Democratic party just isn't going to fly. 

That's why the San Jose Mercury News has endorsed Ro Khanna as well, saying "Silicon Valley -- whose economy, like the 17th District, stretches into the East Bay -- needs more than a congressman who mostly votes the right way... Silicon Valley's other representatives, Congresswomen Anna Eshoo, D-Palo Alto, and Zoe Lofgren, D-San Jose, are older than 65, and both are invaluable voices in Washington -- respected leaders on valley issues as well as defenders of progressive values. When they meet with us, they are insightful; we always learn something. This is not the case with Honda."

Ro is one of us. He's committed to reforming immigration so our talented friends who happened to be born elsewhere can still come here to create new businesses and jobs. Startups die every other day because of our antiquated and special-interest-ridden immigration policies. He's on our side when it comes to SOPA, PIPA, net neutrality and a maintaining a free and open Internet. 

If you're in the 17th District (Fremont to Sunnyvale), you have a chance to make history. Register to vote, and consider Ro Khanna for Congress. This race matters, and it's looking like a few hundred votes will swing this one way or another. 

Listen to the people for whom you're building: A low income housing development that chose bath tubs over hot water

I was watching an urban planning documentary called Urbanized on Netflix recently. It brought up a fascinating example of participatory design in the context of building homes. A housing development in Santiago, Chile called Elemental ran into a problem. The builders had to make a tough decision: should they build in hot water heaters, or should they build in bath tubs? The budget could only support one or the other at the outset.

A typical top-down approach would dictate that of course you'd want hot water. A first world view of the situation would say that you'd rather shower standing up with hot water than sit in a bathtub and have to heat water separately. 

Yet that's the exact opposite of what future residents of Elemental actually wanted. Architect Alejandro Aravena went out into communities and talked with residents and discovered what typical bureaucrats would never find - that people moving to the low income housing from slums would unanimously choose bathtubs instead. Hot water heaters and gas furnaces cost money, and are unfamiliar. Bathtubs, on the other hand, were very familiar (in fact what residents typically did in their existing living environments due to the extra privacy) and didn't generate additional energy cost. 

Further, hot water was one of the things that people typically added later, once they had acclimatized to the new living environment and improved their station in life. 

It seemed to me this was the sort of thing you could only tell by actually talking with the people who would use your creations. It is the ideal situation for us to create things for ourselves. But when you aren't doing that, you have to be extra careful about the assumptions and values you bring to the table. 

Always-on video recording will prevent tomorrow's Missouri police state — light is a disinfectant

The ongoing Missouri police fiasco has prompted a lot of discussion about the need for police transparency. My friend Jeff Lonsdale writes:

Active duty police officers need to be automatically recording everything they do. With recordings, incidents such as those happening in Ferguson can be quickly resolved one way or another. When tested in Rialto, California, recording reduced both complaints filed against police officers and the incidents where use of force was required. There will still be cases where police officers use excessive force in murky situations but by and large transparancy via recorded police and citizen interactions should protect the innocent parties, see more guilty parties punished and cause better behavior all around.

Smartphones, wearables, and always-on high bandwidth connectivity is converging in the next ten years to make this happen, not just for police officers, but for private citizens too. 

Always-on video is already shining light in Russia with the omnipresence of on-dash cameras. They say light is the best disinfectant, and that's exactly what video can be in the future. It's time to build. 

Software eats apparel — What MTailor means for how you'll buy clothing in the future

Miles and Rafi are two founders in the current YC batch who just launched their new startup, MTailor. It's a iPhone and iPad app that lets you get accurate measurements of your body so you can order made-to-measure dress shirts that fit you perfectly. You put the iPhone or iPad on the floor at an angle, and the app walks you through the 30 second process of turning around in place in front of the front-facing camera. The amazing thing is that the entire process is 20% more accurate than what a professional tailor would do in person. 

That's pretty damn cool. The founders are Stanford CS and math grads and devised and perfected the computer vision algorithms themselves. In the past similar founders would have tried to find some way to license the tech, but MTailor is building a new brand from scratch. It's an ambitious way for the company to fully create as much value as possible without middlemen and enterprise sales. 

Mass customization is finally hitting the mainstream, thanks to software. In the past it has always been difficult and time consuming for clothing to be made with specific measurements. It was a tedious process to begin with since it was hard for people to get accurate measurements, and even experienced tailors have trouble getting it right. Add a whole lot of waiting to the mix too — it'd take six weeks or more to wait for a shirt to be made. Those two problems taken together reduce demand. And as a result, the cost goes up, even further reducing demand. 

That's where a new software capability can come in, like MTailor's computer vision algorithms, and radically change the equation. Since it's easy to measure yourself, the major stumbling block is removed. And with a steady stream of orders, they can bring the price and wait down. That's exactly what the team has done, coming in at the $69 price point and a 2 week lead time. I was once a die-hard made-to-measure Indochino dress shirt fan, but I know where all my future purchases are coming from. 

That's the power of better, cheaper, faster, and how software eats the giant apparel business. Keep an eye on MTailor and try download their app. It's available right now. 

How a bug in Windows might be costing humanity over 600 years of wasted time per day

There's a Windows bug that I have been hitting for over a year that has driven me insane. My Downloads folder takes 20 seconds to show up in my File Explorer in Windows 8. To fix, I had to set the folder's type to "General Items" instead of "Photos" because Windows was trying to generate thumbnails over and over again. 

A program manager at Microsoft made the wrong decision. Microsoft has been a program manager culture for decades now. (Former president of Windows division Steven Sinofsky writes at length about it here.) PM's at Microsoft have the cat-herding role of trying to make things happen with little authority. As with most things that are started with very good intentions, there are unintended consequences. Bugs like the one above are one of them.

How did a bug like this stay in Windows for years? It's not like they didn't know about it. Someone on the team hit this problem and filed it as a bug in their bug DB. There are a lot of really dedicated software engineers and software testers out there who really do care about building great stuff at Microsoft. But it is the PM who has the unenviable task of sorting through hundreds if not thousands of these types of bugs and figuring out what gets fixed and what doesn't. On the day in question, the PM of the File Explorer probably saw this alongside 50 other bugs that were must-fix. They had a specific date to ship, and a fixed number of engineering resources. They knew their bug count had to glide down to zero by X date, and some bugs had to sadly be resolved "Won't fix." If a PM doesn't hit their schedule and ship, they get fired. This bug had to be punted. 

Say a billion users hit this bug and lose 20 seconds every day -- that's 633 years of human life wasted in the world. Normal people using their computers have no idea why this bug is happening, and just assume the computer is thinking. These wrong calls on small decisions have wide reaching impact when you multiply by the reach of software today.  

Great software is built by people making the right call on thousands of small decisions. We who create software have a responsibility to our users. Make the right choice. Fix the damn bug. And if you run a company that builds software, know better than to put a good person in a role that forces them to make the wrong decision. 

What Instacart taught us: Don't be a dreamer, be a creator

YC S12 company Instacart is well on its way to being one of the iconic companies of our times. How it started has been covered before: Instacart founder Apoorva Mehta hacked his way into YC by sending me a six pack of beer. It was a cold and refreshing pack of 21st Amendment's Back in Black, as pictured above. Apoorva sent it over using his own service two months after the deadline for applications at YC. The hack wasn't the thing that mattered, really, though. What mattered is what Apoorva had already built something amazing. 

I had already seen dozens of startups say they were going to do grocery delivery. But we had never actually seen someone build an app with thousands of products in it, with a few real drivers using the app to deliver real groceries every day. I downloaded the app that afternoon and was blown away at how much had already been done. That was absolutely remarkable. Everyone else was dreaming. Apoorva created it. And then everyone wanted it. 

That's the key lesson here. Don't be a dreamer. Manifest that dream. Build it. Write the code. Spin up the service. Hire. Create.

Living in a paperwork-less future

"Once you're living in the future in some respect, the way to notice startup ideas is to look for things that seem to be missing” —Paul Graham

"The future is already here — it's just not very evenly distributed." – William Gibson

Imagine: it is the near future. A mortgage application is a one-tap action on your smartphone. So is renewal of your driver’s license with the DMV. Paying your last parking ticket is a just one tap too. A car loan? A visa application? Life insurance? Car insurance? An estate plan? All a few taps. 

The promise of this has been a long time in the making. The next 20 years is not going to be like the last 20 years, however. Now's the time. Why? Because of a number of overwhelming new platform forces: 

Smartphone adoption — Late adopter industries are being unlocked because now far more people on the planet can have an always-on, cheap, ubiquitous computer. Android devices are $150 with no contract, today. A cheap tablet or smartphone can replace kiosks, point of sale devices, and PC’s on desks. You'd be crazy to say construction would be an industry where you grow a software business even 5 years ago, but that's exactly what Plangrid is doing bringing blueprints online. 

Digital signing — HelloSign is at the forefront of this. Instacart recently used the HelloSign API to be able to onboard their contract delivery workers far faster than ever. If you can sign legal documents instantly, you bring the friction of getting business down to nearly zero. 

Payments systems — StripeCrowdtilt OpenBalanced and WePay are at the forefront of this. It’s never been easier for a startup to stand on the shoulders of these giants and be able to do real transactions instantly. Coinbase does the same allowing anyone to receive payment using bitcoin, unlocking worldwide commerce with a universal currency. These things just didn’t exist even a few years ago. 

Legacy API's — Any system must work with all legacy systems. They’re getting filled in by both platform API services like Lob, Twilio, Plivo, and HelloFax. Not everything in the stack has to live in the future. 

Anti-fraud — Services like Sift Science are bringing world-class machine learning and anti-fraud technologies to every site and service. Online verification and personal data room services are coming online that will let people do large transactions with people they just met. Airbnb was able to use Facebook and Twitter data to verify people and make them more human, such that you could do business with them. LendUp is doing the same for payday lending, adding more data to the antiquated to the scammy world of payday lending. 

Killing paper-based purchase orders, application forms, approval queues and replacing them with all-digital direct-to-consumer experiences is happening right now, and in every walk of life. 

In the 1980's, Walmart started using IT to start working directly with manufacturers and cut waste out of their supply chain. Today, the company represents 2% of annual US GDP. That same thing is finally coming to all the other industries, and they're being brought to the world by small teams of talented founders building on the fundamental platforms mentioned above. 

New services are better, cheaper, and faster. As a result, these new businesses will be built shockingly fast. It’s not twilight for young startups trying to make something great. It is dawn.