Curated Insights 2018.08.17

Not enough people are paying attention to this economic trend

Haskel and Westlake outline four reasons why intangible investment behaves differently:

  • It’s a sunk cost. If your investment doesn’t pan out, you don’t have physical assets like machinery that you can sell off to recoup some of your money.
  • It tends to create spillovers that can be taken advantage of by rival companies. Uber’s biggest strength is its network of drivers, but it’s not uncommon to meet an Uber driver who also picks up rides for Lyft.
  • It’s more scalable than a physical asset. After the initial expense of the first unit, products can be replicated ad infinitum for next to nothing.
  • It’s more likely to have valuable synergies with other intangible assets. Haskel and Westlake use the iPod as an example: it combined Apple’s MP3 protocol, miniaturized hard disk design, design skills, and licensing agreements with record labels.

For example, the tools many countries use to measure intangible assets are behind the times, so they’re getting an incomplete picture of the economy. The U.S. didn’t include software in GDP calculations until 1999. Even today, GDP doesn’t count investment in things like market research, branding, and training—intangible assets that companies are spending huge amounts of money on.


How Box conquered the enterprise and became a $1.7 billion company in a decade

However, what most people failed to understand—and continue to misunderstand to this day—is that Dropbox was never launched as a competitor to Box. The use cases were completely different. Box.net and Dropbox may have shared some similar underlying technologies (and an uncomfortably similar name), but the focus of Dropbox was cloud-based file management for the consumer market. Box was focused on file sharing. By the time Dropbox launched in 2007, Box.net had already largely abandoned the consumer market in favor of the enterprise. There were other key differences between the two products, such as the necessity of installing a dedicated Dropbox directory on a user’s local machine versus Box.net’s entirely cloud-based interface. Additionally, the two companies’ target markets and business models couldn’t have been more different.

Levie knew SharePoint was Box’s biggest competitor, so he did what any inventive, irreverent entrepreneur would do—he took out a billboard advertisement on a stretch of highway on Route 101 between San Francisco and Silicon Valley. The ad promised SharePoint users that Box would pay for three months of SharePoint access if they didn’t prefer Box. In February 2009, Box went one step further in its media assault on Microsoft by erecting another billboard, this one highlighting the many aspects of SharePoint that were most unpopular among its user base.

While the enterprise market represented a unique chance for Box to pivot away from the increasingly competitive consumer market, essentially shifting the focus of the entire company was no small undertaking. Until that point, Box had used a freemium business model. This worked fine for the consumer market, but it was completely unsuitable for the enterprise. This meant Box would not only have to radically redesign its product from the ground up but also restructure its entire business model.

By acquiring Increo, Box immediately gained access to Increo’s innovative document collaboration tools. This was crucial. It wasn’t enough for Box to offer cloud-based storage or integrations with Salesforce and Office. It had to offer additional value as competing tools vied for dominance.

The consumerization of enterprise IT driven by Box and other forward-thinking companies wasn’t merely an attempt to cultivate a unique value proposition or drive adoption. It reflected much broader shifts in computing in general. The advent of Web 2.0 apps created a new design paradigm that placed emphasis on ease of use and accessibility across multiple devices over complex file management tools. Smartphones fundamentally changed the way we think of computing. For an enterprise software company like Box to be at the forefront of trends in usability was impressive.

OneCloud was an excellent example of how consumer-focused design informed Box’s broader strategy. The company had built a platform for developers in 2011 known as the Box Innovation Network, which functioned similarly to an app marketplace. OneCloud was an extension of this idea, only it was intended exclusively for mobile devices. This would later become a predictable cycle in Box’s development. New features were added to the product to meet emerging needs, and those features were presented to users in ways that directly mirrored those of consumer apps and sites.

What’s more important, however, is how well Box converted its free users to paid subscribers. Consumer apps like Evernote convert free users to paid plans at a rate of approximately 3%. Box was converting free users to paid plans at a rate closer to 8%, including major corporate customers such as Bank of New York and ambient advertising powerhouse Clear Channel. As a result, Box achieved revenues of more than $11M in 2011.

Because most of Box’s sales calls came from companies that had already been using the product, Box’s sales teams were typically able to close 60% of those deals within two weeks—an impressive figure, especially considering the often months-long sales cycles typically associated with the enterprise market.

Box has done an excellent job of not only carving out its own niche in an increasingly competitive space but also by applying design and UX principles of consumer-focused SaaS products to redefine how enterprise software looks, feels, and works. With its keen focus on usability, ease, and simplicity, Box has become a leading force in the consumerization of the enterprise and has shaped how other enterprise software companies approach their products.

Ad tech firm poised to surge 50%

Bid factoring is essentially a linear equation that enables marketers to apply multipliers to different targeting parameters. This approach makes it easier to value each user individually and dynamically, allowing marketers to more easily reach their target users. Bid factoring saved time for marketers through automation and removed the need to store tons of line item permutations, therefore lowering data storage costs.

When Green started The Trade Desk, his goal was to “build a company for the next 100 years.” He did not want to follow the same mistakes that other companies in the space made such as having a conflict of interest by being on both the buy and sell side. Green decided to build a demand side platform because he believed the demand side of the advertising transaction will always have the advantage. In advertising it will always be a buyer’s market because it is easy to add supply by having an extra impression on a web page or additional 30-second spot to a commercial break to meet increased demand. This basic economic reality means advertising supply is more elastic than demand and will forever put the buy side in the power position.

The Trade Desk would also be transparent and not charge unsustainable take rates. Green believed once the digital advertising industry matures, total transaction costs to purchase a digital ad would be $0.20-$0.30 for every $1.00 spent, with roughly $0.15-$0.20 going to the DSP and $0.05-$0.10 being split between the SSP and the ad exchange. The Trade Desk could have charged much higher take rates but decided to charge customers what it believed would be the fair end-state price for their services. While take rates could become lower as competition potentially increases, similar to what happened with discount stock brokerages, barriers to entry and the DSP’s ability to provide increasing value to advertisers overtime should preserve prices.

As the ad market has grown, the number of auctions has increased exponentially. In order for a DSP to win an auction, it now takes many more looks. For each ad campaign, costs have increased while revenues remained fairly flat, increasing operating leverage. DSPs that have half the ad spend as The Trade Desk will struggle because they will incur the same amount of expense per ad campaign but monetize less, making it much more difficult to be profitable if you are a smaller player and don’t have the scale.

Every day The Trade Desk’s customers log into their platform to use the data and analysis to value ad inventory and run marketing campaigns. Advertisers provide their customer data and publishers provide their user data, which The Trade Desk uses to help advertisers value media for their specific needs. As The Trade Desk accumulates more data over time, its insight and analysis add more value to its customers, creating a self-reinforcing virtuous cycle.


Nvidia’s new Turing architecture is all about real-time ray tracing and AI

Nvidia describes the new Turing architecture as “the greatest leap since the invention of the CUDA GPU in 2006.”

“Hybrid rendering will change the industry, opening up amazing possibilities that enhance our lives with more beautiful designs, richer entertainment and more interactive experiences,” said Nvidia CEO Jensen Huang. “The arrival of real-time ray tracing is the Holy Grail of our industry.”

The new RT cores can accelerate ray tracing by up to 25 times compared to Nvidia’s Pascal architecture, and Nvidia claims 10 GigaRays a second for the maximum performance.

With NGX, Nvidia today also launched a new platform that aims to bring AI into the graphics pipelines. “NGX technology brings capabilities such as taking a standard camera feed and creating super slow motion like you’d get from a $100,000+ specialized camera,” the company explains, and also notes that filmmakers could use this technology to easily remove wires from photographs or replace missing pixels with the right background.


Tesla’s autonomous opportunity is severely underappreciated

We estimate that net revenue for autonomous platform providers – those companies that own the software technology stack for autonomous ride-hailing services – should exceed $2 trillion by 2030, roughly equal to our expectations for automaker revenue at that time. Unlike their auto-manufacturing peers, however, autonomous platform providers should see software-like margins, be less capital-intensive, and enjoy network-effect-driven regional competitive dominance. So, while autonomous platform providers may generate the same revenue as automotive manufacturers, ARK believes these providers will generate six times the operating earnings and consequently will prove to be substantially more valuable. In fact, ARK estimates autonomous platforms will be worth more than the entire $4 trillion global energy sector.

An enhanced Autopilot package with the ability to self-drive costs $5,000 upfront or $6,000 for customers who choose to wait and buy later. Payment for this feature alone can be thought of as nearly pure profit on every Tesla sold. In addition, once Tesla launches the Tesla Network, its autonomous ride-hailing network, it could collect platform fees, similar to Uber’s model today, from every autonomous ride charged to the consumer. Given a rate of $1 per mile to the end consumer and over 100,000 miles per year per vehicle, Tesla could benefit from $20,000 in high-margin platform fees per car per year. Over a five-year lifetime, a single Model 3 could generate $40,000 in net cash flow. Even investors optimistic about Tesla’s prospects project the Model 3 cash flow at $4,000 and one-time in nature. In effect, each Model 3 sale could generate 10 times more cash flow than investors currently understand.

Google’s targeted ads are coming to a billboard near you

Digital outdoor ad spending is growing at 15 percent annually, and will overtake traditional outdoor outlays by 2020, according to PwC. But Google is the 800-pound gorilla that’s not yet in the room. It would give the company another major edge over Facebook, which doesn’t have the same access to location-based mobile data.


Alibaba tweaks a controversial legal structure

There are three problems with VIEs. First, key-man risk. If the people with nominal title die, divorce or disappear, it is not certain that their heirs and successors can be bound to follow the same contracts. Second, it is not clear if the structure is even legal. China’s courts have set few reliable precedents on VIEs and the official position is one of toleration rather than approval. Third, VIEs allow China’s leading tech firms to be listed abroad, preventing mainlanders from easily owning their shares and participating in their success.

Alibaba’s proposed change is aimed at tackling the first problem, key-man risk. At the moment four of its five VIEs are nominally owned by Jack Ma, the firm’s leader, and Simon Xie, a co-founder and former employee. After the restructuring, the two men will no longer be the dominant counterparties. Instead the VIEs will be owned by two layers of holding companies, which will sign contracts with Alibaba. These holding companies will ultimately be nominally owned by a broader group of Alibaba’s senior Chinese staff. The idea is that if anyone gets run over by a bus, then the scheme will not be disrupted, because nominal control is spread among a wider group of people. The new approach is far from perfect but it is an improvement. If all goes to plan it will be completed by 2019. Other tech firms may feel pressure to follow.

$1b+ market map: The world’s 260 unicorn companies in one infographic
60+ startups disrupting IKEA in one market map

SoftBank’s Son says WeWork is his ‘next Alibaba’

It is rare for Son, who casts a wide net with his startup investments, to commit so much resources to a single company. But he said WeWork is more than just a renter of office space: it is “something completely new that uses technology to build and network communities.”

The use of shared space to forge connections is not unique to WeWork. The company’s edge lies in the steady flow of data it collects on members, which is shared with other locations and can be accessed by users of the WeWork app around the world. The idea is that more data means more innovation — a model that underlies Son’s excitement about the company.

What MoviePass can teach us about the future of subscription businesses

Pricing is so powerful that playing with it requires great skill and precision. MoviePass should have done its price experimentation at the outset and on a local basis. It could have optimized the price points and tested alternative pricing models quietly, instead of jerking millions of customers around. Even a slight tweak — such as moving to a club pricing model like Costco’s — might have solved its cash-flow problems.

These kinds of tweaks could also have enabled the company to consider regional pricing strategies, given that its cost of goods (the full price of movie tickets, which it pays theater operators) varies from $8 in Nebraska to over $15 in New York. This case is also a good reminder that the United States has local profit pools. It is silly to think that a one-size-fits-all national strategy is the right approach for a market as ethnically and economically diverse as the United States.

MoviePass failed to recognize how the behavior of superconsumers, customers who are highly engaged with a category and a brand, differs from that of average consumers — and how, if not anticipated, this difference can create problems for a company’s cost model. It can especially be a problem if the company uses a “buffet” model of fixed price and unlimited quantities, as MoviePass did.

Quantum computers today aren’t very useful. That could change

Quantum computers are, however, far more prone to errors than binary machines. Instead of using electric signals to generate a series of zeros and ones like a conventional computer, quantum computers rely on the real-world, mechanical behavior of photons, which are packets of microwave energy. The machines require a complex, multi-layered refrigeration process that brings quantum chips to a temperature just above absolute zero. By eliminating certain particles and other potential interference, the remaining photons are used to solve computational problems. The true magic of this system is how photons can become entangled and produce different but related results. Scientists only partially understand why it works the way it does.

A quantum chip doesn’t look like much with the naked eye. Through an optical microscope, though, you can see the quantum logic gate that makes everything possible. The team here is working on a process of stringing together 16-qubit chips to execute on the 128-qubit design. Essential to this is a new kind of quantum chip that communicates results in three dimensions instead of the current two, which allows Rigetti to fit the chips together like puzzle pieces and turn them into a single, more powerful computer. “What we’re working on next is something that can be scaled and tiled indefinitely,” Bestwick said.

Why the future belongs to ‘challenge-driven leaders’

The consensus view of Mr. Marchionne, relayed by hundreds of tributes, is that he possessed an unusual blend of vision, technical expertise, analytical rigor, open-mindedness and candor. The remembrances also agreed on something else: he was a bona fide eccentric. “God bless you, Sergio,” Morgan Stanley analyst Adam Jonas told Mr. Marchionne during a January conference call. “We’re never going to see anyone like you again.”

The trajectory of great ideas

“Being right is the enemy of staying right because it leads you to forget the way the world works.” – Jason Zweig. Buddhism has a concept called beginner’s mind, which is an active openness to trying new things and studying new ideas, unburdened by past preconceptions, like a beginner would. Knowing you have a competitive advantage is often the enemy of beginner’s mind, because doing well reduces the incentive to explore other ideas, especially when those ideas conflict with your proven strategy. Which is dangerous. Being locked into a single view is fatal in an economy where reversion to the mean and competition constantly dismantles old strategies.

Survivorship bias on wheels

One last thing: When it was introduced as new in 1984, the 1985 Testarossa listed for $90,000 (but dealers charged huge premiums over list due to “Ferrari fever.”) You can still find Testarossas for that original list price — meaning the net returns over 43 years has been precisely zero — before maintenance, storage and repair costs.

As a comparison, in 1985, the benchmark S&P500 was about 200, and it closed yesterday at 2,821.93. That generated an average annual return of about 8.5%, returning 1,400% price appreciation since then, and, with dividends reinvested, over 3,000% total return (in nominal terms, like the chart above, neither is adjusted for inflation).

Selecting investments after the fact is easy; ask yourself this question: What car do you want to buy as an investment for the next 34 years to be sold in 2052?


Curated Insights 2018.05.20

The spectacular power of Big Lens

There is a good chance, meanwhile, that your frames are made by Luxottica, an Italian company with an unparalleled combination of factories, designer labels and retail outlets. Luxottica pioneered the use of luxury brands in the optical business, and one of the many powerful functions of names such as Ray-Ban (which is owned by Luxottica) or Vogue (which is owned by Luxottica) or Prada (whose glasses are made by Luxottica) or Oliver Peoples (which is owned by Luxottica) or high-street outlets such as LensCrafters, the largest optical retailer in the US (which is owned by Luxottica), or John Lewis Opticians in the UK (which is run by Luxottica), or Sunglass Hut (which is owned by Luxottica) is to make the marketplace feel more varied than it actually is.

Now they are becoming one. On 1 March, regulators in the EU and the US gave permission for the world’s largest optical companies to form a single corporation, which will be known as EssilorLuxottica. The new firm will not technically be a monopoly: Essilor currently has around 45% of the prescription lenses market, and Luxottica 25% of the frames. But in seven centuries of spectacles, there has never been anything like it. The new entity will be worth around $50bn (£37bn), sell close to a billion pairs of lenses and frames every year, and have a workforce of more than 140,000 people. EssilorLuxottica intends to dominate what its executives call “the visual experience” for decades to come.

For a long time, scientists thought myopia was primarily determined by our genes. But about 10 years ago, it became clear that the way children were growing up was harming their eyesight, too. The effect is starkest in east Asia, where myopia has always been more common, but the rate of increase has been uniform, more or less, across the world. In the 1950s, between 10% and 20% of Chinese people were shortsighted. Now, among teenagers and young adults, the proportion is more like 90%. In Seoul, 95% of 19-year-old men are myopic, many of them severely, and at risk of blindness later in life.

Del Vecchio paid $645m (£476m) for Ray-Ban. During the negotiations, he promised to protect thousands of jobs at four factories in the US and Ireland. Three months later, he closed the plants and shifted production to China and Italy. Over the next year and a half, Luxottica withdrew Ray-Ban from 13,000 retail outlets, hiked their prices and radically improved the quality: increasing the layers of lacquer on a pair of Wayfarers from two to 31. In 2004, to the disbelief of many of his subordinates, del Vecchio decided that Ray-Ban, which had been invented for American pilots in the 1930s, should branch out from sunglasses into optical lenses, too. “A lot of us were sceptical. Really? Ray. Ban. Banning rays from the sun?” the former manager said. “But he was right.” Ray-Ban is now the most valuable optical brand in the world. It generates more than $2bn (£1.5bn) in sales for Luxottica each year, and is thought to account for as much as 40% of its profits.

The Moat Map

Facebook has completely internalized its network and commoditized its content supplier base, and has no motivation to, for example, share its advertising proceeds. Google similarly has internalized its network effects and commoditized its supplier base; however, given that its supply is from 3rd parties, the company does have more of a motivation to sustain those third parties (this helps explain, for example, why Google’s off-sites advertising products have always been far superior to Facebook’s).

Netflix and Amazon’s network effects are partially internalized and partially externalized, and similarly, both have differentiated suppliers that remain very much subordinate to the Amazon and Netflix customer relationship.

Apple and Microsoft, meanwhile, have the most differentiated suppliers on their platform, which makes sense given that both depend on largely externalized network effects. “Must-have” apps ultimately accrue to the platform’s benefit.

Apple’s developer ecosystem is plenty strong enough to allow the company’s product chops to come to the fore. I continue to believe, though, that Apple’s moat could be even deeper had the company considered the above Moat Map: the network effects of a platform like iOS are mostly externalized, which means that highly differentiated suppliers are the best means to deepen the moat; unfortunately Apple for too long didn’t allow for suitable business models.

Uber’s suppliers are completely commoditized. This might seem like a good thing! The problem, though, is that Uber’s network effects are completely externalized: drivers come on to the platform to serve riders, which in turn makes the network more attractive to riders. This leaves Uber outside the Moat Map. The result is that Uber’s position is very difficult to defend; it is easier to imagine a successful company that has internalized large parts of its network (by owning its own fleet, for example), or done more to differentiate its suppliers. The company may very well succeed thanks to the power from owning the customer relationship, but it will be a slog.

How much would you pay to keep using Google?

Part of the problem is that GDP as a measure only takes into account goods and services that people pay money for. Internet firms like Google and Facebook do not charge consumers for access, which means that national-income statistics will underestimate how much consumers have benefitted from their rise.

Survey respondents said that they would have to be paid $3,600 to give up internet maps for a year, and $8,400 to give up e-mail. Search engines appear to be especially valuable: consumers surveyed said that they would have to be paid $17,500 to forgo their use for a year.


There is another

Spotify has better technology, merchandising (like discovery playlists), and brand. Unlike Apple Music, being a pure-play (as opposed to being owned by a tech giant) gives Spotify more cred among purists, young people, and influencers. The instinct / T Algorithm cocktail has resulted in a firm with 170M users, 75M of whom are premium subscribers. The firm registered €1B this quarter, representing 37% growth. Spotify accounted for 36% of premium music subscribers globally.

What takes Spotify to $300B, and true horseman status? They launch video, and become the most successful streaming entertainment firm, full stop. Netflix’s legacy is on the second most important screen, TV. Spotify was raised on the most important – mobile. Netflix needs to become Spotify before Spotify becomes Netflix. Nobody has cracked social and TV, and as half of young people no longer watch cable TV, if Spotify were to launch video and captured any reasonable share and engagement via unique playlists, then cable and Netflix would begin ceding market cap to Spotify.


Subscriptions for the 1%

The problem with these minuscule conversion rates is that it dramatically raises the cost of acquiring a customer (CAC). When only 1% of people convert, it concentrates all of that sales and marketing spend on a very small sliver of customers. That forces subscription prices to rise so that the CAC:LTV ratios make rational sense. Before you know it, what once might have been $1 a month by 20% of a site’s audience is now $20 a month for the 1%.

There is a class of exceptions around Netflix, Spotify, and Amazon Prime. Spotify, for instance, had 170 million monthly actives in the first quarter this year, and 75 million of those are paid, for an implied conversion of 44%. What’s unique about these products — and why they shouldn’t be used as an example — is that they own the entirety of a content domain. Netflix owns video and Spotify owns music in a way that the New York Times can never hope to own news or your podcast app developer can never hope to own the audio content market.

The Apple Services machine

It is this hardware dependency that makes it impossible to look at Apple Services as a stand-alone business. The Services narrative isn’t compelling if it excludes Apple hardware from the equation. Apple’s future isn’t about selling services. Rather, it’s about developing tools for people. These tools will consist of a combination of hardware, software, and services.

Apple currently has more than 270 million paid subscriptions across its services, up over 100 million year-over-year. Apple is in a good position to benefit from growing momentum for video streaming services including Netflix, HBO, and Hulu. It is not a stretch to claim that Apple will one day have 500 million paid subscriptions across its services. Apple isn’t becoming a services company. Instead, Apple is building a leading paid content distribution platform.

Tencent Holdings Ltd. delivered two major milestones when it reported its earnings Wednesday: record quarterly profits and more than one billion monthly active users on its WeChat platform. The social media and gaming giant, which has been leery of barraging its users with ads, also declared it had raised the maximum number of ads that customers see on WeChat Moments from one a day to two. The app has become China’s most popular messaging service and is integral to driving everything from gaming and payments to advertising for Tencent.

MoviePass: the unicorn that jumped into Wall Street too soon

“The growth-at-all-costs strategy is being funded these days by the venture community, not the public market. The last time we saw the public markets fund a growth-at-all-costs strategy was the 1999 internet bubble, and we all know how that ended.”

The prospect of steep declines in a company’s valuations once it hits the public markets is one reason why U.S. companies are waiting longer to go public. Overall, U.S. companies that have gone public this year have done so at an average market capitalization of $1.1 billion, according to Thomson Reuters data, a 44 percent increase from the average market cap during the height of the dot com craze in 1999. At the same time, companies are now going public 6.5 years after receiving their first venture capital backing on average, more than double the three years between initial funding and going public in 1999.

Cerebras: The AI of cheetahs and hyenas

The specialist starts out with a technology optimized for one specific task. Take the graphics-processing unit. As its name denotes, this was a specialist technology focused on a single task–processing graphics for display. And for the task of graphics, graphics-processing units are phenomenal. Nvidia built a great company on graphics-processing. But over time, the makers of graphics-processing units, AMD and Nvidia, have tried to bring their graphics devices to markets with different requirements, to continue the analogy to hunt things that aren’t gazelle. In these markets, what was once a benefit, finely tuned technology for graphics (or gazelle-hunting), is now a burden. If you hunt up close like a leopard and never have to run fast, having your nose smooshed into your face is not an advantage and may well be a disadvantage. When you hunt things you were no longer designed to hunt, the very things that made you optimized and specialized are no longer assets.

Intel is the classic example of a generalist. For more than 30 years the x86 CPU they pioneered was the answer to every compute problem. And they gobbled up everything and built an amazing company. But then there emerged compute problems that specialists were better at, and were big enough to support specialist companies—such as cell phones, graphics and we believe AI. In each of these domains specialist architectures dominate.

We are specialists, designing technology for a much more focused purpose than the big companies burdened with multiple markets to serve and legacy architectures to carry forward. Specialists are always better at their target task. They do not carry the burden of trying to do many different things well, nor the architectural deadweight of optimizations for other markets. We focus and are dedicated to a single purpose. The question of whether we—and every other specialist– will be successful rests on whether the market is large enough to support that specialist approach. Whether, in other words, there are enough gazelle to pursue. In every market large enough, specialists win. It is in collections of many modest markets, that the generalist wins. We believe that the AI compute market will be one of the largest markets in all of infrastructure. It will be the domain of specialists.


This $2 billion AI startup aims to teach factory robots to think

What sets Preferred Networks apart from the hundreds of other AI startups is its ties to Japan’s manufacturing might. Deep learning algorithms depend on data and the startup is plugging into some of the rarest anywhere. Its deals with Toyota and Fanuc Corp., the world’s biggest maker of industrial robots, give it access to the world’s top factories. While Google used its search engine to become an AI superpower, and Facebook Inc. mined its social network, Preferred Networks has an opportunity to analyze and potentially improve how just about everything is made.

At an expo in Japan a few months later, another demo showed how the tech might one day be used to turn factory robots into something closer to skilled craftsmen. Programming a Fanuc bin-picking robot to grab items out of a tangled mass might take a human engineer several days. Nishikawa and Okanohara showed that machines could teach themselves overnight. Working together, a team of eight could master the task in an hour. If thousands — or millions — were linked together, the learning would be exponentially faster. “It takes 10 years to train a skilled machinist, and that knowledge can’t just be downloaded to another person” Fanuc’s Inaba explained. “But once you have a robot expert, you can multiply it infinitely.”

China buys up flying schools as pilot demand rises

In September Ryanair axed 20,000 flights due to a rostering mess-up made worse by pilot shortages. This forced the low-cost carrier to reverse a longstanding policy and recognise trade unions and agree new pay deals — a move that it said would cost it €100m ($120m) a year from 2019.

China is on course to overtake the US as the world’s largest air travel market by 2022, according to the International Air Transport Association.

US aircraft maker Boeing predicts China will need 110,000 new pilots in the years through to 2035, and its airlines are expected to purchase 7,000 commercial aircraft over the next two decades.

China’s aviation market grew by 13 per cent last year, with 549m passengers taking to the skies, double the number who flew in 2010. Growth is being driven by the rising middle class, an expansion of routes by Chinese airlines and the easing of visa restrictions by foreign governments keen to attract Chinese tourists.

California will require solar power for new homes

Long a leader and trendsetter in its clean-energy goals, California took a giant step on Wednesday, becoming the first state to require all new homes to have solar power.

The new requirement, to take effect in two years, brings solar power into the mainstream in a way it has never been until now. It will add thousands of dollars to the cost of home when a shortage of affordable housing is one of California’s most pressing issues.

Just half a percent

If you save $5,000 a year for 40 years and make only 8% (the “small” mistake), you’ll retire with about $1.46 million. But if you earn 8.5% instead, you’ll retire with nearly $1.7 million. The additional $230,000 or so may not seem like enough to change your life, but that additional portfolio value is worth more than all of the money you invested over the years. Result: You retire with 16% more.

Your gains don’t stop there. Assume you continue earning either 8% or 8.5% while you withdraw 4% of your portfolio each year and that you live for 25 years after retirement. If your lifetime return is 8%, your total retirement withdrawals are just shy of $2.5 million. If your lifetime return is 8.5% instead, you withdraw about $3.1 million. That’s an extra $600,000 for your “golden years,” a bonus of three times the total dollars you originally saved.

Your heirs will also have plenty of reasons to be grateful for your 0.5% boost in return. If your lifetime return was 8%, your estate will be worth about $3.9 million. If you earned 8.5% instead, your estate is worth more than $5.1 million.

Keep your investment costs low.
Slowly increasing your savings rate over time.
Consistently saving while treating investment contributions like a periodic bill payment.
Bettering your career prospects to increase your income over time.
Avoiding behavioral investment mistakes which can act as a counterweight to the benefits of compounding.

Curated Insights 2017.07.23

The limitations of deep learning

…the only real success of deep learning so far has been the ability to map space X to space Y using a continuous geometric transform, given large amounts of human-annotated data. Doing this well is a game-changer for essentially every industry, but it is still a very long way from human-level AI.

To lift some of these limitations and start competing with human brains, we need to move away from straightforward input-to-output mappings, and on to reasoning and abstraction. A likely appropriate substrate for abstract modeling of various situations and concepts is that of computer programs.


Machines poised to take over 30% of work at banks, McKinsey says

Cognitive technologies — applications or machines that perform tasks once requiring human thought — are now cheap enough that banks can deploy them across operations facilitating trades or other capital-markets business. Automating tasks will “free up capacity” for staff to focus on higher-value work, such as research, generating new ideas or tending to clients.

Machine learning — which uses algorithms to identify patterns in large sets of data — can help sales and trading staffs understand positions faster and predict what flows will look like.

Natural language processing can perform legal and regulatory tasks by scanning through records, emails and recordings to translate them into structured data.

Cognitive agents can act as in-house personal assistants or service centers; think of help desks for trading staffs that have issues with their systems.

Robotic process automation — in which machines handle repetitive tasks — is particularly effective in banks’ middle offices, where it can help with end-of-day valuations and extract data.

Smart workflow tools — including document scanning and automated data entry — can speed the process of signing up new clients.

Netflix provided a new set of documentation, along with its customary earnings report, discussing how it accounts for its spending on content

Netflix surges 11%: sub adds crush estimates; discloses ‘content accounting’

“In continued success, we will deploy increased capital in content, particularly in owned originals, and, as we have said before, we expect to be FCF negative for many years. Since our FCF is driven by our content investment, particularly in self-produced originals, we wanted to provide some additional context on our content accounting at our investor relations website.”

Amazon Prime and other subscription businesses: How do you value a subscriber?

“[Understanding] the actual unit economics in the underlying business…requires analyzing the ‘true’ contribution margin of the business; not simply looking at gross or net revenue and the proper contra-revenue treatment, and not even looking just at gross margin as defined by the company. Many companies embed costs that are truly variable (for instance customer support, marketing, credit card processing) below the gross margin line. If you want to know if the business model truly hunts, you must pay careful attention. Otherwise, you may have simply found a company that is simply selling dollars for $0.85.”


Amazon is buying products from some US retailers at full price to build global inventory

The new program, which follows a similar rollout in Europe, is the latest move by Jeff Bezos to build up a complete catalog, even if Amazon can’t make much money on the products in question. In some cases, Amazon is approaching these third-party merchants after the manufacturer has declined to distribute the products through Amazon.

“When items are unavailable in a particular geography, we provide customers with selection from another marketplace. This offers customers a wider selection of great brands and helps sellers increase sales.”


Vanguard, the Amazon of asset management

What Vanguard’s founder, Jack Bogle, and company do have going for them is a unique ownership structure. Fund investors double as the shareholders. This allows Vanguard to essentially operate at cost, spending incremental profit on lower fees.

Vanguard has benefitted from a killer combination in recent years—low cost and quality performance. This is because although the firm distributes ETFs and actively managed funds, they specialize in passive, index-based investing—a style which has surged in popularity amid widespread underperformance across the active manager community.

Investing is a game of probability. Why would anyone want to pay 6x more for a product with a 90% likelihood of being inferior? The average actively managed mutual fund fee is 0.72%—6x higher than Vanguard’s 0.12% annual fee. And roughly 90% of those funds are underperforming Vanguard’s ultra-cheap option.


In urban China, cash is rapidly becoming obsolete

Ant Financial and Tencent were set to surpass credit card companies like Visa and Mastercard in total global transactions per day in the coming year. The key is that both companies are able to provide payments on the cheap, partly by allowing smaller vendors to make use of a simple printout of a QR code or their phone, instead of an expensive card reader. A back-end system that stores a record of user accounts, instead of having to communicate with a bank, also keeps costs down.

As the country builds its entire consumer economy around two private smartphone payment platforms, it is slowly locking out people unable to get onto those networks, and locking itself into those companies.