Uber and The Network Effect

Uber. Love the company or hate it, Uber has staked its claim as the defining company of our decade. Its story arc, whether it ends in a dramatic failure or lands with aplomb, will be retold in the same manner the histories of companies like Google, Facebook, and Microsoft are.

However, whether the story ends well or not will be partly influenced by a highly technical question, far divorced from the drama of investor-litigation and ousted CEOs.

Does Uber benefit from a network effect?

The presence or absence of network effects is critical question for a company, its investors, and anyone who is possesses an interest in a company’s potential to make a durable profit over time.

Why? For a company to demonstrate a large return on invested capital (ROIC) over time, it has to maintain a durable competitive advantage over competitors. Just making a product or peddling a service that people want — hard enough as it is — is usually not enough to guarantee that a company can sustainably maintain profit in excess of the cost of invested capital over a long period of time. Rather, to accomplish this goal a firm must also contain an element of uniqueness. It has to do something or own something that’s not easily copyable. Otherwise, the profits that it makes will be directly eaten away by competitive forces of direct rivals and new entrants, and indirectly through competition from substitutes, powerful suppliers and powerful buyers. This is an unfortunate fact of doing business. Adam Smith’s model of competition — the one college freshman learn in ECON 101 — makes for an efficient market, but doesn’t make for a cash-rich, long-lasting business. All the great business practitioners, analysts and theoreticians appreciate the power of durable competitive advantage. Warren Buffett and Charlie Munger talk about it through their concept of ‘economic moats’ in their investments. Bill Gurley, one of the best VCs of all-time, referenced the idea when discussing the company characteristics that make for a high Price/Revenue multiple. Michael Porter, maybe the most important business theorist ever, wrote an entire book about it titled Competitive Advantage. The importance of the concept is such that all the great businesses of our time, the ones that are the subject of biographies and magazine exposés, have extremely deep and wide economic moats. Apple, Google, Intel, even companies like Ford, GM, and WalMart; all have durable competitive advantage.

Does Uber?

That’s the real, underlying question we’re trying to answer when we ask: “Is Uber subject to a network effect?”. Network effects, which are also called demand side economies of scale, result when a product or service becomes more valuable as a function of the more people use it. That is, the product provides more value to user n+1000 then it gave to user n. You can sniff out their presence by answering the question: “Does the product/service get better directly as a function of adding new users?” Network effects are one way in which companies can maintain a durable competitive advantage. Others include: patents/IP, brands, supply-side economies of scale, experience curves, switching costs/lock-in, access to distribution channels, and favorable government regulation (there are others as well, but these are the main ones).

But compared to these other methods, strong network effects hold an outsized position of importance in comparison. Why? Strong network effects erect nearly insurmountable barriers to entry against competitors and therefore yield commensurately impressive returns to investors and founders of the companies whose product benefits from them. If you gave me the best team of software engineers, tech executives, marketers in Silicon Valley, a better product, and an endless war-chest (and told me to leave because I know nothing about building companies), I would not be able to displace Facebook from its de-facto monopoly in social networking. The strongest network effects are truly irreplaceable. They cannot be bought, with money alone, with a better product, or with talent. To really appreciate their power, compare them to the other sources of competitive advantage.

Patents present a powerful barrier to entry — the law. But firstly, they expire. This generally forces companies to find more creative ways of generating durable competitive advantage after their key patents expire. Additionally, patents are rarely broad enough to exclude any type of new product from being created. Patents for new molecular entities can be circumvented by different molecular entities that perform the same job or differ in slight ways. Historically many of the great patent holders have been circumvented eventually.

Other IP constructs, mainly copyrights, last much longer than patents do, but are substantially less broad in their restriction, and generally are less important in governing the life of technology-based companies. Rather, they mainly control the ownership of content.

Supply side economies of scale — what we traditionally think of as scale economies — arise when the marginal cost of producing a product or service decline as a function of the number of units produced. This is a well characterized economic phenomenon, one that you can find operating in a variety of famous companies, most notably companies like Wal-mart, Ford (the original version), Amazon, and Standard Oil. These companies rely on cost leadership as their main strategic imperative — and if that’s your strategy, generally scale is a boon. However, despite playing some role in almost all durable companies, scale economies eventually run out of fuel, usually well before a company dominates the majority of the market. That is they are subject to diminishing marginal returns over time. Generally, the curve that describes the relationship between output and cost is parabolically shaped. That is eventually, the economies of scale are trumped by the diseconomies — usually in the form of logistical concerns — associated with producing a massive amount of a product.

Additionally, this source competitive advantage is more tenuous to maintain, even at the lower market shares that a company can achieve. Firstly, supply side economies of scale can be bought with a large enough war chest. They also have durability problems over time. As market leadership is being built, small cost advantages as a result of supply side scale advantages can parlay into fast growth, but over time, they become commensurately less important. For example, Standard Oil ruthlessly chased scale during the growth phase of the oil market, but eventually, scale ceased to be a competitive advantage — modern oil refining companies do not chase scale over all else. But most importantly, supply side economies of scale are important mainly when cost is the primary consideration of customers. Supply side economies of scale create cost advantages — but if cost no longer becomes the dominating concern of customers, the moat created by a persistent cost advantage can shrink and do so quickly.

The Limits of Scale

A full comparison of the sources of competitive advantage is beyond the scope of this article. However, it is worth pointing out that demand side economies of scale suffer from none of the drawbacks listed above. Theoretically, if there are more customers to be had, a market leader with strong network effects can continue adding more value to users relative to its competitors. This is why in markets that are subject to strong network effects, one company tends to own an outsized proportion of the customers, on the order of 80+% (often called ‘winner take all dynamics’). Additionally, while network effects can vary in strength, the strongest networks often improve value to users nonlinearly (the most popular relationship is O(n*logn). ), meaning that the race to market leadership dominates all other concerns.

So what about Uber? Is it a market leader because it presents strong network effects as a barrier to entry. At first glance, the answer is no. For value to increase with the number of users, value to consumers must not be independent — the value yielded by a product must be dependent on other individuals also using the product. However, this does appear to be the case with Uber. I care very little if other individuals are using Uber’s car-hailing service so long as I get a car to me in less than 5 minutes. The same is true for Uber drivers — Uber having more drivers does not increase the value of Uber to other drivers. So at first it appears I have written a very long article for no reason.

But the presence of both drivers and riders (customers) complicates things a bit.

Uber is part of a new generation of companies that builds multi-sided platforms or multi-sided markets. Platform products bring together 2 or more parties in such a way that benefits both parties, with one or both sides of the platform being charged. Sometimes this takes the form of a literal marketplace. An example of this is Ebay, one of the pioneers of the model. But it can be less obvious as well. Microsoft’s Windows operating system was a platform that brought together developers (of applications that ran on Windows OS systems) and users. Google built a search engine, but it monetized a platform that connected users and advertisers.

Uber builds an application that lets users call a car to take them where they want to go. But it also is a platform with two sides: riders and drivers, with the number of riders vastly exceeding the number of drivers. Uber might not have direct network effects, but it definitely contains a network dynamic between the riders and drivers. That is, the number of riders influences the value of the platform to drivers and vice versa. As a rider, I might not care about other riders directly, in the way that I care about whether my friends use Instagram, Snapchat, or Facebook, but I certainly do care (even if implicitly) about the number and location of drivers on Uber’s platform. Too few or too far away, and I will not like the delay or inconvenience of the inability to call a driver in an arbitrary location and perhaps prefer a competitor. Drivers too, care about the number and location density of riders — that’s their source of money!

So critically, the influencers of this interdependence are the number of drivers and riders, and their location (which itself is a function of density). How many drivers/riders are there andover what location are they dispersed? This makes itself seen directly through pickup times and location on the rider side (can I get a ride anywhere I am? and how long do I have to wait?), and utilization rate (how often am I making money through driving?) on the driver side. Lastly we also have to consider price (this though, is generally a function of cost and so more a function of a supply side economy of scale) . More riders leads to a higher density of riders and therefore a higher utilization rate for drivers (more people to service, less time in between services), more money for drivers, leading to more drivers overall on Uber (vs. a competitor) which improves the service (in terms of pickup times and localization) for the riders leading once more to more riders. Additionally, the more drivers are utilized the more Uber can lower the price of a ride because there’s more volume on the network to offset the drop in price (this has been demonstrated — Uber’s prices have substantially dropped over time such that they are now cheaper on average than taxis ever were).

This ‘interlink’ between the different sides of a platform create a dynamic known as a cross-side network effect, a subset of network effects termed ‘indirect network effects’ (also called ‘chicken and egg’ problems) by Google economist Hal Varian. To hammer home the example, consider the DVD. I don’t directly care whether others own DVDs or not, but I am influenced by whether a set of my favorite content is on DVD versus some other technology, which itself is influenced by whether a majority of users use DVDs or not.

So Uber does not contain direct network effects on either side of its market, but it does contain a cross side indirect network effect. The important question then becomes: what does this mean for its competitive advantage? As it turns out, this dynamic is being played out around the globe as we speak, so we will probably get the answer to it in due time. However, in the meanwhile we can make some predictions.

To make these predictions we have to get a little bit deeper in our analysis of the nature of this cross-side network effect. Unlike its presentation the details of a network effect matter because it determines how strong it is, and thus how defensible a company’s position is. It’s not a ‘yes or no’ proposition.

Network Effects Over Space

The discussion surrounding Uber and network effects has generally centered around locality. Most agree that there is some degree of indirect network effect, but individuals doubt whether the network effect is strong enough to fend off competitors or to justify its valuation of $69.5 Billion. That doubt of the strength of the network effect is partly due to considerations of distance. Within a given geographic location — usually the city and surrounding suburbs — people are willing to grant that Uber probably benefits from a network effect. But this effect, they argue, is independent of location. Whether you are the market leader in NYC has no bearing on whether you are the market leader in San Francisco. In the worst case, this city-by-city independence would lead to a fractious geographic structure where there are individual winners in cities, but the market looks impossibly fractured on a province or national scale. The market dynamics would then resemble a military invasion — more similar to true combat than the industrial warfare that is traditionally waged between competitors. I suspect that this argument is more true than even Uber’s most ardent defenders will admit. It would seem to explain the dynamics of ride-sharing internationally, where competitors have emerged in Russia (Yandex), China (DiDi), India (Ola), and Singapore (Grab). However, these dynamics do not seem to be emerging in the U.S., where Uber seems to be entrenched in most metropolitan areas. Whether due to networking effects or not, it appears that the limit to Uber’s market leadership is national, not metropolitan. This leads to a broader discussion of what could lead Uber’s indirect network effect to persist outside of cityscapes.

The first reason this could occur is through people moving. If I use Uber in one city, I’m likely to use it in another city. This effect is most likely a result of Uber’s brand, which we’ll dive into detail about in the next section. But if it’s true that the Uber name travels, it does make it more likely that Uber will become a market leader in more cities as a result of its leadership in another city. But the effect does seem weak — certainly weaker than the intra-city network effect. Remember, Uber’s network effect is a function primarily of location and pickup time on the rider side and utilization rate (proxy for money made) on the driver side. If people’s preferences travel, then more riders in one city leads to more riders another city whenever they travel. But what percentage of riders from one city travel and bring their preferences with them? The network effect depends heavily on the sheer number of riders choosing Uber over a competitor in a new city and that number seem small.

Investor Bill Gurley also claims that these travelers make launching easier in new cities (e.g. I used Uber in NYC on a business trip, but I live in city Y that doesn’t have Uber). But again, these effects — in terms of the raw magnitude of riders necessary to catalyze the indirect network effect, seem comparably tiny. That being said we must acknowledge a key fact: ride-sharing might be reaching a more mature state in 2017, but in 2014, it was very young. And as in all network effect businesses, small edges in market leadership at the start of the growth cycle can perpetuate themselves through self-propagation. A small amount of travelers might seem insignificant now, particularly in mature cities where Uber has already ‘won’ or ‘lost’. But when Uber was launching, these travelers probably had an outsized influence on catalyzing the network effect. The economic phenomenon I refer to is the concept of critical mass in network effects. How many riders is enough to catalyze the positive network externality? How many riders/drivers is necessary to attract the number of drivers to meet the ‘good enough’ standard for location, pickup time, and utilization rate? The number is unique for all markets, and I don’t know of a way to try to obtain an estimate. But if small enough, travelers could be enough to catalyze the network effect, which would explain Uber’s dominance in most American metropolitan areas, but not internationally.

Two additional reasons present themselves as ways in which cities are not independent of each other.

First, brand allegiance by existing customers as they travel might be tiny, but Uber’s reputation amongst ‘non consumers’ (note the subtle difference) is not. Prior to Uber’s very bad, no good 2017, Uber launching a city would yield almost instant (positive) recognition, ensuring that there was a ready set of riders available — kicking off the network effect from the get go.

Secondly, travelers might be small in number, but they are high volume users and they maintain, as a group, persistent presence. That is, they use the service a lot while they are traveling, and as a group there are always a good percentage of travelers.

Outside of travelers, there are the traditional benefits to market leadership, seen across industries: it’s easier to raise capital (Uber has raised a bunch of it), it’s easier to partner with other complementary firms, and that customers tend to want to keep you in power. But again, these are competitive advantages around scale, not network effects.

But the effect that I think most probably explains Uber’s national dominance as opposed to individual metropolitan dominance is the first-mover advantage. At times, too much is made of the first-mover advantage. As many people have noted — Friendster caught the social network wave early, only to be thoroughly annihilated by Facebook. Venture capitalists like to say that it’s not first to market that wins, but first to product-market fit that does. But here’s the thing: Uber’s key value proposition remains mostly unchanged from city to city. Thus, after proving out product market fit in its home base of operations, San Francisco, it could win by being first to market in other cities. And if it was first to market, it was the first to get riders and drivers on board, meaning the network effect already began churning before it’s competitors could even get a start. The rich get richer.

This effect seems to explain the discrepancy in Uber’s success domestically, where it is the market leader, versus internationally, where Uber retrenched in China, ceded Russia, and is still locked in a battle with Ola in India.

The commodity problem.

The second concern that gets mentioned with regard to Uber’s network effect is that it peddles a commodity. It’s been made by some pretty smart individuals like Christopher Mims at the Wall St. Journal, as well as by Sam Gerstenzang, a former member of a16z’s (invested in Lyft) investing team.

According to Mims, “Drivers are completely mercenary and driven by price; they have no specific loyalty to Uber”.

He claims that, “Uber is entering what is essentially a frictionless market (for both drivers and riders) in which its services are a commodity. The company’s phenomenal growth so far (we don’t know the actual numbers, but doubling in revenue every 6 months is what Kalanick claims) has been built on the back of low-hanging fruit — expansion into new cities, particularly where taxi availability is low — and levels of dissatisfaction among taxi drivers that may be temporal”.

Gerstenzang follows this with by differentiating Uber from its marketplace model brethren, saying, “Uber is not a marketplace (like Airbnb or eBay), because the “products” on its shelves are indistinguishable — diversity doesn’t matter. I don’t care if there are 10 drivers or 10,000 drivers as long as they come to my house in two minutes.”

Mims and Gerstenzang are making two slightly different but related points, but both implicitly or explicitly call into question the strength of Uber’s competitive moat.

The first point is made by both Mims and Gerstenzang: Uber sells a commodity, where the service being sold (the ability to be picked up anywhere and dropped off) is indistinguishable between competitors and thus price becomes the dominant factor by which both riders and drivers make user decisions.

The second point, hinted at by Mims regards the loyalty of drivers to Uber’s platform. He goes on in the article to describe how smart Uber/Lyft drivers have multiple phones and work for multiple platforms, making decisions purely based on which platform creates a better priced inbound request. To couch this concern in the language of strategy, Mims argues there are no switching costs to lock-in the driver and thus the barrier to entry for competitor recruitment of drivers is low.

Note: This discussion has taken a slight turn away from our previous discussion of network effects. We’ve been talking about network effects. Switching costs are another barrier to entry that companies attempt to create for their products. I will talk about the connection between the two later on.

Let’s handle the first. Mims and Gerstenzang point out that what Uber sells is a commodity and thus its product is indifferentiable from a competitors. Why is that bad? Ultimately because it’s very hard to build pricing power around a hard-to-differentiate product. Warren Buffet and Charlie Munger define pricing power as the ability of a business to raise prices by 10% and not see a mass defection of its customers. It is a proxy metric for measuring the strength of a firm’s moat. Companies with strong sources of competitive advantage — like Buffett’s classic investment in See’s Candies — usually can raise prices slightly without damaging their profits. Traditionally, commodities are thought to be difficult products to build moats around. If one product is indifferentiable from another, price becomes the only factor around which competition occurs and thus businesses selling these types of product fail Buffett’s pricing power test.

So are they right? I’d argue not quite. The argument that Uber selling a commodity immediately implies an inability to maintain a competitive advantage takes too narrow a view on the different sources of competitive advantage. Product differentiation is only one of many strategies that can be taken to build a moat. In fact, I would argue that companies — quite large ones in fact — have built durable competitive advantage (remember this is the exact same as an ‘economic moat’) selling the exact type of commodities that Uber is purported to sell. For a history of this we only have to turn to the consumer packaged goods (CPG) industry. This industry, which sells us things like shampoo, razors, cereals — pretty much anything you find at your local grocery store — has been selling commodities for half a century. Anyone can make cereal — the barriers to entry seem quite low. Yet, the companies who make these goods have found ways to differentiate all the same. Rather than differentiating purely on their product, they differentiated by sales channel, distribution mechanism, brand loyalty, and many other factors. You can buy an equivalent of an Oreo two shelves down from the branded version, but Nabisco still makes a consistent profit.

Specifically, in the case where the underlying product is relatively undifferentiated, branding becomes critical. Brands as a source of competitive advantage are generally given too much credit. But in the case where the goods are relatively undifferentiated, prices are similar, and purchases are habitual, most individuals default to one brand over another. I for example, always use Colgate toothpaste, irrespective of whether Crest’s price is a bit lower. Because it requires too much mental energy to make exclusively rational cost-benefit judgements for these repeat purchases, consumers tend to make associations between brands and the jobs that they do. That is, whenever I think about furniture, I immediately ‘hire’ Ikea to perform that job for me. This association between ‘job’ and ‘brand’ defines the way in which brands can confer competitive advantage.

Thus, a practical switching cost is introduced by Uber’s brand, locking riders into the platform. Theoretically, Mims argument is sound — Lyft is only a free download away. But when riders need to get from Point A to Point B, very few riders are going to make the deliberate comparisons between price — they will use their default. The switching cost and stickiness of Uber over its competitors derives from consumer laziness and desire to avoid mental work.

The second argument however, holds much more weight because it discusses the lack of switching costs on the driver side of the marketplace. Drivers, unlike riders, have every incentive to carefully and rationally consider which platform to drive on. And his anecdote about drivers having multiple phones and acting as price mercenaries gives Mims’ theory undeniable weight. Brand means nothing on this side of the market. So I agree that for drivers at least, Uber appears to operate in a frictionless market. But does that undermine Uber’s competitive advantage? Common sense says yes — if Uber drivers are constantly able to operate on more than one platform, that reduces the pricing power Uber has over its drivers. It cannot lower prices for riders, is forced to provide bonus payments to incentivize drivers to stay on the platform and maintain the liquidity necessary to serve its population. In the worst-case, Mims’ theory spells doom for Uber by creating a negative externality. Just as an increased number of riders can incentivize more drivers and vice versa — see the description of the network effect — a sudden drop in the number of drivers or riders could send the cycle spinning in the opposite direction.

But again, there is more complexity to consider. The only metric that matters to drivers is utilization rate: how much of their time can be spent driving and what are they get paid for that work? But that metric is a function of rider density — the number of requests coming from Uber as compared to its competitors. If a service has substantially more riders, most of the requests will come from that service, and drivers will have no choice but to service that platform. The switching cost for drivers derives from the fundamentally finite supply of time. Any time spent on an Uber ride is necessarily time not spent on a competitor ride. Thus while price matters — ultimately the supply of riders trumps all. So long as the underlying value proposition is sound and drivers can’t just decide to stop ride-sharing en masse — and ride-sharing’s explosion globally would suggest that it is — ultimately drivers will be forced to go where the riders are. This trends aligns well with the competitive drivers of other marketplace companies who rely on similar indirect network effects. For companies like Google, eBay, Amazon, etc., the only thing that matters is winning a critical mass of demand. That critical amount of demand allows the platform company to monopolize supply as a monopsonist (a powerful buyer). If all the demand comes from one platform, where is the supply to go?

Additionally, Uber can take steps to attempt to introduce ‘stickiness’ into the driver side of the network as well. Critically, we here take a step into speculation. Amazon for example, ran into the same stickiness problem as it scaled — other e-Commerce sites were only a link away (and they didn’t have a monopoly over supplier time to work with either). They countered through a loyalty program that targeted not suppliers, but customers. You might have heard of it! By ensuring that customers were repeat, high volume purchasers on Amazon’s platform, suppliers had no choice but to sell there. Uber could try something similar.

But what of the network effect? We’ve spent a long time talking about the nature of Uber’s product and switching costs. But how does this intersect with our prior discussion of network effects?

Network effects alone do not necessarily guarantee a powerful competitive moat — Mims is correct point out that switching costs present a critical additional barrier to entry that when combined with network effects (both direct and indirect) can make for a truly terrifying and nigh-on unbeatable competitive advantage. A great example of this would be Microsoft’s OS dominance (close to 98% market share at its peak with a $800B inflation-adjusted market capitalization). High switching costs and an indirect network effect led to quite a bit of profit and one very rich man. Mims — and technology industry followers in general — love to point to Groupon and its failure as the quintessential example of a company that leveraged networking effects but suffered from low switching costs. But as we’ve pointed out, in this case the network effect monopolizes drivers’ time. The network effect introduces a switching cost in of itself (not always the case).

That being said, Uber selling a commodity does imply that certain features from which it could derive its indirect network effects will remain closed off. For example, eBay’s network effect was based partly on volume of merchandise available, but also on diversity. The more suppliers it could attract to the network, the more diversity was available for buyers. This feature clearly is not available to Uber and thus it would not be invalid to argue that Uber’s network effect could be weaker than Ebay’s or other analogous companies for this very reason. In particular, because the ‘value’ part of Uber’s network effect is limited in breadth and the underlying product is reliant on brand-imagery to provide lock-in, it puts an emphasis on the first-mover advantage I mentioned earlier. Thus, while Mims’ argument on its own doesn’t mean Uber’s valuation isn’t justified, it does raise an adjacent concern by defining the competitive dynamics of the ride-sharing market quite narrowly: get to riders and get their first. This means that in areas where Uber recruited riders first and achieved critical mass, Uber is likely to have the type of pricing power that Mims’ argues against as a result of amassing rider share that enforces switching costs naturally. But in places where the competitive situation is more ambiguous, there is a much darker future for Uber.

Ultimately, I think the concerns over commoditization have distinct solutions that can be implemented. But, to dismiss the concerns would do a disservice to anyone trying to come to an accurate valuation for Uber.

So what does this discussion suggest about Uber’s future value? Guessing Uber’s value presents, in my opinion, a consummate challenge for an analyst interested in startups. Some of the best, including NYU Stern Professor Aswath Damodaran (who wrote the literal textbook on corporate valuation) and Bill Gurley, took swings at it. One has to perform a detailed analysis to break out Uber’s contribution margin on a per drive and per rider basis (i.e. perform a LTV calculation), while also not losing sight of broader theoretical concepts that might change those numerical calculations in the future. To be right about Uber’s ultimate value is to ace the startup forecasting test in a way that suggests one truly has understood all the inputs that go into determining a young, growing company’s value.

Critically, I have not performed a full analysis of Uber. This piece was dedicated to only one part of Uber’s eventual valuation — analyzing the characteristic of Uber’s network effect. That analysis can serve to forecast Uber’s potential market share, and aid in attempting to understand whether Uber will be able to make the unit economics work. It even could suggest strategies for Uber to take with regards to pricing. But it cannot answer the question of “how high is up?” and it doesn’t consider any other potential competitive advantages that Uber could have (e.g. scale). Let’s not even get into the situation in which Uber successfully builds an autonomous vehicle.

But there are some key takeaways to glean.

Firstly, Uber does present a network effect as a barrier to entry — albeit an indirect one. This network effect is a cross-side network effect between riders and drivers. However, the network effect has limitations. It appears to have limits in scale — exhausting at the national level in practice, and potentially the metropolitan level in theory. Additionally, Uber’s network effect is limited in breadth — based purely around rider/driver density and corresponding pickup times, utilization rate, and price. These limitations, along with the habitual nature of purchases in the ride-sharing market have placed the onus on being the first-to-market as a way of catalyzing the network effect and its associated competitive advantage. This emphasis on being first-to-market appears to explain the discrepancy between Uber’s national dominance but international retrenchment. In addition to the competitive advantages of Uber’s network effect alone, the effect is powerful enough it seems to introduce switching costs of its own onto drivers through the monopolization of time, overcoming some of the problems associated with Uber selling a ‘hard-to-differentiate’ service.

Secondly, if this characterization of Uber’s network effect accurately represents reality, it suggests that Uber has made strategic mistakes in its growth strategy:

  • In domestic markets, rather than attempt to lock drivers into Uber’s platform through volume based bonuses, Uber could directly change the amount that gets paid out to drivers on a per-ride basis. Uber should be worried about getting enough drivers into the market to supply their growing numbers, but ultimately, it is the number of riders and thus the proportion of inbound ride requests originating from Uber that controls the lock-in for drivers.
  • In international markets, rather than burn cash trying to play catch-up to incumbents, they should understand the power of the first-mover principle. Because the network effect is limited in breadth to a battle of liquidity, entering too late into a market spells doom. To be fair, Uber has recently retrenched — understanding the limits of its network effect internationally by conceding in China and in Russia.
  • Lastly, although the network effect exists, some of its weaknesses do not seem to have been addressed properly. Unlike Amazon, whose leadership understood the limits of its network effect to keep customers on Amazon’s e-commerce platform and took steps to increase its ‘stickiness’, Uber has not attempted to improve rider or driver lock-in to Uber’s platform in a successful way.

Thirdly, Uber’s mistakes in running company culture, and its other recent missteps could have a potentially ruinous effect on Uber’s domestic market leadership. Uber’s network effect and quick moving strategy allowed it gain market leadership domestically. And network effects, by definition, encourage market leaders to remain market leaders. But they are not unbreakable. Specifically, note that Uber’s network effect directly depends on a majority of riders using its platform, kept there by weaker switching costs than one might ideally like. They are there because of brand. But Uber’s missteps could cause users to overcome the habitual use of Uber in favor of rivals. And thus, the number of riders could shrink, correspondingly reversing the network effect. This situation is one Lyft — which has raised close to $3 Billion itself — would be ideally positioned to benefit from.

So count me as one of those who is if not worried about Uber, is somewhat disappointed. I don’t think the company will be worth over $100 Billion over the next couple of years, as early-backer Benchmark claims to believe. However, I don’t think Uber — yet — is in danger of losing its domestic dominance. The network effect, however limited, appears to be strong enough to maintain Uber’s grip on the U.S. But the bull-case for Uber, the one in which it becomes a global monopoly, completely redefining the nature of transformation and becoming a business on the scale of Amazon, Google, and Facebook, looks increasingly unlikely. And that is a shame. Uber’s missteps are entirely of its own perpetuation, and I by no means feel sorry for Uber’s management. Rather, I feel dismayed that Uber’s grand vision will likely never come to pass.