Quick commerce and the seductive lies of network effects
From Uber to Zepto - why some models work in local markets and others don't
Uber is looking to buy Expedia. This is a continuation of Uber’s bet on building demand-side advantages—customer loyalty, network effects, and the elusive super-app vision.
But if we look at the few businesses that have performed well in local markets, they have done so by doing the very opposite - by building asset-intensive supply-side advantages.
Network effects have become the fallback ‘how-we-win’ argument, often applied indiscriminately to any platform connecting producers and consumers. More producers bring more consumers, creating a nice, little flywheel.
Local market players like Uber, Doordash, Zomato, Zepto are often explained in terms of network effects.
Or in terms of data-driven learning. Gain enough users and the data will be invaluable.
But what if these theories don’t apply to local markets at all? What if adding users actually creates inefficiencies? What if more data leads to diminishing returns?
And how do you create advantages when your network doesn’t create defensible network effects and your data isn’t useful beyond a point?
Through this analysis, we’ll look to answer several questions:
Do network effects matter at all for local businesses?
When do dark kitchens make sense for food delivery and when do they not?
Why do densely populated markets like India favour Zomato, Swiggy, and Zepto over Ola in terms of margin expansion?
Can quick commerce’s 10-minute deliveries sustain?
Who wins when quick-commerce players enter traditional e-commerce categories?
Why Uber’s potential Expedia acquisition continues its super-app fallacy when defensibility in local markets lies at an entirely different source.
Let’s dive in!
The Uber Fallacy: Network effects misapplied
For years, Uber has been positioned as a business driven by network effects.
The logic seems intuitive: more drivers attract more riders, and more riders attract more drivers.
To be fair, I’ve been guilty of over-emphasising Uber’s network effects myself, as in this video from 2017 on the occasion of joining the ING Group’s Global Innovation Council.
David Sacks’ widely circulated diagram only reinforced the belief that Uber’s value grows exponentially as its network expands.
However, this does not reflect the economic reality of Uber’s business model.
In a network characterized by true network effects, each additional user increases the value of the network for every other user. But Uber’s network does not behave this way. As the number of drivers in a given city increases, the marginal benefit decreases after a certain threshold. Adding more drivers does not significantly reduce wait times for riders or improve the experience in any other way.
The benefit of additional drivers is capped by the commoditized nature of the service - i.e. the only benefit is reduced waiting time, there are no scope benefits of variety or choice.
Another limitation is the lack of captivity on drivers and riders - or low multihoming costs on both sides, as I’ve written earlier.
This is further complicated by the fact that Uber’s reputation system doesn’t reward its drivers so there is no incentive to stay back. Drivers can easily switch between Uber and competing ride-hailing services and riders “multi-home,” as well. This lack of stickiness undermines any potential for a lasting network effect. On Airbnb, hosts are incentivized to stay due to the value they build over time - better reviews lead to higher earnings -creating a reinforcing cycle of engagement.
Uber - Looking for advantages
Some may argue that Uber benefits from learning effects - improvements in routing or dispatching efficiency as it gains more data over time. However, much of Uber’s so-called “learning” is more effectively substituted by external sources such as Google Maps, and third party data providers, giving it no unique advantage there.
Data learning effects also decline drastically owing to the fact that almost all benefit to users translates into two things: low waiting times and shortest time to destination - the second of which is available as a relatively ‘public’ resource via Google Maps or Waze.
Thus, Uber’s growth is not protected by network effects or learning effects. In fact, it has no substantial moat to defend against competitors with similar access to technology and drivers.
Building local advantages
Uber’s defendants point to the ‘local nature’ of its network effect - the fact that a rider in city A cannot benefit from a driver in city B. As we’ve noted above, it is not the local nature of the network effect, but the
(1) commoditized benefit,
(2) low multi-homing costs,
(3) relatively low learning advantages beyond a certain scale, and
(4) absence of supply-side producer captivity
that comes in the way of Uber’s defensibility
Interestingly, the issue here is that looking for network effects distracts us from looking at the real source of advantage in local businesses - economies of density.
Economies of density + High fixed cost base + Production advantages
Economies of density emerge when a business can create efficiencies by focusing its operations in areas with concentrated demand.
Economies of density differ from economies of scale, though the two are often confused.
Economies of scale occur when a company’s per-unit costs decline as production increases—spreading fixed costs over a greater number of units. In contrast, economies of density concentrate resources - infrastructure, logistics, labor - within a limited geographic area.
The higher the local demand, the more intensively businesses can use these resources, driving down the cost per transaction.
To create barriers to entry, economies of density should be accompanied by high fixed costs.
This is why Uber was so keen to sell the self-driving fleet dream. Not because it would lower the payouts required to drivers and improve margins. But because self-driving fleets create a major barrier to entry - good ol’ fixed costs.
There’s a third factor that improves defensibility: production advantages. This was the other part of the self-driving dream - that learning advantages in self-driving would ensure that first to scale would maintain that lead.
So the winning formula in local markets involves:
High economies of density, combined with
High fixed cost base, combined with
Production/inventory advantages
Of course, that hasn’t really played out for Uber - self-driving fleets remain a dream.
So Uber’s next big bet on this three-part formula is…
Food Delivery
Predictably, Uber (and ride-hailing, in general) wants to do what every business struggling with network effects tries to do - become a super app.
As I’ve explained before, most super-apps fail.
But Uber’s expansion into food delivery - or Zomato’s expansion into quick commerce - isn’t a super-app play.
It is a play on economies of density + fixed costs + production/inventory advantages.
As I’ve explained before in The economics of food delivery:
When looking at the health of a food delivery platform, you need to focus on two key metrics:
1. Customer retention
2. Peak customer spend i.e. maximum spend per customer per time period (per week or per month)
The higher the customer retention and the faster you can move a cohort of customers to peak customer spend, the more likely you are to make the unit economics work.
On the cost side, the marginal cost of delivery plays an important role. Food delivery platforms work with very thin margins per order. Hence, route optimization and number of orders per route are important criteria.
Route optimization starts to benefit from some economies of density, when a company can cluster deliveries in a tightly defined area, reducing travel time and costs per delivery. The more orders concentrated within a given area, the faster and more cost-efficient each delivery becomes. This creates a self-reinforcing cycle: better delivery times drive more demand in that area, which further lowers costs and improves performance.
These are not network effects. While both involve a feedback loop, economies of density are grounded in geographic proximity and operational efficiency rather than the relatively abstract value of connecting more users.
But in the absence of fixed costs and production advantages, this still doesn’t create barriers to entry.
Food delivery players found their answer to fixed costs + production advantages: cloud kitchens.
As I explain in the same article:
The ‘delivery platform + dark kitchen’ integration playbook essentially works out as follows:
1. Aggregate consumer demand and delivery logistics as a delivery platform.
2. Partner with restaurants, build network effects, and learn from market-wide data to identify popular dishes, ordering patterns etc.
3. Create a dark kitchen based on this data and optimise cooking operations around demand patterns.
Step 2 is critical. F&B is a high risk business. Individual restaurant owners take the risk of starting new businesses. They pay high rents for premium real estate. Delivery platforms learn from the risk taken across the ecosystem and build stronger demand models than any individual restaurant could.
Dark kitchens finally create the elusive fixed costs.
Using data from restaurant orders, food delivery platforms capture demand patterns, and select warehouse placement as well as menu items - based on high-demand items in that locality - accordingly. This develops production advantages.
Economies of density work particularly well for such businesses where demand determines ‘stocking’ and availability of ‘stock’ determines faster delivery, which in turn creates a good customer experience driving more demand.
As unit economics improve, dark kitchens can also aggressively offer discounts, which spins this flywheel further.
The Quick Commerce Confusion
Quick commerce—offering ultra-fast delivery, often within 10 minutes—was met with skepticism when it first appeared.
Who needs delivery in 10 minutes? And will this persist beyond Covid?
As it turns out, the issue isn’t whether there was a demand-side consumer problem to be solved. The real issue that quick commerce attacks is whether there is a supply-side model that can substitute India’s unorganized retail at scale.
Indian cities are population-dense and small kirana stores solve daily needs. Since the mobile internet boom in India, and leading up to Covid - 2016-2020 - such stores would operate on Whatsapp, delivering small orders to nearby homes in response to messages from customers.
India’s unorganised retail is hyper-local and comes with all the problems of ‘last mile’ delivery.
Economies of scale cannot be created in such markets. But economies of density can.
Quick commerce is an infrastructure bet, not a market bet
Quick commerce isn’t about winning customers with 10-minute delivery; it’s about building a supply-side infrastructure that leverages economies of density.
Rapid delivery proves this efficiency, but once established, the same infrastructure can be used for other, less time-sensitive products, like electronics.
Advantage is created by securing micro-warehouses, or “dark stores,” in those areas. By positioning small distribution hubs close to dense clusters of demand, quick commerce companies can meet the intense need for rapid fulfillment more efficiently than traditional e-commerce models.
Once a company establishes these local warehouses, the operational model shifts.
Once a company invests in strategically located hubs, those fixed costs—rent, inventory, and infrastructure—create a barrier for competitors.
Each additional delivery then becomes cheaper, which increases demand and lowers costs even further.
Quick commerce isn’t about network effects; it’s about optimizing real estate to build advantage in specific high-density areas.
The defensibility of quick commerce lies in the fixed infrastructure—the micro-warehouses. It is the density of demand that drives efficiency and makes the model work, not the size of the network.
Quick commerce vs traditional e-commerce
Quick commerce players - Zepto, Blinkit, and others - are increasingly moving into categories dominated by e-commerce firms.
Misunderstanding quick commerce as a demand-side play would make this expansion seem illogical. But understanding it as a supply-side economies of density play changes how we look at this.
To understand the e-commerce vs quick commerce battle, we need to better understand economies of scale vs economies of density:
If you have economies of scale, adding more users helps expand your margin.
If you have economies of density, adding more users in existing locations only helps expand your margin.
This means that a business with economies of scale - traditional e-commerce with a nation-wide supply chain - can benefit from growth in its user base.
A business with economies of density - quick commerce - is instead faced with a trade-off - grow new locations or reinvest in existing ones.
In hyperlocal markets which have scaled out, quick commerce works very well. But its backers are constantly faced with the trade-off between doubling down on those markets vs growing new ones.
In the battle with e-commerce, though, quick commerce holds an ace.
Not all e-commerce customers are equally profitable. Some are less profitable than others. In fact, the most profitable customers and the least expensive deliveries tend to concentrate in some of the larger, densely populated cities.
This, in turn, means that quick commerce is well positioned to take the most attractive buyers away from e-commerce and is naturally not suited to serving the less attractive buyers who also happen to live in areas of low geographical density.
This is the fundamental conundrum facing e-commerce firms - if they lose their most profitable customers, margins drop sharply. This can trigger a downward spiral, where losing those valuable customers forces companies into a fierce battle to win them back. But if the e-com company lacks a key competitive advantage—like a hyperlocal warehouse network for faster delivery—it ends up burning cash without gaining advantage, pushing the business deeper into trouble.
If Uber gets into travel…
As I write this post, Uber is reportedly interested in buying Expedia.
This is a step in the wrong direction and continues to chase Uber’s ‘super-app’ pipe dream.
Since this news came out, LinkedIn thought leaders have been speculating on a similar online travel acquisition for Zomato.
This is reading Zomato’s business completely upside-down. Zomato is much better positioned to enter a far more lucrative business - e-commerce.
For a business that has mastered economies of density, online travel is possibly the worst business it can get into.
If anything, Zomato is best positioned to continue expanding its economies of density play and use its last-mile warehouse footprint for a growing range of use cases.
The real hyper-local advantage
When you see a hyper-local business, dont’ look for network effects. Don’t even try to explain it away with global network effects vs local network effects. Don’t try to use the catch-all data argument.
And definitely don’t go about looking for a super-app.
Look for
(1) economies of density
(2) built around a fixed cost base with
(3) ownership of production/inventory in high demand categories.
Ride-hailing, food delivery, and quick commerce are not driven by network effects, as is often claimed. Their success hinges on economies of density.
These hyper-local businesses thrive in environments where concentrated demand allows for operational efficiencies. And food delivery and quick commerce create sustainable advantages there if they build a defensible position around warehousing infrastructure and stock/inventory control.
That is where real defensibility is built out in hyper-local markets.
Your thoughtful commentary reminds me of the famous movie line from Jerry Maguire: “You had me at hello.” But in your case, your article lost me at your first question about Uber: “Why does Uber still struggle more than a decade after achieving so-called network effects?”
Over the past two years, Uber has become a profitable, cash-generating machine whose stock price has nearly quadrupled. So rather than try to explain Uber’s ongoing struggles, it may be more productive to try to understand the drivers behind its remarkable turnaround. Here’s my view on that question… https://len-sherman.medium.com/the-inconvenient-truths-ubers-ceo-doesn-t-want-you-to-know-2fc0cd742b24