The chicken-and-egg problem

The indirect network effect from the previous lesson has a vicious flip side. If two sides each attract the other, then at the start, when neither side is present, each is waiting for the other, and nothing moves.

A new marketplace has no buyers, so no seller wants to list, so there is nothing to buy, so no buyer comes. A ride app has no drivers, so riders wait too long and leave, so drivers earn nothing and quit. The very feedback loop that makes a mature platform unstoppable makes an infant platform stillborn.

This is the chicken-and-egg problem, and it is the primary reason platform businesses fail. Not bad technology, not a weak idea, they simply never get both sides spinning the flywheel, and an empty platform is worthless to everyone, so everyone leaves.

The entire early strategy of any platform is one question: how do you get the flywheel turning from a dead stop, when the thing that will eventually make it valuable, the other users, does not yet exist? Everything else, the technology, the features, is secondary to solving this. A brilliant platform nobody is on is just an expensive empty room.

Platforms mediating two groups are two-sided markets, studied formally by Jean-Charles Rochet and Jean Tirole, and their pricing follows rules that look wrong by ordinary business logic.

In a normal business you price each product above its cost. In a two-sided market the platform sets a price for each side, and it will often deliberately price one side below cost, or free, or even pay them to participate, and recover everything from the other side. Charging one customer nothing and their counterpart a lot is not a mistake here; it is the optimal design.

The reason is that the two prices are not independent. Cutting the price on one side brings more of that side, which makes the platform more valuable to the other side, who will then pay more. So the platform is not pricing two products; it is tuning a system, and the right question is not "what does each side cost to serve" but "which side, made cheaper, most increases the whole platform's value".

This is why so many digital services are free to one group. The free users are not charity and not a loss leader in the usual sense. They are the inventory that the paying side is actually buying access to.

If you subsidize one side and charge the other, the pivotal decision is which is which. Get it backwards and the flywheel never starts. A few principles decide it.

Subsidize the side that is:

• More price-sensitive. The side that will bolt over a small fee should be the free one; the side willing to pay should carry the cost.
• Harder to attract but more valuable once present. Often you pay to seed the scarce side that the other side actually comes for.
• The one that creates value for the other. Content creators make a platform worth visiting for viewers, so you court creators and monetise the audience their content assembles.

The classic pattern is to make the money side subsidise the value side. Media platforms are free to readers and charge advertisers, because readers are the value advertisers pay to reach. Many marketplaces lean on sellers and keep buyers cheap or free, because a marketplace thick with buyers is one sellers will pay to be in.

There is no universal answer; it depends on which side is scarce, which is sensitive, and which draws the other. But the question is always the same, and answering it wrong is one of the most common ways a two-sided business dies with a good product and an empty room.

Knowing whom to subsidize does not by itself put the first users on the platform. Several concrete tactics recur, each a different way to fake or force initial scale.

• Single-player value first. Build a tool that is useful to one side alone, before any network exists, then add the network once that side is present. A tool people use solo has no chicken-and-egg problem, and the users it accumulates become one side of the eventual marketplace.
• Fake one side. Provide the missing side yourself at the start: seed a marketplace with your own inventory, or a content platform with content you commissioned, so early users find it non-empty.
• Come for one thing, stay for another. Attract users with a standalone feature, then reveal the network once they are in the door.
• Marquee users. Land a few high-profile members of one side whose presence alone pulls the other.

The common thread is that you cannot wait for a two-sided network to appear on its own, it will not. You manufacture the appearance of a working platform, by hand and at a loss, long enough for the real flywheel to catch. Almost every large platform has an unglamorous origin story of exactly this kind of brute-forced early scale.

A counter-intuitive tactic deserves its own step, because it contradicts the instinct to launch everywhere at once.

The instinct is wrong for network effects. Spreading a fixed number of early users across the whole world means every individual market is empty: a hundred drivers spread over a country is nothing everywhere; a hundred drivers in one neighbourhood is a working service. Since the network effect is often local, the value is only felt by people close enough to actually match, launching narrowly concentrates users to the point where the effect ignites.

So the winning move is frequently to dominate a tiny market first: one city, one campus, one niche. Reach the density where the platform genuinely works there, let the flywheel spin in that small arena, then repeat the playbook market by market.

This reframes what "scale" means for a platform. Total users is the wrong metric early on; density within a market is the right one. Ten thousand users spread thin is a collection of dead platforms; a thousand users packed into one market is a live one that can then be cloned outward. The general principle: for a network-effect business, win small and dense before you try to win big and thin.

Even after ignition, a two-sided platform has to keep the sides in balance, and imbalance in either direction is corrosive.

Too many sellers and too few buyers, and sellers get no sales and leave. Too many buyers and too few sellers, and buyers cannot find what they want and leave. The platform's ongoing job is not just growth but matched growth on both sides, which is genuinely harder than growing a one-sided business, where you only manage a single funnel.

This creates a subtle ongoing tension. Growth on the two sides rarely arrives at the same rate, so a platform is constantly leaning into whichever side has fallen behind: throttling one, incentivising the other, adjusting prices to rebalance. Manage a marketplace and you are perpetually fixing an imbalance rather than simply pushing a single growth number.

The deeper point is that a two-sided business is a balancing act, not a growth chart. A one-sided company wants one number to go up. A platform wants two numbers to go up together and in proportion, and the moment they diverge, the weaker side starts to leave, which pulls the stronger side down after it. Health is the ratio, not the total.

Pull the pieces together and a striking conclusion follows: for a network-effect business, the cold-start period is often the whole game.

The logic chains cleanly. Network effects create winner-take-most dynamics, covered next. Winning depends on igniting the flywheel first and fastest. Igniting it means solving chicken-and-egg through subsidy and brute-forced early density. Therefore the company that solves the cold start best usually wins the market, and the advantages compound from there into a lead that is very hard to reverse.

This is why platform competition looks so different from ordinary competition. Two normal businesses can coexist for decades, each improving its product. Two platforms racing to ignite the same network effect are often in a fight where the first to reach critical mass takes most of the market, because their growing network becomes the reason the other's users defect. Being second-best at scale is a fine outcome in most industries and close to fatal in a strongly-networked one.

So the enormous early spending platforms are famous for, subsidising rides, giving away the product, paying to seed both sides, is not irrationality or vanity. It is a rational race to critical mass, where the prize for finishing first is a market that then defends itself.

The cold start is a dead stop: each side waits for the other. The escape is to manufacture one side, concentrate users into a dense small market, and subsidize whichever side unlocks the flywheel.