Rails Before Marketplaces

Jumia listed on the New York Stock Exchange in 2019 as the Amazon of Africa, one marketplace meant to sell phones, fridges, and groceries to a continent.¹ By the end of 2025 it had booked $2.2 billion in accumulated losses, and at the close of 2023 it had shut Jumia Food across all seven markets that ran it, Nigeria among them. Bolt closed its food business in Lagos the same month.² The streets those companies were built on kept running on three things that never needed a venture round to work: airtime, the phone call, and the ride to the other side of the city.

The mistake was not ambition or capital; Jumia raised more than almost anyone on the continent. The mistake was order. It opened a marketplace of convenience before the rails a marketplace of convenience runs on had been laid. Amazon did not arrive first in America either. The things that made Amazon possible arrived first, and most of them took decades.

Necessity and convenience

Two kinds of marketplace sit at opposite ends of how badly you need them. Amazon marks the convenience end: you reach for it because it beats the shop down the road, but the shop is still there, and on any given day you can skip the order. The game is discovery, the substitute always at hand. Ride-hail marks the other end. In Lagos a private car waits behind a wall most households never clear, on the order of forty vehicles per thousand people against more than six hundred in the United States, and the bus does not reach across a city that doubled while no one was building rail.³ You cannot skip the trip to work, so you take whatever moves you.

The difference is elasticity. Raise the price of a convenience and the customer walks to the shop; raise the price of a necessity and they still have to get to work, so they find the money. How much the customer can refuse decides which marketplace a place with none of them can build first.

The behavioral loop

Emerging markets are full of proof that infrastructure can skip a generation. Kenya put money on phones before it put branches on corners, and the bank branch was simply skipped. The leapfrog is real for wires, towers, and rails, but it stops at behavior.

A habit is built only by repetition. You send money through a phone, it arrives, the person on the other end confirms it, and the next time you trust it a little more. Trust compounds one cleared transaction at a time, and no marketing budget can pay it forward. W. Brian Arthur described how small early leads lock in and compound under increasing returns, the same dynamic that makes a recycled SIM worth more than a credit bureau.⁴ An established habit is just as hard to dislodge, and just as slow to build, because the loop only turns as often as the customer comes back.

Frequency is the forge. A daily necessity runs the loop every day and sets the habit before the funding runs out. A monthly convenience runs it twelve times a year and burns through its money waiting for trust to harden. M-Pesa looks like the counterexample, a leap from almost nothing to 35.8 million active users moving 38 trillion shillings a year.⁵ It was nothing of the sort. Kenyans had spent years buying and sending airtime, learning that value could live on a SIM and move between phones; M-Pesa extended that habit rather than inventing it.

The spine

Before any of these marketplaces, two things had to be in every hand: a way to talk and a way to pay. In Africa they came bundled in the same object and often the same purchase. About 95 percent of mobile subscribers in sub-Saharan Africa are prepaid, topping up in small amounts many times a month, so airtime became the first digital value most people ever held.⁶ It is traded for cash, sent across borders, and spent like currency, to the point that the IMF counts airtime top-ups as a channel for sending money home.⁷

Airtime was the spine, never a marketplace. What rode on it first was conversation. Trade across much of Africa still moves by voice, because a deal closes on a relationship and a relationship is built in a call rather than a contract. Fewer than five percent of firms in sub-Saharan Africa take a dispute to court; the enforcement a courtroom supplies elsewhere is supplied here by knowing the person on the other end.⁸ So the phone carries the relationship first and the money second, and that sequence is the ground everything else stands on. Voice still earns its keep: in Kenya the average subscriber talks around 200 minutes a month, and voice and messaging make up a quarter of Safaricom's service revenue.⁹

Ride-hail and Jumia

Put the two side by side in the same cities and the order shows itself. Ride-hail arrived in African cities between 2013 and 2016 and met demand it never had to manufacture. Uber alone passed a billion trips on the continent by 2022, up from fewer than two million riders in 2017 to more than thirty million.¹⁰ Crossing Lagos is not optional, a car is out of reach, and the bus does not come, so people took whatever filled the gap. The demand was inelastic, the loop turned daily, and paying a stranger through a phone to drive you became ordinary because it had to.

Jumia ran the experiment backwards. It opened a marketplace of convenience, physical goods carried to a door, before the rail that carries them existed: no dense and cheap logistics network, no settled habit of paying upfront for a box that lands days later, no addresses across much of the city. Capital did not fix the order, it funded the wrong sequence for longer. The food businesses that held on in Nigeria, Chowdeck among them, did it by keeping unit economics tight and growing only where the loop could close, not by outspending the gap.¹¹ Transport could be built first because you cannot walk to work. Commerce had to wait, because you can always put off the parcel, and the financing loop is what later turned the rides themselves into a renewable asset instead of a one-time burn.

The Brazil test

The obvious objection is income. Brazil is richer than Nigeria, so of course delivery worked in São Paulo and stalled in Lagos. The objection breaks on the timeline. iFood launched in Brazil in 2011 and now clears more than a hundred million orders a month, but it did not land on empty ground.¹² MercadoLibre had run since 1999 and Mercado Pago since the early 2000s, drilling one habit into Brazilians for a decade: pay online and the thing arrives.¹³ Account ownership climbed from 56 percent of adults in 2011 to 84 percent in 2021.¹⁴ The rail went down first, and only then did convenience scale on top of it.

Brazil had a second rail that money cannot buy on a schedule. The country had eaten in self-service, pay-by-weight restaurants since the 1980s, when the first comida a quilo spot opened in Belo Horizonte and the format spread nationwide within the decade.¹⁵ A Brazilian already trusted that food cooked by a stranger and sold by the kilo met a known standard. The expectation that outside food was safe and consistent was in place long before an app asked anyone to order it. Lagos in 2015 had neither the payment rail nor that demand habit, and no subsidy installs a decade of behavior in a quarter.

The same order shows up in the rich world, only earlier. Carlota Perez, reading two centuries of technological revolutions, found that each one installs its infrastructure before its golden age of applications begins.¹⁶ America laid the delivery rail first. FedEx flew its first overnight run in 1973, 186 packages to 25 cities, and spent two decades making "a stranger will bring my package tomorrow" an ordinary thing to believe.¹⁷ Amazon launched in 1994, onto a country that already trusted the parcel.

Bits before atoms

If necessity decides which marketplace comes first, physics decides which kind. A marketplace made of bits scales the moment the spine is there, because nothing has to be carried. A marketplace made of atoms waits on the slowest rail of all, the one that moves physical things: trucks, warehouses, addresses, and the belief that a box will arrive.

Nigeria proves it in film. The country has roughly ninety cinemas and three hundred screens for more than two hundred million people, and those cinemas sold about 2.66 million tickets in the whole of 2024.¹⁸ The theatrical rail was never built, so the film market routed itself onto the one rail that already reached every phone. Nollywood, by output the second-largest film industry in the world, now releases full features straight to YouTube as their first and main window.¹⁹ Omoni Oboli put Love in Every Word online for free in March 2025 and watched it pass 32 million views inside a year; channels like Ruth Kadiri's carry millions of subscribers and hundreds of millions of views.²⁰ No cinema, no disc, no shelf. The content rode a rail Google had already paid to build, drawing on the oldest demand there is, the wish to watch something.

Nigeria is not alone in putting films on YouTube; India's dubbing channels are larger. What makes the Nigerian case sharp is that YouTube is the primary first-release window for new, original commercial films, where the giant channels elsewhere mostly re-run libraries that already had a cinema or television life. Bits scaled first because they skip the rail that takes longest to lay; atoms are still waiting on it.

What to build first

The lesson for anyone building in these markets is a rule for choosing, not a product to chase. An operator's edge lies less in building a necessity than in picking the right one, the necessity where the behavioral loop is cheapest to close. Two conditions make it cheap. Either the demand is inelastic enough to force the habit on its own, the way crossing a city does, or the habit already exists, the way Brazil's kitchens had taught a country to trust outside food. Find one of those and the loop closes on the customer's own momentum instead of on your balance sheet.

After that, build bits before atoms. The marketplace that carries nothing crosses first; the one that carries boxes waits for roads, addresses, and delivery trust. And when the time comes to move atoms, lay that rail on purpose, the way FedEx did, rather than opening the store and hoping the logistics appear. Within a single marketplace the hard question is which side to build first. Across marketplaces, the harder one is which marketplace the rails can carry at all. Jumia answered that one last. The order is the strategy.