The launch machine, and what it is built to measure

What the machine actually does

Rocket Internet's operating method was not, at its core, about growth. It was about signal extraction. The Berlin-based company built to clone and accelerate proven internet business models in markets where the original had not yet arrived, deploying capital, engineers and a standardised operational playbook simultaneously across multiple countries. The point of doing it simultaneously — rather than sequentially — was not efficiency. It was comparison. When you launch the same product in a dozen markets on the same week, with roughly equivalent spend, you get something close to a controlled experiment: the markets that work tell you what conditions the model actually requires, and the markets that fail tell you what it cannot overcome.

Tripda launched across roughly a dozen countries in 2014, concentrated in Latin America, Southeast Asia and Germany. That roster was not random. It was a test of a specific hypothesis: that intercity carpooling could reach liquidity in markets where either the train was slow and expensive, or the coach network was thin, or household car ownership was rising fast enough to produce a supply of drivers with empty seats. Brazil was a named market. Germany was another. Those two countries alone contain almost the entire range of conditions the model might face — a large middle-income country with an underbuilt intercity rail network on one side, and a dense, deregulating European transport market on the other.

The arithmetic of corridor liquidity

Liquidity in a carpooling marketplace is not a vague concept. It has a precise meaning at the corridor level: enough drivers departing a given city pair on a given day that a passenger searching that route finds a realistic choice, and enough passengers searching that a driver posting a seat expects to fill it. Below that threshold the market does not clear — the driver posts and waits, the passenger searches and leaves, and the platform records not a transaction but an abandonment.

The machine is built to find that threshold fast. Each market launch generates three numbers that matter: posting rate (drivers listing seats per week on a given corridor), search rate (passengers querying the same corridor), and match rate (the fraction of searches that produce a booking). A corridor where posting and search rates are both low is simply absent — there is no market to build. A corridor where posting is high but match rate is low signals a supply-demand imbalance that better marketing on one side might correct. The interesting corridor is one where match rate climbs week-on-week without additional paid intervention: that is organic liquidity, which is the only kind that produces unit economics worth caring about.

Average occupancy on a long-distance car journey sits close to one and a half people, which means drivers are already absorbing the fixed cost of the trip regardless of whether any passenger sits beside them. The marketplace is trying to move that number toward the physical limit of the car — typically four seats — by monetising slack that already exists. The arithmetic is not complicated, but it is ruthless: a market where car ownership is too low has no slack to monetise, and a market where the population of inter-city travellers is too small never produces enough search volume to clear even a single corridor.

What gets closed and why

The eighteen-month window Rocket Internet gave Tripda was not arbitrary sentiment. It is roughly the minimum time needed to let network effects either ignite or fail to ignite on a corridor. In the first three months, posted supply is thin and match rates are artificially low because the platform is paying drivers incentives to list. In months four through nine, incentives are wound down and organic behaviour is observable. By month twelve, the corridors that will reach self-sustaining liquidity have usually shown their trajectory; the ones that will not have plateaued at match rates too low to cover cost of acquisition for either side.

The close signal is not a single bad number — it is the absence of a positive trend in organic match rate combined with a customer acquisition cost that does not fall. When the cost of acquiring a new driver stays flat or rises while the probability that driver completes a match stays flat or falls, the market is not learning. A marketplace that does not learn as it grows is not a marketplace — it is a subsidy programme.

Germany illustrates one specific failure mode the arithmetic cannot cure. The 2013 liberalisation of the German long-distance coach market created a legal, cheap, frequent alternative at roughly the same price point as a shared seat, operated by carriers with more reliability guarantees than any reputation system could replicate. In that environment, carpooling is competing on a dimension — price — where the coach is at least its equal, and losing on dimensions like schedule reliability and simplicity. BlaBlaCar had already established itself in France before coach deregulation became a German complication, which meant it had brand and liquidity as defensive assets. A new entrant launching into Germany in 2014 faced a deregulated coach sector as an incumbent competitor, not a gap.

Brazil presented a different problem. Car ownership was rising through the 2000s and into the 2010s, which in theory produces driver supply. But intercity distances in Brazil are enormous, the population of travellers on any single corridor outside the São Paulo–Rio de Janeiro axis is thinly distributed, and the density problem — not enough drivers on one corridor on one afternoon — is geometrically harder to solve when a country is continental rather than European in scale. High car ownership and large distances do not automatically produce liquidity; they produce potential supply spread across too many corridors to concentrate.

The measurement is the product

What emerges from the Rocket Internet method, applied to intercity carpooling, is a clear taxonomy of market conditions: the corridor that ignites organically, the corridor that requires permanent subsidy to simulate liquidity, and the corridor that never produces enough volume to be worth the question. The machine is designed to sort markets into those three categories before the subsidy bill becomes the business model.

That sorting is itself the intellectual output of a multi-market simultaneous launch, and it is genuinely useful independent of whether any single market survives. The liquidity threshold for intercity carpooling — the minimum viable posting rate, the minimum viable search depth, the price elasticity relative to coach and rail — is a piece of market knowledge that a single-country launch could not generate in comparable time. The machine is expensive. It is also, as a measurement instrument, faster than any alternative. What it measures is not whether a product is good. It measures whether the conditions that make a product work exist in a given place, right now, at a cost the economics of the model can bear.

That is a narrower question than ambition, and a more answerable one.

Simultaneous multi-market launches are an instrument: the metric is which corridor reaches liquidity, and everything else is closed.