Safaricom grew M-Pesa's agent network in lock-step with its customers.

M-Pesa lets people store and transfer value on a phone, but that value is only useful if a user can turn e-cash into real cash nearby, so the agent network is what made the product work.

Safaricom grew agents in step with subscribers rather than skimping, recruited them for aptitude, trained them on cash-and-e-cash liquidity, and spaced them geographically, so each outlet stayed busy and each customer stayed close.

By 2012 Kenya had roughly 19.5 million mobile-money users and about $8 billion a year transferred, close to a quarter of GDP, five years after launch.

Grow supply exactly as fast as demand

Balanced agent density keeps each outlet profitable enough to maintain its cash float

Proximity matters: users within a short walk transact far more often than those far from an agent

Spreading agents keeps service convenient, which is what makes mobile money stick

An idle or overcrowded agent either stops serving or stops sustaining the float, killing the network

A network is worth nothing until it is dense enough to be convenient: rationed supply and demand to keep both sides healthy, not just one.

M-Pesa became a template for mobile money across Africa and beyond, studied for its operating-model design. Safaricom later opened the agent network to rival operators, and the agent model was replicated in many markets even as regulators pushed for interoperability.