Skip to content

SYSTEM Cited by 1 source

M-Pesa

M-Pesa is Safaricom's mobile-money platform, first launched in Kenya in 2007 and now a continent-wide production payments system. From a distributed-systems perspective it is the flagship instance of the patterns/feature-phone-frontend shape: systems/ussd menus at the user edge, a modern cloud payment platform behind.

This wiki's knowledge of M-Pesa is derived from Werner Vogels' October 2025 All Things Distributed post; the entry exists primarily as an anchor for its cross-references in concepts/appropriate-technology and patterns/feature-phone-frontend rather than as a deep architecture page.

Scale (per Werner, 2025)

  • > $100 billion processed in 2024.
  • 4,000 transactions per second in the production hot path.
  • Real-time ML-based fraud detection in the transaction path.
  • Cloud infrastructure on AWS.

(Source: sources/2025-10-29-allthingsdistributed-what-is-ussd-and-who-cares)

User edge

USSD menu dialed on any GSM feature phone over 2G. Typical flows:

  • Send money to another number.
  • Pay a merchant / bill.
  • Buy airtime.
  • Check balance.

See systems/ussd for the edge protocol.

Backend

The post names only the architectural shape, not specific services:

  • AWS cloud infrastructure (operator: Safaricom).
  • Real-time ML fraud detection somewhere in the transaction path (pre-authorize or near-synchronous).
  • Sufficient backend to sustain 4K TPS with fraud-scoring in the latency budget.

No specific service names, data-plane designs, or consistency properties are disclosed in this source.

Peer systems named in the same post

  • Moniepoint (Nigeria) — 5.2B transactions / $150B in 2024.
  • Mukuru — cross-border remittances.
  • OPay — multi-service mobile wallet.

All share the USSD-frontend + cloud-backend shape.

Significance in this wiki

M-Pesa is the canonical worked example whenever we discuss:

  • concepts/appropriate-technology — the customer has a feature phone; the frontend matches; the backend is modern. Not a compromise, a correct design.
  • patterns/feature-phone-frontend — thin edge, sophisticated backend, scaling in the backend where the scaling is cheap.
  • The long tail of the 3G era: GSMA forecasts ~1/3 of Sub-Saharan Africa connections will still be 3G in 2030, so the user edge remains a meaningful design constraint for at least a decade.

Seen in

Last updated · 200 distilled / 1,178 read