Skip to content

SYSTEM Cited by 1 source

KOKO Networks

KOKO Networks operates an IoT-backed bioethanol cooking-fuel distribution network across Sub-Saharan Africa. The consumer-facing unit is the KOKOpoint, described by Werner Vogels as "an ATM for cooking fuel" — a cloud-connected vending station where customers top up bioethanol using their phone.

From a distributed-systems perspective KOKO is an instance of patterns/feature-phone-frontend applied to physical-goods retail: customer channel minimal (feature phone / USSD / payment rails), backend rich (IoT ingest + inventory state + demand forecasting).

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

Scale (per Werner, 2025)

  • > 700 KOKOpoint fuel stations connected.
  • Each station streams real-time inventory data to KOKO's cloud platform.
  • Demand forecasting in the backend uses this telemetry.
  • KOKO describes the deployment as "Africa's largest deployment of IoT technology for consumer fuels."

Architectural shape

Customer (feature phone)   KOKOpoint station (IoT)
        │                           │
        ▼                           ▼
    Payment                    Inventory telemetry
    rail                       (real-time stream)
        │                           │
        └────────► KOKO cloud ◄─────┘
                        │
                        ▼
                  Demand forecasting
                  + restock routing

The customer dials a short code / pays via mobile money; the station itself is an IoT endpoint pushing inventory state to the cloud, which aggregates across all stations for supply-chain decisions.

Why it matters in this wiki

KOKO is the worked example that the patterns/feature-phone-frontend shape generalizes beyond payments. The user is the same Sub-Saharan African customer with a feature phone; the business problem is retail fuel logistics rather than money transfer; the shape of the system is unchanged.

This broadens concepts/appropriate-technology from "mobile money" to "any consumer service in feature-phone-dominant markets."

Caveats

  • The ATD post is a thesis piece, not an architectural deep-dive. Specific cloud provider, IoT protocol, and forecasting stack are not disclosed.
  • No TPS, latency, or volume numbers given — only the >700 station count and the qualitative "real-time" descriptor.

Seen in

Last updated · 200 distilled / 1,178 read