Summary

  • Michuki Mwangi helped carry a shared African Internet community vision into IXP programmes and public explanation; the evidence does not make him the sole author or owner of the goal.
  • The familiar 80% is not a single comparable result: the 2010 aspiration, a two-country local-traffic estimate, a CDN’s own locally served share, an IXP’s peak Gbps and a later count of local websites have different populations and denominators.

Analysis

Three different statements can all carry an “80” without measuring the same thing. The 2010 Africa 80/20 vision aimed for 80% of Internet traffic in Africa to be exchanged locally, with 20% routed from outside the continent. A 2020 chart in an Internet Society study showed a large CDN serving an increasing share of its traffic locally in Kenya and Nigeria, with the lines passing 80%. In the same year, Michuki Mwangi described IXP building as “80% human engineering” and 20% network engineering. One was a regional traffic ambition, one a provider-specific ratio, and one a memorable way to describe the work of getting competitors to cooperate.

Their shared numbers invite comparison; their units refuse it.

That distinction does not diminish Mwangi’s role. The Internet Society biography places him in the work: he joined the organization in 2008, had managed the .KE registry, led AfTLD and volunteered technical support for the Kenya Internet Exchange Point (KIXP). Its 2014 annual report associated him with the 80/20 vision and described work with the African Union and AfriNIC on the African Internet Exchange System (AXIS). His own 2020 account begins with the peering community’s aspiration, not a personal invention.

The more accurate portrait is a technical convenor and advocate within a coalition of operators, IXPs, institutions and public bodies.

The goal itself was easy to remember and hard to operationalize. Internet Society sources variously described 80% of African traffic as “locally accessible,” “locally exchanged,” or not routed from outside the continent. Those phrases overlap, but they do not by themselves specify whether the numerator is traffic exchanged at an IXP, traffic served from an in-country cache, or traffic whose full path remains inside Africa. Nor do they settle whether the denominator is all traffic consumed on the continent, traffic generated within each country, or a sample of popular services.

A target can coordinate advocacy before it becomes a measurement protocol. Trouble begins when the headline survives but the protocol is left implicit.

The 2020 Kenya and Nigeria study is valuable precisely because it names a bounded comparison. Commissioned by the Internet Society and authored by Michael Kende, it revisited two leading exchanges, KIXP and IXPN. Its summary says each country moved from about 30% localized traffic in 2012 to nearly 70% in 2020. It also describes the institutional work around those changes: multisite exchanges, carrier-neutral data-centre nodes, more content caches and points of presence, removal of mandatory peering, and the trust needed for networks to make selective agreements. Mwangi’s account foregrounds that same human layer.

An IXP is a switch and a community of firms willing to connect to it.

The report’s other headline numbers belong in separate columns. KIXP’s peak traffic rose from 1 Gbps to 19 Gbps; IXPN’s rose from 300 Mbps to 125 Gbps. Those are exchange throughput rates, not the fraction of all Internet traffic localized in either country. The study also estimated annual savings of US$6 million in Kenya and US$40 million in Nigeria. Those are reported case-study estimates, not a continent-wide balance sheet. Together the figures show meaningful change in two ecosystems, but they cannot be divided into one another to manufacture an African 80% score.

Figure 3 makes the denominator problem visible. Its source is a single large CDN. It plots the percentage of that provider’s traffic served locally after it deployed points of presence in Kenya and Nigeria. The local share climbs from zero and later exceeds 80% at points in the series. That is useful evidence about one provider’s infrastructure and traffic mix. It is not the proportion of every byte used by Kenyans or Nigerians that stayed in-country, much less the proportion of all African traffic exchanged on the continent. The chart title contains the word “traffic”; the chart’s source note contains the limit.

The Internet Society’s 2021 overview widened the frame without turning it into a single traffic-weighted continental series. It counted 19 African IXPs in 2010 and 46 in 2020. It counted countries with more than 70% localized traffic: one in 2010, South Africa, and three in 2020, adding Kenya and Nigeria. The same overview said most countries still had to raise local exchange from roughly 20% toward 80%. The country count is a useful measure of breadth and a reminder that two striking case studies do not stand for a continent. It is not the share of Africa’s total traffic that was local.

By 2024 the Internet Society was describing a 50/50 vision for selected economies: at least half of locally generated traffic should remain local by 2025. That is a new formulation, not proof by itself that the old 80/20 ambition was either achieved or abandoned. Its measurement note explains a persistent problem: operators may know which services use the most bandwidth, but traffic-volume data is rarely public.

Where sFlow data is unavailable, the proposed method looks at the top 1,000 websites in a country, estimates whether each is served locally, and compares the count of local with external sites using technical clues and local vantage points.

That proxy can illuminate hosting and content distribution. It does not count bytes. A low-volume local news site and a high-volume video service each contribute one website to a simple site ratio; actual traffic weights them differently. The method is useful if reported as a website-locality indicator. It should not silently become a measured share of all traffic. The distinction is not pedantry: a city can improve a proxy while most bytes remain on foreign platforms, or carry substantial local bytes through a few services that a top-site count underweights.

There is another operational boundary. In a 2026 Internet Society essay, Hanna Kreitem warns that remote or cross-IXP peering can make a network appear connected to a local exchange while traffic still travels through distant infrastructure. That is an attributed current argument, not a retrospective measurement of Kenya’s 2012 baseline. It nevertheless sharpens the test: membership, a port, an ASN or a “local peering” label is not enough to prove that the relevant path stayed in-country. The path and the content-serving location matter.

So what did Mwangi’s 80% accomplish? It gave operators, technical communities and public institutions a compact reason to convene around local interconnection. The Kenyan and Nigerian cases document growth, while the 2021 overview shows that progress was uneven across countries. The evidence cited here does not supply one comparable, traffic-weighted, continent-wide series with which to certify that the 2010 target was reached—or failed—as a single African result. Refusing that verdict is not refusing the achievement. It keeps the achievement attached to the systems that were actually measured.

The next target should travel with its measurement receipt: which countries and networks are in scope; whether the denominator is all consumed bytes, locally generated bytes or selected websites; what “local” means for caches and full routes; where and when observations are taken; how missing traffic data is handled; and how the estimate changes when the method changes. Keep peak exchange throughput, website-locality, locally served CDN share and country counts as distinct indicators. A target can rally a community. Only a stated and repeatable method can tell it what has changed.

Sources