Reaching the Offline Majority

The march toward a connected world has been deeply uneven. For all the talk of a borderless internet, a vast share of humanity remains structurally excluded — not only those beyond the reach of any signal, but a far larger group who live under coverage yet cannot use it, lacking a smartphone, affordable data, reliable electricity, or the literacy a screen demands. This review asks a practical question on behalf of those people: setting aside the dream of universal broadband, what actually works to educate them now, with the tools they already hold? The evidence points to two answers, and both are humble.

1. The scale of the divide

According to the International Telecommunication Union, the United Nations agency for digital technologies, an estimated 2.6 billion people — about 32 percent of the world's population — remained offline in 2024, down only modestly from a revised 2.8 billion the year before.¹ The ITU's 2025 figures put the number still near 2.2 billion.² The divide tracks poverty almost exactly: in high-income countries roughly 93 percent of people are online, against only about 27 percent in low-income countries.¹

The usage gap is bigger than the coverage gap

It is tempting to assume the problem is missing cell towers. It is not, mostly. Industry analysis from the GSMA distinguishes the coverage gap (no signal at all) from the much larger usage gap (living under a mobile network but not using mobile internet).³ The usage gap is now the dominant form of exclusion, driven by the unaffordability of internet-capable devices, unreliable electricity to charge them, the cost of data, and gaps in literacy and digital skill. The implication is decisive: building more towers will not, by itself, reach these learners. Any solution must meet them on the hardware and conditions they already have.

2. Tutoring by ordinary phone

The most striking evidence concerns the simplest tool: the basic feature phone, which even very poor households often own. During and after the pandemic's school closures, researchers tested whether brief, structured tutoring delivered over plain phone calls could teach foundational skills.

The landmark result comes from a set of randomized controlled trials across five countries — India, Kenya, Nepal, the Philippines, and Uganda (building on the "ConnectEd" model first trialed in Botswana⁴). Pairing targeted instruction with short weekly phone calls produced average learning gains of 0.30 to 0.35 standard deviations — large by the standards of education research — with similar results whether government teachers or NGO staff made the calls.⁵ Because it needs no smartphone or data, the approach proved extraordinarily cost-effective, delivering up to four years of high-quality instruction for every $100 spent, ranking in the top percentile of evaluated education programs.⁵

A separate randomized experiment across 200 villages in rural Bangladesh ("telementoring," in which volunteers mentored mothers and tutored children by phone) found treated children scored about 0.75 standard deviations higher on a standardized test, with the gains persisting a year later.⁶

Automating the call: Interactive Voice Response

Live tutoring is powerful but labor-bound. Interactive Voice Response (IVR) — automated, pre-recorded audio lessons a caller navigates with the phone's keypad — trades some warmth for near-limitless scale. A randomized trial in Bangladesh delivered IVR audio lessons through a toll-free number to 1,763 primary-school children across 90 villages and improved literacy and numeracy scores by 0.60 standard deviations, with the largest benefits going to academically weaker children and those with less-educated caregivers — exactly the learners most systems leave behind.⁷

3. The offline "server-in-a-box"

Where richer material is needed — video, interactive practice, a whole digital library — but no internet reaches, a second family of tools applies: low-cost micro-servers such as Kolibri, RACHEL, Kiwix, and Internet-in-a-Box, often run on a Raspberry Pi. Each broadcasts its own local wireless network, letting any nearby phone or laptop browse a large store of open educational resources entirely offline, consuming no mobile data.⁸ Their value is structural: they conform to local constraints rather than fighting them, and they keep working when connectivity is absent — or deliberately cut.

The peer-reviewed evidence here is promising but more modest than for phone tutoring, and honesty requires saying so. A study of offline servers and laptops in ten Indigenous schools in Guatemala ran for only a single, pandemic-shortened year — too short to establish statistically significant gains in test scores. What it did document, qualitatively, were consistent advantages: more standardized and differentiated instruction, more chances to practice, higher engagement, and access to quality resources where printed materials were scarce.⁹ Broader reviews of educational technology in low-income settings reach a compatible conclusion: the hardware helps most when paired with trained teachers and good content, not on its own.¹⁰

4. What the evidence asks of us

Three conclusions follow. First, universal broadband is not coming soon enough to wait for; more than two billion people need a path to learning today. Second, closing the coverage gap will not be enough, because the usage gap — born of poverty, devices, and electricity — is larger. Third, the interventions that fit these realities are not speculative. Feature-phone and IVR tutoring deliver some of the most cost-effective learning gains ever measured (0.30 to 0.75 standard deviations across trials), and offline servers provide a durable, internet-free bridge to high-quality material.

This is the reasoning behind Reach Them All — a free phone tutor for the learner who is alone with a question, and an offline box of light for the village a signal never reaches. The science says both can work. Whether they reach anyone is, finally, a question of will.