Marshini Chetty, Richard Banks, A. Brush, Jonathan Donner, Rebecca Grinter
Most readers of this magazine live in an always-on, ultra-connected world. Many of us enjoy all-you-can-eat bandwidth at reasonably high speeds. But what happens when your Internet use is palpably constrained? What happens when you only have a fixed amount of bandwidth per month and where every byte you access uses up this precious resource? Or worse, what happens when you have to share that bandwidth pool with the three other Internet users in your household? These are the questions raised when we begin thinking about how users interact with the Internet in metered bandwidth situations. By answering these questions, we can help Web developers, interaction and application designers, and even Internet service providers (ISPs) customize their experiences for situations in which each byte delivered has a monetary value. In this article, we report initial insights from our experiences studying how households in South Africa use metered bandwidth.
To be clear, a metered Internet connection is not the same as a slow Internet connection. The HCI community has already approached the issue of bandwidth speed. For example, researchers have explored how families adjust when internal bandwidth bottlenecks are made more visible  and others, such as M-Lab and Glasnost [2, 3], have focused on how to make external network speeds more transparent to end users. Recent studies also describe how slow Internet speeds affect how users in resource-constrained settings browse and experience the Web, with their interactions being more planned, purposeful, and deliberate . On the other hand, the structure of bandwidth pricing has received less attention. Yet, metered bandwidth, consumed either as a pay-as-you-go per megabyte or in relatively small monthly bundles, is worthy of exploration by the HCI community, because it imposes different constraints to slow internet speeds.
Low-bandwidth caps are common for digital subscriber line (DSL) users in South Africa, Australia, and India, and in some entry-level broadband plans in Canada. Elsewhere, the prospect of lower caps on home broadband is sometimes discussed, as ISPs seek ways to mitigate the high demand for bandwidth by more customers. Most important, with the rapid spread of Internet-enabled mobile devices and wireless modems for laptop data access, more users will be connecting to the Internet via a metered connection than we might have ever imagined. Some will retain the convenience of an all-you-can-eat plan at their home or workplace and will be metered only when using their mobile devices. For others, particularly for resource-constrained users in the developing world, the only means of Internet access will likely be though a metered channel, via a cybercafé, a mobile device, or a metered DSL connection at home.
We have chosen to study bandwidth caps in South Africa because most broadband users in the country have known no other way of accessing the Internet at home; the rituals and practices to cope with shared, metered bandwidth are well understood in connected South African homes. Due to relatively scarce undersea international data connections and a somewhat sleepy domestic market, broadband has been quite expensive. Prices have begun to fall more rapidly in the past few months, as more capacity (and more alternatives, including some all-you-can-eat plans) are coming online, but as of late 2010 the price of broadband was still out of reach for the half of South African homes living below the poverty line , and a considerable expense for those with a middle-class income .
In 2010 most broadband providers offered varied monthly plans in South Africa, with caps ranging from 1GB to 5GB, generally fashioned in a use-it-or-lose-it styleif a user doesn't consume the resource within the month, the balance does not roll over to the next month. If users exceed their allocation, they are moved to a slower connection or their Internet access is terminated. Of course, if they have the means, users can "top up," or buy another gigabyte of data, which is also "use it or lose it" for that month. If not, they must live with limited or no Internet access for the remainder of the month. In some cases, operators offer websites where users can track monthly balances and purchase top-ups. Many will send an SMS or email reminder when the cap is looming. Which plan a household ultimately chooses is related to the gains perceived from investing in a broadband service. More so, plan choices correlate with how affordable the rates are for the caps, on top of the expense of the telephone line which is required to deliver the DSL service.
We set out to investigate how households choose their plans and how these metered plans affect their patterns and norms of Internet use. Questions we aim to answer are: How does Web behavior change or differ when there's a bandwidth cap? What strategies do households use to manage their caps? What information do users require to help them best use their cap for their needs? And finally, how well do users approximate their needs and match them to the best fitting plan? We believe studying how metered usage affects home computing is important for the following reasons:
Caps shape how, when, and for how long people browse. As we saw in work by Wyche et al., constraints can change how users surf the Internet . For instance, Kenyan Internet users in the workplace were more deliberate in their browsing style, planning their online activities before getting on the Web to avoid excess metering charges at Internet cafes. They were also frustrated by the time spent downloading large files and email attachments, often sent by colleagues in better-resourced settings, because of slow speeds and bandwidth limitations. In our work in South Africa, we're noticing the constraint of metered bandwidth similarly affects how people browse the Web. For example, our participants tend to avoid high-bandwidth sites at the beginning of the month so they can stretch their caps to the end of the month, when the meter starts running again.
Caps force people to put a price on their Internet use. Caps cause users to approximate the monetary value of their household's Internet use. In our work we looked at what plans people chose and whether they actually ended up using their allocated bandwidth for the month or consistently went over their cap. By doing so, we seek to understand how well the bandwidth amounts offered by providers match up to the online needs of our participating households. Moreover, we wish to understand how users view and ascertain the value of various types of digital bits and content. By understanding the value that users give to each bit, we can better predict the desire for and choice of certain forms of online content over others, and design accordingly.
Caps create a need for tools that show people how they are using their bandwidth. Although ISPs offered basic per-day or per-month reports on bandwidth consumption, most of our participants were not sure which specific online behaviors were eating into their caps. Yet to help them stick within a budget, avoid "top up" gigs, and have bandwidth last to the end of the month, they require transparency about the bandwidth costs of various sites and actions.
Because of the inherent limitations of last-mile and wireless-access technologies as well as congestion from increased demand by users, applications, and devices, it is likely that metered bandwidth plans will persist. In the mobile space in particular, metered caps ensure a quality of service for everyone and help to control traffic on already-burdened infrastructure. This makes metered bandwidth use an interesting topic of investigation, not only in infrastructure poor settings.
Our fieldwork in Cape Town has taken us to a wide range of homesfrom a massive mansion with sprawling gardens, to a four-room home in which seven people shared a single 3GB monthly connectionand to a variety of situations in between. We are struck by how almost everyone we spoke to was aware of the bandwidth constraint, adopting similar strategies to manage this limited resource, regardless of income level. In this way our research illustrates a way to blur the boundaries between the traditional dichotomies of studies of "household" use in developed markets versus those of "resource constrained" users in the developing world, the focal point of many ICT4D and HCI4D studies.
Moving forward, we suggest the HCl community should consider the implications of metered bandwidth situations on matters of inclusion, usability, and usefulness. Specifically, we need to better understand how a cap affects Internet use, what tools and information people need to help them deal with metered usage, and how existing websites and tools can be more accommodating of people who are on usage-based bandwidth plans. Whether via landlines or, increasingly, mobile phones, metered bandwidth may challenge the freewheeling metaphors that we have allowed to become synonymous with Internet use, such as "browsing" and "surfing." By examining Internet use under this form of constraint, we will broaden the HCI4D discourse by opening up conversations and avenues for research along a continuum of users, beyond the easy dichotomies of developed and developing. In doing so, we may improve the ways in which we design and build experiences for the next billion Internet users.
1. Chetty, M., Banks, R., Harper, R., Regan, T., Sellen, A., Gkantsidis, C., Karagiannis, T. and Key, P. Who's Hogging the Bandwidth?: The consequences of revealing the invisible in the home. Proc. of the 28th International Conference on Human Factors in Computing Systems. (Atlanta, GA, April 1015). ACM, New York, 2010.
2. Max Planck Institute for Software Systems. Glasnost: Bringing Transparency to the Internet; http://broadband.mpi-sws.org/transparency/
3. Measurement Lab; http://www.measurementlab.net/
4. Wyche, S.P., Smyth, T.N., Chetty, M., Aoki, P.M. and Grinter, R.E. Deliberate interactions: Characterizing technology use in Nairobi, Kenya. Proc. of the 28th International Conference on Human Factors in Computing Systems. (Atlanta, GA, April 1015). ACM, New York, 2010.
5. CIA World Fact Book; https://www.cia.gov/library/publications/the-world-factbook/geos/sf.html/
6. The Broadband Bible; http://www.the-broadbandblble.co.za/wordpress/wp-content/uploads/2009/12/TheBroadbandBlble.pdf/
Marshini Chetty is a Ph.D. candidate in the Human-Centered Computing program at Georgia Tech. Her research focus is on home computing in resource-rich and resource-constrained settings.
Richard Banks is a senior interaction designer at Microsoft Research in Cambridge, UK. He spent a decade working on the interface design of a broad set of Microsoft's products before joining the research division in 2006.
A. J. Bernheim Brush is a researcher at Microsoft Research in Redmond, WA. Her research area is HCI with a focus on ubiquitous computing in the home and continuous sensing on mobile devices.
Jonathan Donner is a researcher in the Technology for Emerging Markets Group at Microsoft Research India. His research is focused on the spread of mobile telephony in the developing world.
Rebecca E. Grinter is an associate professor in the School of Interactive Computing at the Georgia Institute of Technology. Her research interests include human-computer interaction and computer supported collaborative work.
©2011 ACM 1072-5220/11/0300 $10.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2011 ACM, Inc.