Mobile Data A Development Use Case

A study funded by Bill & Melinda Gates Foundation shows the way.

Mobile phones have become ubiquitous, used not only by relatively wealthy consumers in developed markets but also increasingly by people in the world’s poorest countries. In 2012, there were 5.9 billion active mobile connections globally which has been forecasted to increase to 7.6 billion in 20171. As the power of mobile devices has increased and their cost has fallen, more and more people around the world have found them to be critical tools that enhance their daily lives.

Mobile devices generate a range of data about their users. Information about identity, location, social patterns, movement, finances and even ambient environmental conditions can be derived from the data logged in mobile systems. As this data is uniquely detailed and tractable, it can capture information not easily found from other sources at a scale that would be difficult to recreate through other means. In particular, mobile is one of the only large-scale, digital data sources that touch large portions of low income populations in developing countries, implying that solutions identified in one market can easily be experimented with in another. While this data is personal and private, if it is analyzed under proper protec-

tions and anonymization protocols, it can be used to enhance the lives of poor people around the world across a range of dimensions.

For example, mobile data has been used by researchers, mobile operators and governments to help plan emergency response after natural disasters, enhance access to financial services for the poor, track the spread of infectious disease, and understand migration patterns of vulnerable populations. Indeed the full range of ways that mobile data can be used to improve the lives of poor people is only beginning to be explored.

Economic development projects require keen insight into people’s lives— their habits and behaviors, their health and prosperity levels, and their needs and aspirations. It is critical to clearly understand the problems of the poor before trying to help fix them. Therefore, having

Mobile devices generate a range of data about their users. Information about identity, location, social patterns, movement, finances and even ambient environmental conditions can be derived from the data logged in mobile systems

granular data that captures the experiences of poor communities, along with the analytical techniques needed to decipher that data, allows researchers and development practitioners to improve the accuracy, effectiveness, and reach of their initiatives. Practitioners in the field of economic and social development can better monitor and track the progress of their programs in almost real time, bring projects to scale at a lower cost, gather rapid feedback from the field, collaborate more effectively with stakeholders, and demonstrate impactful outcomes.

While the opportunity to use mobile data for development goals is increasingly accepted, challenges and barriers remain. Data needs to be available and accessible. It needs to be presented in a format that can be understood and utilized. Operators’ commercial objectives need to be considered. Most importantly, data must be used in a way that does not infringe on data protection rules and an individual’s right to privacy.

Many operators, researchers, and governments have explored ways to

deal with these challenges—some more successfully than others—through anonymization, aggregation, opt-in/opt-out models, regulations, and legislation. The areas of privacy and data sharing are especially critical and are evolving every day, meaning it will take time for a consensus to be found. Still, many of the relevant parties are gradually coalescing around uniform practices.

Application Impact

Mobile data offers a view of an individual’s behaviors in a low-cost, highresolution, real-time way. This provides tremendous opportunities for creative uses in development programs.

The table below depicts the three primary types of analyses, and some of their applications: Ex-post – Evaluation and Assessment (e.g., estimating local wealth via past mobile phone activity); Current – Measurement and Real-Time Feedback (e.g., tracking population movements to understand where to deploy aid workers); and Future – Prediction

Mobile is one of the only large-scale, digital data sources that touch large portions of low income populations in developing countries

and Planning (e.g., predicting where and when liquidity is needed along a mobile money agent’s network) In general, the more predictive the analytics can be, the higher impact the analysis will have, but there are high impact applications for development and commercial practitioners across this spectrum.

To date, most mobile data for development efforts have focused on public health or emergency services initiatives. On the commercial side, the focus has been on customer segmentation for better-targeted marketing and churn reduction. One area where development and business interests meet is in the provision of financial services for the poor, since this is a social good that benefits significant populations in the developing world while also boosting revenues for banks, mobile operators, and other service providers.

Mobile Operators Are Beginning to

Exploit Mobile Data: Operators are increasingly exploring big data analytics to improve operations, develop new products and services, and generate more revenue. At the same time, greater numbers of software vendors and other big data analytics specialists are developing more effective, real-time, and sophisticated techniques and tools for capturing the full potential of operators’ vast data stores.

Two overarching models are emerging, which are not mutually exclusive:

Driving internal capabilities: In this model, the operator uses big data analytics to drive operational improvements, develop better products and

services inside its core businesses, and deliver a differentiated customer experience. Mobile network operators have primarily applied this model to develop better approaches to network optimization, customer retention, and churn reduction. For example, an operator in Rwanda and Ghana worked with third-party analytics consultants on anonymized subscriber data to conduct social network analysis to develop a predictive model that identified potential mobile money users based on their communities.

Creating new products and services:

Many different products or services could be developed relying on the platform of a mobile operator’s dataset. For example, operators could sell insights to third parties (e.g., in the US, Verizon Preci-

Much like the commercial sphere, the development world is only just beginning to understand the full potential of this type of data

sion Market Insights offers measurement solutions for a varied clientele, including media owners, advertisers, and venue owners.) Similarly, AirSage captures signaling data, CDRs, and other network traffic from operators, anonymizes and aggregates it, and provides insights to third parties.

Operators often engage in both models, where the typical evolutionary path is to begin by focusing on driving internal capabilities, and eventually expand efforts to include external opportunities as well.

Even in developed markets, most operators are only in the early stages of exploring the possibility of newly emerging analytical tools, with the majority of applications in the early phases of test and deployment. Facing the challenge of a maturing core business, many operators view monetization of their vast amounts of data as a key growth opportunity but are daunted by the task of managing and extracting value from the data.

Current Uses of Mobile Data in Development Programs

By leveraging rich operator datasets and

state-of-the-art analytic techniques, mobile data can help address a wide range of development needs across finance, agriculture, health, education and other spheres.

Much like the commercial sphere, the development world is only just beginning to understand the full potential of this type of data. The “pilot program uses” Table identifies some of the numerous examples of collaborations between operators and researchers who analyze CDR datasets to provide a window into the activities of a population. For example, the multinational operator Orange organized a D4D (Data for Development) Challenge that encouraged researchers to explore development applications using a modified (to protect privacy) set of Orange’s CDR data on Ivory Coast subscribers. The result was a widely praised competition with over 80 research entries from leading academics and practitio-

No significant program has yet been brought to repeatable scale leveraging mobile operator data for social good purposes

ners that showcased a wide variety of uses for this data.

While there have been a number of interesting research collaborations and some promising proof of concept studies, no significant program has yet been brought to repeatable scale leveraging mobile operator data for social good purposes.

The use cases identified in the above Table demonstrate the scope of what is already possible with mobile data. While they have been proven academically or with a limited pilot, they have yet to be developed for systematic, large-scale use. Here we explore two use cases in greater depth.

Use Case 1: Disaster Relief

Using mobile data to estimate population flows in the wake of natural disasters and emergencies to determine where to send relief.

A natural disaster, conflict, famine, or major epidemic often results in en masse migration of populations from the affected areas. A challenge faced by relief organizations is how to effectively model population movements in such emergencies so that relief efforts can be organized and more effectively deployed. As long as the mobile infrastructure has not been completely wiped out, mobile data can provide the information needed to estimate population movements in near real

time, which can help practitioners optimize the distribution of aid and relief services (See Table for Use Case 1).

Flowminder is a nonprofit entity based in Stockholm that functions as a clearinghouse for aggregating, analyzing, and disseminating mobile phone location data to NGOs and relief agencies during disaster relief and reconstruction efforts. After the Haitian earthquake of 2010, a team from Flowminder and researchers from several US and Western European universities analyzed cell tower data from 2 million SIM cards linked to Haitian operator Digicel to estimate population flows in the wake of the earthquake and a subsequent cholera epidemic. They found that those who left Port-au-Prince after the earthquake did not merely flee chaotically to the nearest “safe” zone, but instead had highly predictable travel patterns. Typically, survivors went to the location where they had spent the most recent Christmas and New Year’s holidays, areas where they had strong social bonds. The cholera outbreak that began just months after the earthquake allowed researchers to validate their finding that people’s travel patterns during more stable times predict their escape routes during crises. The Flowminder team’s work provided strong evidence that estimating population movements during disasters and outbreaks using mobile data can be done rapidly and accurately.

Use Case 2: Financial Inclusion

Increasing access to financial services by using mobile data to generate financial profiles of unbanked persons.

Proponents of financial inclusion are beginning to see mobile data as an excellent way to build financial profiles of people who lack a conventionally documented financial history. Poorer, unbanked people have little to no record of past borrowing behavior and volatile income and expenditure patterns, making it difficult for banks and others to provide them with financial services such as savings products and access to credit. As a remedy, an individual’s mobile usage data can provide proxy indicators, such as airtime usage, top-up history, mobile transaction data, and P2P transactions, to create an alternative financial profile.

The alternative profiling models already in use suggest the wealth of information embedded in mobile subscriber data.

A number of operators and banks have already begun to offer financial products that rely on mobile data indicators. In Kenya, for example, Safaricom and Airtel have formed partnerships with financial institutions to expand mobile money services to include short-term and longer-term microcredit provisioning. A number of aggregators, analytics services firms, and financial services specialists have developed mobile databased automated profiling models and related services.

Future Opportunities to Leverage Mobile Data

The range of potential mobile data for development use cases goes well beyond the pilots that have been explored to date. Conversations with operators and researchers revealed a number of high potential applications which have yet to be tried. previews only a few of these specific possibilities.

Should Mobile Data for Philanthropic Use Be Free?

One issue that often comes up when discussing the use of mobile data for development purposes is whether or not operators should be allowed to charge for the data they share. While many development use cases will no doubt justify some reasonable payment for access to data, there are also cases where the data should be shared for free.

The United Nations Global Pulse has put forward the idea of “data philanthropy,” where operators would have a duty to share data for certain limited uses when the public good is urgent and clear. Global Pulse argues that these cases actually make business sense for a number of reasons:

First, companies should want the best for their clients, if only because when their clients do better, they can afford more mobile services. In many cases, mobile data may hold clues to upcoming problems, from disease outbreaks to agricultural crop failures. Global Pulse points to the following example:

“Imagine you are CEO of that company, and you have just completed construction of a number of costly new cell towers in a region that appears to be a promising market. Unbeknownst to policy makers, many in this community are being affected by an on-going, low-level food crisis. By the time this becomes public knowledge, your new customers are no longer able to afford your services.”

Second, public backlash from refusing to share data could be significant, while, the goodwill generated among the non-profit and public sectors when sharing data could be of significant value. Recent blog discussions and press articles have picked up on the idea.

Future debates will have to work out when data should be shared freely. Postdisaster scenarios seem to top the list, as do predictions of major economic shocks or disease outbreaks, but there are other uses where the mandate to share is less clear (e.g., ongoing monitoring of food prices).