Big data continued to win the buzzword popularity contest in 2015. Big data for development (D4D) has been the darling of the tech for development community in recent years, yet there is a continuous challenge in matching the demand side (development community, researchers, analytical companies) with the supply side (the mobile network operators who have the data). In many ways the hype has also misled people into assuming that preparing and analysing this data is an easy thing to do- it’s not.
It is hard to find meaning and value in the data- it requires a niche set of skills, a clear problem statement, trust between partners and the right regulatory conditions. There are other good reasons for a measured approach: customers entrust mobile operators with their data, so privacy and security of mobile users personal data is paramount. As big data for development is a relatively new area of inquiry and application, efforts must be made to mitigate risks. However we have seen promising examples when the stars align and there is the right mix of potential for social impact, a shared willingness and set of capabilities and assets among partners and a clear research question and social problem to solve.
A recent partnership between Telenor Research, in cooperation with the Harvard T.H. Chan School of Public Health, Oxford University, the U.S. Centre for Disease Control, and the University of Peshawar have demonstrated how Big Data can be harnessed to anticipate and track the spread of dengue fever.
The published study, “Impacts of human mobility on the emergence of dengue epidemics in Pakistan” analysed anonymized call data from more than 30 million Telenor Pakistan subscribers during the 2013 dengue outbreak, using the large volume of data to accurately map the geographic spread and timing of the epidemic. I had the opportunity to speak to Kenth Engø-Monsen, Senior Data Scientist at Telenor Research and Mai Oldgard, Senior Vice President, Head of Sustainability to get their insights into the study, the outcomes and the partnerships and how this approach might be replicated.
There are so many areas where the link between call detail record (CDR) data and human mobility can be explored and operators like Telenor receive numerous requests to engage in research- how did you decide to focus on this particular challenge?
Telenor is motivated to help solve social problems and empower communities. It is essential with big data projects that there is a clear problem statement to guide the research, rather than being led by the data itself. From Telenor’s perspective, it made sense to look at dengue fever as it has a wide impact in Asia, where Telenor has a large footprint. The World Health Organisation estimates that half of the world’s population is at risk of contracting the disease, and it is emerging in Pakistan.
Dengue represents a significant public health and development challenge globally, negatively impacting community health and economic growth. It was important to the research team to find a problem where they could develop something meaningful and impactful from assets and expertise that were already in place- making a social contribution driven from their core competency that aligned with their vision of empowering societies in the markets where they operate.
It was also important to Telenor that the people who work on this challenge every day- the Ministry of Health and others, could actually make use of the big data tools that were developed and in future benefit from the insights that real time big data can provide. Telenor takes the approach that big data alone will not solve the problem, but that it can be used to strengthen or support existing approaches.
How did the process begin?
Patience was important- the project itself took two years from inception to the publication of the study in September 2015. Several elements have to fall into place to ensure a successful project. In the first year, negotiations between Telenor Research and Telenor Pakistan were underway to gain support for the project. Much of the focus was placed on due-diligence to ensure that the project was undertaken inline with privacy principles and regulatory requirements. Time was spent resolving legal questions and solving issues, as well as assembling the right project teams and partners for the kick-off.
Controlling for privacy and risk and meeting with regulatory requirements were a priority for the team- what approvals were necessary and what is the potential for replicability of the methodology used?
There were on-going discussions about how to process the CDR data to ensure privacy. It was determined that the best approach would be for the CDR data to remain within Telenor Paksitan’s data warehouse with a clear line of demarcation where only Telenor would have access to the raw data. The Pakistan Telecoms Authority (PTA) scrutinized the project to ensure that it had complied with national regulations and security laws in the country. Trust had to be built with PTA through face-to-face discussion and presentations and a significant amount of due diligence was undertaken- the stakes are high- if we don’t do this properly as a mobile operator, we could have our licence revoked. This is something that it is important for potential partners to appreciate. The methodology used could be replicated and this is the approach we advocate.
What were some of the critical success factors that enabled this project?
Having a framework and a “recipe” for privacy was important. A year ago, we were involved in the development of the GSMA Guidelines for the Use of Big Data for Ebola Response however their application is much wider. We shared our recipe for how we had approached privacy in the research in Pakistan and we felt proud that the GSMA incorporated these principles as a basis. The guidelines suggest that the MNO should handle the data because it is their obligation and expertise- then work closely with data scientists to ensure the data is extracted and processed in the correct ways- it is a resource we need to protect.
It is important to note that the data you want to extract, and the processes required are dependant on the research questions and problem you want to solve- having this clear as a starting point can help ensure that the right aggregates and perspective are captured and that it is done in a privacy preserving way.
The privacy principles give a methodology on how we can do this. It lowers the barriers and builds confidence. Our dengue study is a good demonstration case in line with the existing GSMA principles and this would be a valuable area for mobile operators to work with the GSMA on developing further.
There is a lot of demand from academics, humanitarians and policy makers around the data philanthropy agenda- how do you select partners and what messages do you have for researchers and others who seek to partner with mobile operators on big data for development?
We select partners for their skill set. We know what we bring: analytics capabilities, data expertise. Complementarity and finding the right collaborators has been key to the success. Having a team that had the analytical skills as well as an understanding of the broader business priorities for Telenor was essential to navigating between and within the different stakeholders. Let the MNO handle the data because it is their job and obligation. Good privacy design and working closely with the scientists to ensure the data is processed in the correct ways is essential. A lot of organisations approach mobile operators and request that they dump telco data to outside servers for research and development purposes. There are solutions that don’t necessitate a data dump but the hype has led many to believe that everyone can and should get their hands on CDR data. Define the problem and engage the MNO to help understand how it can help to solve it.
What is Telenor’s long-term vision and next steps for big data for social good?
Gaining support and demonstrating internally and externally the value that this kind of initiative can have will be important to unlocking resources to match future demands. Since we launched this study, we have had a lot of attention within the Telenor Group. We have an epidemic use case that we want to replicate in several countries depending on local legislation. I’m always humble that while we have a report that proves a concept, that’s only part of the pie. Putting it into production so that its running regularly and being integrated into real- time systems to produce the right info is highly ambitious but that is our vision. We need to create a viable system that taps the right info from the telco system in a privacy and security preserving way and produces the necessary information to the stakeholders who need it to improve service delivery, health outcomes and development.
From the GSMA perspective, this study is a great story of collaboration for an excellent cause. It demonstrates the potential and impact of mobile operator Big Data while also laying foundations for a methodology that can be replicated, highlighting how to prioritise privacy while achieving social impact.
For more information, visit:
http://www.telenor.com/media/articles/2015/defeating-epidemic-disease-with-big-data/