How Big Data and AI saved lives in Indonesia
A real-time big data and artificial intelligence platform that allows policy makers and citizens to understand the levels of physical distancing, movement, and mobility at the village level

Muri is outside the Posyandu (local health centre) near her home in North Jakarta. It’s been closed for services since the start of the COVID-19 pandemic. In the seven months since the first detected case of COVID-19 in Indonesia, more than 222,000 Indonesians have contracted COVID-19, with 8841 reported deaths.
Carrying her 9-month-old daughter, Kinara, Muri tells us about the very real cost, emotional and financial, her family is paying to help slow the spread of this global pandemic.
“I’m scared, because we don’t know people’s condition, so I’m scared to go places,” she said, “My wish is for the COVID to end, my husband to be able to get work as usual again, his salary can get back to normal."
In the first year of the COVID-19 pandemic, with no medicine or vaccine available, governments around the world were pursuing non-medical interventions. According to the best science available, handwashing, mask usage and maintaining physical distance were the most effective tools we had to combat the spread of COVID-19. While the impact of the virus was and is devastating, the implementation of these containment policies carried their own cost. With more than 500,000 schools closed across Indonesia at the peak of the PSBB (Pembatasan Sosial Berskala Besar- Large-scale social restrictions), children were shouldering the cost of controlling this pandemic.
In economies like Indonesia’s, where large segments of the population depend on daily incomes with no financial safety net, lives and livelihoods were at stake. It was critical that policymakers had the best evidence available when making hard decisions, especially those related to social distancing and movement restrictions.
Due to the rapidly changing nature of the pandemic, traditional data that Governments and UNICEF would ordinarily rely on from the Official Statistics of BPS (Badan Pusat Statistik), Indonesia’s National Statistics Office, were either not available or came with delays too great to meaningfully assist policy makers with decision-making.
Recognising the need for real-time evidence to inform policymaking, UNICEF Indonesia’s Data & Analytics team, worked closely with colleagues from UNICEF Headquarters, the UNICEF East Asia and Pacific Regional Office, and in-country partners at the University of Indonesia (UI). Together they created ‘COVID-19 Mobility Insights’, a real-time big data and artificial intelligence platform to allow policy makers and citizens to understand the levels of physical distancing, movement and mobility at the village level, in real time.
The development for this real-time big data and artificial intelligence platform was made possible thanks to the cooperation with Meta’s Data for Good team.
How we did this
The data platform was built using anonymised, aggregated mobile phone data from users of our partner services, Cuebiq and Facebook. In a nutshell, it told decision-makers where physical distancing was working, how it related to the rate of transmission, and what the impact was on the most vulnerable.
To do so, the platform relies on anonymized and aggregated data of Facebook users’ movements and uses high-resolution density maps to reliably estimate the total population that stays at home. Using Cuebiq’s data, we calculated people’s ordinary night-time location (where they spend most of their time between the hours of 8pm and 5am). Then, by observing their mobile locations during the day, we calculated how much time they spent at home each day. Using this, we were able to provide granular, real-time data about the level of compliance to the restrictions, at all administrative levels.
Working with our colleagues at the University of Indonesia, we established a strong relationship between the levels of stay at home and a reduction in case rates. According to analysis by UNICEF and UI, in an area such as Jakarta a 1% increase in stay at home could save up to 500 lives.
This real-time data from the COVID Mobility Insights Platform was shared daily with the office of the President of Indonesia and the Governor of Jakarta, becoming a key decision-making support tool in the reintroduction, targeting, and lifting of large-scale physical distancing restrictions.

“The big-data stay-at-home measures were powerfully associated with transmission” says Paul Pronyk, the former Chief of Child Survival and Development for UNICEF Indonesia who oversaw this initiative.
“What was even more important is that these figures were both real-time and predictive. For example, if policy makers tightened restrictions one day, we could observe changes in mobility instantly. If greater numbers of people stayed at home, we could expect lower caseloads in 7-10 days. In the absence of a vaccine, this precision made a huge difference.”

There are 24.79 Million people living in poverty across Indonesia. By utilising data provided by TNP2K and BPS, we were able to answer the critical question: Is the ability to physical distance correlated with income or poverty? The answer is: Yes. Our analysis suggests that regions with the highest reported levels of poverty also have the lowest observed levels of physical distancing.
Paving the way
The technology upon which the platform is based has since expanded to a dozen other countries in the region and beyond., It has been used by UNICEF Country Offices, as well as piloted and tested in emergencies such as Typhoon Odette in the Philippines in 2021, mainly to track population displacement, with unprecedented accuracy and swiftness.
“This data has so much value for capturing the displacement of particularly self-evacuees, which we traditionally have been very bad at tracking. When we have insight into how people are being displaced beyond IDP or other camps, we are better able to preposition staff and supplies, and we become better at offering the help needed where people actually are.” says Anthony Mockler, a data scientist in UNICEF East Asia and the Pacific's Frontier Data Tech Node.
Another key benefit is that data can become available just a few hours after a crisis hits. The first 72 hours of an emergency are the most crucial: emergency response actors must make critical decisions during that timeframe regarding staff and supplies, and usually without a lot of data to work on. They often act in an information vacuum, and base much of their decisions on assumptions.
But with this technology, they now have more information from the ground and in quasi-real time, taking guesswork out of the equation. Exploring frontier data sets such as the Mobility Insights Platform is a real opportunity to improve the quality of decision-making. Plus, they are easily replicable and scalable.
What does the future hold
But what about representation? Because data is collected through Facebook, a valid question arises around who is being surveyed versus who isn’t. The most vulnerable groups in society are not necessarily active on social media, or even own a smart phone.
“A traditional survey instrument deployed by UNICEF reaches up to 10 thousand people. In Asia, Facebook penetration is 60%, so we can survey hundreds of millions of people every day. Are the most vulnerable represented compared to other survey methods? We can definitely improve on that. However, these are larger, better, more complete data sets than any we could hope to gather. The scope and reach are so huge, we can likely get better information or data than we’ve ever had access to before, and it will likely be just as relevant to the most vulnerable. But ensuring the vulnerable population groups are meaningfully represented is something that we continue to work on. We are adjusting the methodology to accurately reflect the most vulnerable and have academic work under way to better understand where it is least and most representative.”
Thanks to Indonesia’s high levels of digital opportunity, UNICEF Indonesia has positioned itself as an organisation with advanced Machine Learning and Data Analytics capacity, and a long history of leveraging technology for public good. UNICEF is uniquely placed to undertake this transformative big-data work, providing levels of insight that would have previously required extensive funds to gain.
Thanks to the aggregated and anonymized mobility data provided by our partners at Cuebiq and Meta’s Data for Good team, UNICEF were able to materially assist in the formation of evidence-based policy, ensuring a lower disease burden and a brighter future for Muri, Kinara, and every child.
UNICEF Indonesia wishes to express its sincere gratitude to the key donors that have contributed to this work, including Meta, Nokia through the Finnish Committee for UNICEF and the COVID-19 Solidarity Response Fund.