The Humanitarian Action Challenge – a joint effort by HumanityX, The Hague Humanity Hub, and ImpactCity – ran from October through December 2018. It encouraged multi-disciplinary partnerships to deliver innovative technology-driven solutions to some of the most pressing problems faced by the humanitarian community today. Non-profit Elva and startup Notilyze teamed up to tackle one of the challenges. In their guest contribution below, they share their story.
Managers of refugee camps have one of the toughest jobs in the humanitarian sector. Working with small, overstretched teams and barely sufficient budgets, they often carry the responsibility for the wellbeing and safety of thousands of individuals. With the residents of their camps relying on a steady inflow of tents, clothes, foodstuff, medication, and much more, supply management is one of their key tasks. It is also one of the hardest. As camp populations are always in flux, it is near to impossible for a camp manager, or any mere mortal for that matter, to accurately predict how many supplies of any kind will be needed in any given month.
As a result, basic supplies sometimes go completely out of stock, leaving thousands staring at empty shelves. On other occasions, warehouses are stocked with supplies that no one needs. This of course leads to hefty storage bills, while some products simply expire on the shelves. To tackle this crucial problem, camp managers need smart forecasting tools, helping them to predict demand faster and more accurately.
Optimizing supply flows through machine learning
Our prototype is a first step in that direction. As part of the Humanitarian Action Challenge, and with the support of the International Organization for Migration, Elva and Notilyze have developed Camp Forecast, a machine learning tool that needs only seconds to forecast the demand for humanitarian supplies within a refugee camp – less time than it takes a sleep-deprived camp manager to jot down a rough calculation on the back of a napkin, while yielding better results. Through optimizing their supply flows, Camp Forecast helps humanitarian organizations to save both money, time, and effort.
Machine learning tools like Camp Forecast hold tremendous promise for the humanitarian sector, as they potentially free up significant amounts of money for addressing humanitarian needs. In 2017 for instance, the total humanitarian aid budget was 27.8 billion USD. Even a tiny efficiency improvement of 0.1% would therefore free up 27.18 million USD to be utilized to have a better effect.
A true cooperative effort
Getting from idea to prototype in just three months was no easy task. It would not have been possible without the support and advice we received from the Action Challenge organizers and from IOM staffers. In this sense, the project is truly a cooperative effort. Following their advice, the team kept the Camp Forecast prototype simple, ensuring that camp managers won’t need to learn any new skills or collect any new data to use it. It currently runs on already existing data on camp populations, collected from IOM’s Displacement Tracking Matrix.
Camp Forecast keeps learning over time, and will therefore give increasingly accurate forecasts. Although we might never achieve the perfect forecast, Camp Forecast will hopefully take some of the pressure off of camp managers and give them more time to spend on other tasks, perhaps even including the rare occasional work break.
The beta version of the tool is now up at http://campforecast.elva.org. Beta testers from different humanitarian organizations can apply to pilot the tool in their camps by sending an email to firstname.lastname@example.org.