Predictive Supply Chain for Vaccines no title.jpg

How to deliver the right vaccine, in the right amount, at the right time.

The majority of vaccine stockouts in low and middle-income countries are the result of inaccurate forecasts (World Health Organization, 2017). A stockout is more than an empty shelf; a stockout represents a child who will not receive vaccination. A stockout is a life; a child unable to be vaccinated due to a system failure.

With no alternative in view, ministries of health have accepted double-digit rates of vaccine wastage in the hope of increasing coverage. Supply chains can do better. Supply chains can machine learn to anticipate shifts in consumption. Stockouts and wastage increase the financial burden of health ministries. Reducing stockouts and cutting wastage means saving lives.

macro-eyes was awarded Global Grand Challenges funding from the Bill & Melinda Gates Foundation and USAID to design and pilot the first predictive supply chain for vaccines. The goal: maximize childhood vaccination coverage and minimize vaccine wastage by accurately predicting utilization in the month ahead, translated into vaccine deliveries that anticipate the number of children who will arrive at sites to be vaccinated.


A more precise supply chain is a more equitable supply chain. The macro-eyes innovation is three-fold: enable the first highly-responsive, anticipatory supply chain for health; put into use a machine learning model that embraces multidimensional data – leveraging non-vaccine data from satellite imagery to mobile-phone usage statistics – to better capture the rich nuance of demand, and mitigate the risk of relying on any single data element; demonstrate that the time for putting data to work is now. We don’t need to wait for perfect data and ideal data infrastructure, we are ready to save lives now.

vaccine - 1 month supply & demand data

Existing models for determining appropriate vaccine supply are either static, assuming that tomorrow will look like today or vaccine supply-chains rush to restock depleted supply, working at great cost to address problems that are tragically past. Backwards-facing supply chains do not prepare for the supply needs of the weeks ahead. In work in Tanzania with PATH, the Tanzania Ministry of Health, and district leadership, macro-eyes has demonstrated the ability to improve by 70% on the best model for vaccine forecasting on the market. Put another way, macro-eyes enables the Tanzania Ministry of Health to see into the future (and improve it) with 70% sharper vision.

Vaccine Supply Chain Data Fig 1 842x630.png

Current models for forecasting vaccine supply do not anticipate change, at great cost to life. In all countries – and particularly in low-income countries – change is a constant: changes in demand, changes in access, changes in trust, changes in the movement of populations. Systems that incorporate high-performance machine learning can detect the early signs of change across thousands of features. Being able to more accurately predict the future means being able to more effectively shape that future.