Predictive Analytics, Big Data, Machine Learning, Modeling and Simulation…these now commonplace terms in commercial sector supply chain management illustrate how technology is enabling companies to capture, process, and use larger amounts of timely data on a daily basis. Meanwhile, in many low- and middle-income country (LMIC) public health systems, quantification of health commodities—determining the total quantities of products that will be dispensed to users across a country and be needed to maintain inventory levels, and when they need to arrive in the country—is a once-a-year exercise. New data for monitoring and adjusting plans may only become available quarterly and are too often of questionable quality.
Even though supply chain managers in LMICs still face persistent challenges accessing reliable logistics data, many countries are forging ahead in rethinking information flows and computerizing data capture, transmission, and analysis. So what can we learn from the way functions analogous to forecasting and supply planning have evolved in the commercial sector as more and better data become available? As information systems, data, and processing possibilities improve, what changes about these functions?
Certainly, we see the potential for increased sophistication in trying to predict the future when more and better, more timely data become available in the public health supply chain space. But one thing that doesn’t change is human involvement. People, processes, and technology are required to manage complex systems. Human judgment and assumptions inform what the computers do, and people are the ones that take actions based on the outputs of any analysis. If a forecast and supply plan review is only possible monthly or quarterly in the countries where we work, its analog may be a daily (or even more frequent) automated activity in settings where new, trusted data are routinely available.
Indeed, what we in the public health space call forecasting may in the commercial sector actually be handled by software that constantly compares demand signals and planned promotions to current inventories, decides on optimal ordering patterns, and sends recommended orders to human buyers. Similarly, more and better data could strengthen the ability of supply chain and program managers in LMICs to apply greater sophistication to efforts to align supply and demand. Better information enables all players to coordinate activities and activate levers on both the supply provision and demand creation sides, for instance, to capture demand signals and communicate them to suppliers, as well as share supply information back to programs.
There is and will always be a human element to estimating commodity needs. More and more interesting ways are being proposed to use data, but no matter how complex the computer modeling, no matter how well-automated the processes, no matter how ingenious the machine learning, a computer can’t predict when a medicine’s dispensing protocol changes, or anticipate how quickly or completely new product or regimen uptake will occur. What we have learned over the years, beyond the importance of data visibility and analysis, is that human practice controls medicine dispensing and ultimately demand. Computers can’t, in actuality, do all the work.
Thanks to Noel Watson, Joe McCord, and Chris Wright for sharing their insights and expertise in the creation of this post.