Inventory Management for Mobile Money Agents in the Developing World

Abstract

Mobile money systems, platforms built and managed by mobile network operators to allow money to be stored as digital currency, have burgeoned in the developing world as a mechanism to transfer money electronically.Mobile money agents exchange cash for electronic value and vice versa, forming the backbone of an emerging electronic currency ecosystem that has potential to connect millions of poor and "unbanked" people to the formal financial system. Unfortunately, low service levels due to agent inventory management are a major impediment to the further development of these ecosystems. This paper describes models for the agent's inventory problem, unique in that sales of electronic value (cash) correspond to an equivalent increase in inventory of cash (electronic value). This paper presents a base inventory model and an analytical heuristic that are used to determine optimal stocking levels for cash and electronic value given an agent's historical demand. When tested with a large sample of transaction-level data provided by an East African mobile operator, both the base model and the heuristic improved agent profitability by reducing inventory costs (defined here as the sum of stockout losses and cost of capital associated with holding inventory). The heuristic increased estimated agent profits by 15% relative to profits realized through agents actual decisions, while also offering substantial computational advantages relative to the base model.

The Role of Inclusive Innovation in Promoting Social Sustainability

“…These applications (e.g., M Pesa) have created financial inclusion for large (excluded) communities all over the world (Balasubramanian et al. 2017) without them being directly involved in the ideation phase.…”

“…Mobile money applications are results of "technology push," that is, the solution came prior to the identification of the problem. These applications (e.g., M Pesa) have created financial inclusion for large (excluded) communities all over the world (Balasubramanian et al 2017) without them being directly involved in the ideation phase.…”

The impact of poverty on base of the pyramid operations: Evidence from mobile money in Africa

“…The mathematical intricacies of the mobile money inventory problem, including passive replenishment, are explored in detail in Balasubramanian et al (2022), which characterizes the agent's inventory problem and then develops and analyzes a heuristic policy to guide cash and e-float inventory decisions. Acimovic et al (2022) conducted a field experiment using the algorithm developed in Balasubramanian et al (2022) to explore questions surrounding decision support and communication to agents. However, neither paper considers how poverty might affect an agent's inventory decisionwhich is the focus here.…”

“…This would be true for agents that set inventory levels once before the start of each day (i.e., agents for which a single period inventory model would be appropriate), and for agents that have the opportunity to replenish inventory throughout the day. In the single period case (i.e., a “newsvendor” setting originally described in Edgeworth (1888), formalized in Arrow et al (1951), and extended to mobile money in Balasubramanian et al (2022)), given agents' commission‐to‐cost‐of‐capital ratios, cash and e‐float would be considered “high margin goods.” That is, they would have critical fractiles in excess of 0.5. Therefore, as demand volatility increases, the profit‐maximizing inventory level for each currency would also increase.…”

“…In this way, a mobile money agent experiences increases in inventory of cash (e‐float) when they fulfill demand for e‐float (cash). We refer to this as “passive replenishment.” The mathematical intricacies of the mobile money inventory problem, including passive replenishment, are explored in detail in Balasubramanian et al (2022), which characterizes the agent's inventory problem and then develops and analyzes a heuristic policy to guide cash and e‐float inventory decisions. Acimovic et al (2022) conducted a field experiment using the algorithm developed in Balasubramanian et al (2022) to explore questions surrounding decision support and communication to agents.…”