How do they pay as they go?: Learning payment patterns from solar home system users data in Rwanda and Kenya

Abstract

Pay-as-you-go (PAYGo) financing models play a vital role in boosting the distribution of solar-home-systems (SHSs) to electrify rural Sub-Saharan Africa. This financing model improves the affordability of SHSs by supporting the payment flexibility required in these contexts. Such flexibility comes at a cost, and yet the assumptions that guide the PAYGo model design remain largely untested. To close the gap, this paper proposes a methodology based on unsupervised machine learning algorithms to analyse the payment records of over 32,000 Rwandan and 25,000 Kenyan SHS users from Bboxx Ltd., and in so doing gain detailed insights into users' payment behavioural patterns. More precisely, the method first applies three clustering algorithms to automatically learn the main payment behavioural groups in each country separately; it then determines the preferred customer segmentation through a validation procedure which combines quantitative and qualitative insights. The results highlight six behavioural groups in Rwanda and four in Kenya; however, several parallels can be made between the two country profiles. These groups highlight the diversity of payment patterns found in the PAYGo model. Further analysis of their payment performance suggests that a one-size-fits-all approach leads to inefficiencies and that tailored plans should be considered to effectively cater to all SHS users.