skip to primary navigationskip to content

MPhil in Engineering for Sustainable Development

global challenges, engineering solutions

Studying at Cambridge

David Bowly

Pay As You Go Solar Home Systems in developing countries: Using Data Mining to Improve Financial Sustainablity

There is huge demand for Solar Home Systems (SHS) across Africa, and social and private enterprise are stepping in to meet the opportunity. Pay As You Go (PAYG) finance models overcome a major affordability barrier, with the weekly payments making solar a viable substitute for the incumbent light source, kerosene. The major challenge for SHS businesses is not finding new customers - it is getting a high enough proportion of customers finishing their contracts (high ‘utilisation’) to keep business sustainable. Notwithstanding this challenge, PAYG has an untapped advantage: businesses collect a large amount of data on their customers.

This research aims to show how data mining can be used effectively and afford- ably by these businesses. The first question: how do we identify the problems that an individual customer is having with the product and intervene? Secondly: are there any patterns across groups of customers which can give us insight into the success factors for these businesses?

Firstly, a machine learning algorithm is developed to identify sudden changes in customer behaviour, which often indicate that a problem is developing. ‘At risk customers’ are flagged to the distributor, and a follow-up phone call is made to attempt to solve the problem. A field test of this system indicates that it is effective in identifying problems, but incentives must be aligned with agents to have the problems resolved.
Secondly, code to efficiently calculate and plot average utilisation of a group is developed to visualise patterns across groups of customers. The various uses of this code are demonstrated to show trends in Azuri’s customer base. Environmental and operational factors such as geographic distribution and number of customers per distributor are investigated to determine whether any correlation exists with customer success.
All research is performed on open source software packages, making these anal- yses affordable for developing world businesses. The use of data mining has the potential to significantly improve the sustainability of these businesses, and hence their impact across the developing world.