Raising Retail Productivity through Data Pooling: Preparatory and Pilot Work to Design a Mobile App Intervention in Zambia

The rise of the information economy has highlighted how data may be an input into production. Since firms also accumulate data through production, large firms may be able to accumulate data more quickly, giving them an advantage. The advantage is especially stark in retail, where a key dimension of productivity is the ability to forecast demand for individual products. However, accumulating data through scale is challenging in developing countries because retail is fragmented across thousands of small single-establishment shops. Thus this study aims to answer two research questions: first, does a lack of information about potential demand constrain the growth of small shops in developing countries? Second, can a scalable technology that pools data generated by individual shops overcome the informational barrier to growth? 

This exploratory study aims to design and develop an algorithm that will make profitable recommendations to shops in Lusaka, Zambia. This will be done in four steps: 1) Digitising pictures of shop inventories and classifying products from a sample of retailers; 2) Clustering shops by similar characteristics and identifying niche products within each cluster; 3) Collecting product-level data on niche products and identifying profitable products; 4) Running a pilot RCT of the algorithm’s recommendations to treated shops. The impact of the intervention will be measured by sending mystery shoppers to observe whether the recommendations are adopted, and by collecting an endline survey on revenue and profit. After this exploratory phase, the broader project will answer the research questions by designing and testing a user-friendly mobile phone app that incorporates the algorithm with a more developed experiment. Shop owners will input data on their own products and sales and receive recommendations on new products that might be profitable to stock. The recommendations will be made using information aggregated from other users of the app. 

This research is highly policy relevant because improving the productivity of micro, small and medium enterprises (MSMEs), many of which are in the retail sector, has been a top priority across the developing world. This study will identify whether imperfect information about product-level sales is a key constraint on managerial ability and firm growth. Assuming imperfect information is a constraint, the study will design a scalable mobile-based technology to help firms alleviate the constraint collectively. In particular, government agencies could adopt the maintenance and promotion of the app.


Ajay Shenoy

University of California, Santa Cruz

Jie Bai

Harvard University

David Sungho Park

KDI School of Public Policy and Management