Performance analysis of supermarkets: a combined approach using DEA and linear regression

This study aims to analyse the operational efficiency of Brazilian supermarkets through the application of Data Envelopment Analysis (DEA). The research is based on the premise that the lack of knowledge about efficiency among managers may compromise the performance and profitability of the sector, especially during periods of economic crisis. The final sample comprised 673 supermarkets with only one store, thereby avoiding distortions related to multi-store groups. The DEA model employed was the output-oriented SBM, with three input variables (check-outs, sales area, employees) and 1 output (revenue). The EMS software was used for DEA, while JASP was employed for statistical analyses such as Pearson's correlation and linear regression, and Python for bootstrap and cross-efficiency analyses. The results show that only sixteen supermarkets were classified as efficient, while 657 exhibited relative inefficiency. The variables that demonstrated the highest correlation with revenue were number of employees (0.83), number of check-outs (0.82), and sales area (0.91). Based on these findings, specific formulas were developed for measuring efficiency per checkout, per square metre, and per employee, thereby enabling proportional analyses even beyond the original sample. The research provides a practical contribution to supermarket managers by presenting DEA as a strategic tool for efficiency measurement, benchmark identification, and resource optimisation. In addition to documenting efficiency patterns for 673 single-store supermarkets, we introduce Frontier-Normalised Partial Productivity Indices (EFIC)—transparent indices that compare revenue per checkout, per square metre, and per employee against empirically derived efficient baselines. We demonstrate that EFIC closely aligns with DEA efficiency scores.