Benchmarking: An International Journal, v. 25, n. 8, pp. 2723-2759. Abstract Purpose: The purpose of this paper is to assess the efficiency of agricultural production in South Africa from 1970 to 2014, using an integrated two-stage fuzzy approach. Design/methodology/approach: Fuzzy technique for order preference by similarity to ideal solution is used to assess the relative efficiency of agriculture in South Africa over the course of the years in the first stage. In the second stage, fuzzy regressions based on different rule-based systems are used to predict the impact of socio-economic and demographic variables on agricultural efficiency. They are compared with the bootstrapped truncated regressions with conditional α levels proposed in Wanke et al. (2016a). Findings: The results show that the fuzzy efficiency estimates ranged from 0.40 to 0.68 implying inefficiency in South African agriculture. The results further reveal that research and development, land quality, health expenditure–population growth ratio have a significant, positive impact on efficiency levels, besides the GINI index. In terms of accuracy, fuzzy regressions outperformed the bootstrapped truncated regressions with conditional α levels proposed in Wanke et al. (2015). Practical implications: Policies to increase social expenditure especially in terms of health and hence productivity should be prioritized. Also policies aimed at conserving the environment and hence the quality of land is needed. Originality/value: The paper is original and has not been previously published elsewhere.