The convexity axiom in conventional DEA models requires non-increasing marginal product (Banker et al., 1984, BCC). Banker and Maindiratta (1986, BM) suggest a DEA model with log-transformed input and output values that allow for both increasing and decreasing marginal products. By using simulated data and the Canadian oil and gas industry data, we document that the BM model outperforms the BCC model in estimating the efficiency frontier when the production function exhibits increasing marginal product. The BM model also reduces biases in the second-stage analysis of efficiency, providing additional insights that are not available from using the BCC model. Our analysis suggests that relatively small O&G companies may forego efficiency gains by not scaling up their businesses and that using the BM model is desirable when the production function is believed to exhibit increasing marginal product.