Intermediate codes have been shown to facilitate iterative decoding convergence to the maximum-likelihood (ML) error ratio performance in serially concatenated schemes. In this paper, we propose a novel block-based intermediate code as an alternative to classic convolutional intermediate codes. Because it is block-based, our intermediate code facilitates practical implementations with reduced memory and processing requirements. Furthermore, we demonstrate that it is simpler to analyze and optimize the iterative decoding process when our block-based intermediate code is employed, instead of a convolutional intermediate code. Finally, we demonstrate that the proposed block-based intermediate code facilitates significantly reduced error ratios in practical schemes when employing short transmission frames and a limited decoding complexity.